The clinical effectiveness and cost-effectiveness of primary stroke prevention in children with sickle cell disease: a systematic review and economic evaluation

Sickle cell disease

Sickle cell disease is one of the most common severe monogenic disorders in the world. 5 It is now the most common genetic condition in the UK; the incidence rate is estimated to be 1 in 2000 live births. 6 Rates are higher in some urban areas, affecting 1 in 300 live births. 2 There are approximately 5230 children aged < 16 years with SCD in the UK. As SCD is primarily found in black ethnicities (predominantly from sub-Saharan Africa), there is a very unequal geographic distribution of SCD in the UK; the highest density of the affected population is located in inner-city areas, where there is a high proportion of ethnic minority populations. 7 It is thought that 75% of children with SCD in the UK live in or around London (A Streetly, NHS Sickle Cell and Thalassaemia Screening Programme, and D Rees, School of Medicine, King’s College London, London, UK, 2011, personal communication). Streetly et al. 8 estimated the prevalence of SCD in the UK to be approximately 3 per 1000 children (1 : 330 babies with a positive result) in south-east London in comparison with 0.12 per 1000 children (1 : 8333 infants with positive result) in Cumbria and in Lancashire. Babies reported as Black African make up 4% of total births yet represent 61% of all suspected cases of SCD. 8 Carrier rates in Black African babies are 145 per 1000 children (1 : 7), in comparison with 1.85 per 1000 children reported as being White British (1 : 540). 8

Complications of sickle cell disease

Sickle cell disease is associated with a number of serious complications, including acute pain, splenic sequestration and acute chest syndrome.

Childhood stroke and sickle cell disease

For children with SCD, the risk of stroke is estimated to be 300 times higher than in the general childhood population. 23,24 Without screening, up to 10% of children with SCD suffer stroke, usually ischaemic. 25 A further 17–25% of patients suffer often unnoticed ‘silent infarctions’, resulting in neurological disability and damage. 26 Ischaemic strokes are characterised by slurred speech, weakness in limbs, seizures, coma and cognitive impairments. The most common presentation of stroke is acute hemiplegia (the inability to move, experienced on one side of the body). Recovery from stroke varies across children; functionality may be recovered over time. However, 50% of children who experience a stroke are likely to have remaining disability and 18% of these children will be severely disabled. 24 An underappreciated outcome of stroke in childhood is its association with serious intellectual and cognitive deficits; attention, memory and executive function may all be affected. 24 Once individuals have suffered a primary stroke, they have a 30–75% risk of a further (secondary) stroke if not receiving blood transfusions, and these strokes are associated with significant mortality and morbidity. 25

The risks of stroke for patients with SCD are thought to differ across the course of childhood. Data from the USA indicate that the childhood incidence of stroke in those with SCD is 1.02 per 100 patient-years in children aged between 2 and 5 years, and 0.79 per 100 patient-years in children aged between 6 and 9 years,27 with stroke in all individuals with SCD averaging 0.61 per 100 patient-years.

The Baltimore–Washington Cooperative Young Stroke Study23 identified all children aged 1–14 years in Maryland and Washington DC with a diagnosis of ischaemic stroke and intracerebral haemorrhage between 1988 and 1991. They estimated the incidence of stroke among children with SCD to be 0.28%, or 285 per 100,000 children with SCD per year. Stroke incidence in children without SCD has been estimated at 2.3 per 100,000 children per year. 28 Quinn and Miller29 calculated that by 18 years of age 11% of children with SCD will have suffered a clinically overt stroke and a further 20% will have a clinically ‘silent’ stroke. Data for the UK on stroke rates in children with SCD are not readily available but there is no reason to anticipate that they are significantly different to rates reported from the USA. A longitudinal study by Telfer et al. 30 in the UK followed a neonatal cohort of 252 children with SCD from 1983 to 2005 and used Kaplan–Meier techniques to estimate the risk of developing abnormal TCD scores in a cohort of children followed from birth to the age of 16 years. They found the incidence of first stroke to be 0.3 per 100 patient-years. Estimated risk of stroke was 0.7% per 100 patient-years at age 5 years, 2.7% at age 10 years, 4.3% at age 15 years and 12.8% at age 20 years. The majority of these patients had been screened with TCD ultrasonography and other modalities and, during the course of the study, primary stroke prophylaxis had been implemented.

Regular blood transfusion

The primary prevention strategy for stroke resulting from SCD in both adults and children is regular blood transfusion, although the means by which transfusion prevents stroke is unknown. 35 The standard therapeutic goal of regular blood transfusion is to reduce the HbS to < 30% of the total haemoglobin36 and to maintain a haemoglobin level of > 9 g/dl.

Blood transfusions can be delivered using different methods. These include transfusion of packed red blood cells every 3–4 weeks and exchange transfusion (by hand or automated apheresis) every 4–8 weeks. Following 3 years of blood transfusion therapy, maintenance of HbS at < 50% may then be sufficient to prevent future stroke,31 although there is no direct evidence to support this.

Data are lacking regarding the exact number of children with SCD who are currently receiving blood transfusions in the UK. However, cohort data suggest that between 3.6% and 6.7% of children with SCD receive prophylactic blood transfusion for primary stroke prevention each year (M Roberts-Harewood, 2011, personal communication). 30 Data from a large UK centre suggest that 9% of children with HbSS receive prophylactic blood transfusion for primary stroke prevention each year (P Telfer, Barts and the London School of Medicine and Dentistry, London, UK, 2011, personal communication). The paediatric peer review programme suggests that between 1 : 30 and 1 : 10 children receive regular blood transfusion for all causes, of which primary stroke prevention in children with SCD is the most common (A Yardumian, North Middlesex University Hospital, UK, 2011, personal communication).

Blood transfusion is time consuming and regular transfusion is required to reduce and maintain the target levels of HbS. 36 There are also significant risks associated with chronic blood transfusion, including iron overload. Adverse events (AEs) associated with transfusion include alloimmunisation (development of antibodies to foreign red blood cells),37 risk of transfusion-transmitted infections [such as human immunodeficiency virus (HIV), hepatitis B, prion disease or hepatitis C (frequently occurring in developing countries where the rate of SCD is higher)] and haemolytic transfusion reactions. The risks of these events increase over time and must be carefully considered before undertaking a regimen of chronic blood transfusion.

It is estimated that approximately 67% of children who have a first overt stroke will have further overt strokes without transfusion therapy. 4 However, it is likely that between 17.5% and 20% of children will suffer a second overt stroke, despite receiving regular blood transfusion therapy. 38,39

Hydroxycarbamide

Data from non-randomised clinical studies suggest that hydroxycarbamide might be an alternative to transfusion for primary stroke prevention56 and might reduce the risk of stroke in children with SCD. The Royal College of Physicians recommends that children with SCD who cannot receive blood transfusion because of alloimmunisation, autoantibody formation or non-compliance with transfusion or chelation may be considered for treatment with hydroxycarbamide. Hydroxycarbamide increases the concentration of fetal haemoglobin, and is licensed as a treatment to reduce painful crises in patients with SCD. The number of children in the UK currently receiving treatment with hydroxycarbamide for primary stroke prevention is unknown.

A recent systematic review57 considered all published literature on the efficacy, effectiveness and toxicity of hydroxycarbamide in children with SCD, and found an increase in fetal haemoglobin from 5–10% to 15–20% in those children treated with hydroxycarbamide. Haemoglobin concentration increased modestly (1 g/l) but significantly across studies. Treatment with hydroxycarbamide also decreased hospitalisation rates from 87% to 56%. A small study58 has reported on the impact of hydroxycarbamide on the TCD blood velocities of 59 children with SCD and shows that the magnitude of TCD velocity decline was significantly correlated with the maximal baseline TCD value. Recently published data from the BABY HUG trial59 found a significantly lower average increase in TCD velocity in children aged between 9 months and 18 months receiving hydroxycarbamide than in those receiving placebo. The evidence for the use of hydroxycarbamide as a primary stroke prevention strategy is minimal but suggests that hydroxycarbamide may be useful in reducing TCD velocities in children from birth before they become abnormal. Based on the results of the Stroke With Transfusions Changing to Hydroxyurea (SWiTCH; hydroxurea is now known as hydroxcarbamide) trial,60 it is generally accepted that hydroxycarbamide should not be used for secondary stroke prevention.

Bone marrow transplantation

Bone marrow transplantation is reported to stabilise the cerebrovascular disease caused by SCD61 but is not often feasible due to the lack of availability of suitably matched donors. It is estimated that there are only four to five BMTs performed in the UK each year, and < 400 are performed annually worldwide (A Streetly and D Rees, 2011, personal communication). Successful transplantation of non-sickle cell bone marrow cures SCD, and therefore the children receiving BMT are no longer treated or followed up in the same way as children with SCD. This makes estimation of stroke risk in this population difficult.

Search strategy

The search incorporated a number of strategies, combining index terms (for the disease) and free text words for the technologies involved. The search strategies had no language restrictions and did not include methodological filters that would limit results to a specific study design. Details of the search strategies and the number of records retrieved for each search are provided in Appendix 1. All references were exported to an EndNote version 5 (Thomson Reuters, CA, USA) bibliographic database.

The following electronic databases were searched (YD) for relevant published literature for the period 1950 to May 2011:

• CDSR (Cochrane Database of Systematic Reviews)

• CENTRAL (Cochrane Central Register of Controlled Trials)

• DARE (Database of Abstracts of Reviews of Effects)

• EMBASE

• Health Technology Assessment database

• ISI Web of Science Proceedings (Index to Scientific & Technical Proceedings)

• ISI Web of Science Citation Index Expanded (SCIE)

• MEDLINE

• NHS EED (NHS Economic Evaluation Database).

Given the specialised nature of this disease, searches of conferences were not carried out; clinical advice suggested that the only published data would be found in the literature describing the existing clinical trials.

Selection of evidence

The records identified by the electronic searches were assessed for inclusion in two stages. Two reviewers (MGC and JG) independently scanned all titles and abstracts identified by the search to ascertain which articles may be relevant to the clinical review. Full-text versions of all records selected during the initial screening process were obtained to permit more detailed assessment. These were then assessed independently by two reviewers (MGC and JG), using the inclusion criteria shown in Table 8. The inclusion/exclusion assessment of each reviewer was recorded on a pre-tested, standardised form. Disagreements were resolved by discussion and, if necessary, another reviewer was consulted. A flow diagram summarising the selection and inclusion of studies is provided in Appendix 2, Figure 13.

Data abstraction

Data extraction for the review of clinical effectiveness was carried out by two reviewers (MGC and JG). Data were abstracted by one reviewer and then checked for accuracy by a second reviewer. Data presented from multiple reports of single trials were extracted as a single record.

Quality assessment

Two reviewers (MGC and JG) independently evaluated the included studies for methodological quality using criteria based on guidance published by CRD. 67 Any discrepancies in quality grading were resolved through discussion.

Methods of data synthesis

Individual study data and quality assessment are summarised in structured tables and as a narrative description. The primary treatment outcome relevant to this review was incidence of stroke. The differences in patient groups between the included trials precluded a statistical synthesis.

Search strategy

A comprehensive review of the literature was undertaken to identify all published economic evaluations of primary stroke prevention in children with SCD identified to be at high risk of stroke by TCD ultrasonography using the main search strategy outlined above (see Identification of evidence: clinical effectiveness, above).

Selection of evidence

During the clinical effectiveness screening, all papers that appeared to include economic data were identified. Full-text copies of these papers were subsequently obtained and two reviewers (MGC and JG) independently assessed them for inclusion, using the inclusion criteria described in Table 8. Any disagreements regarding inclusion of economic studies were resolved by discussion. No relevant economic evaluations were identified for inclusion in this review. A flow diagram summarising the selection and inclusion of studies is provided in Appendix 2, Figure 14.

Number of studies identified and included

A total of 1337 non-duplicate records were identified by the search strategy (see Appendix 1) and subsequently screened for inclusion in the review. No trials were identified that evaluated the efficacy of hydroxycarbamide or BMT as primary stroke prevention strategies. Two RCTs31,35 made comparisons between blood transfusion and standard care and were included in the review. Data from these trials were published in peer-reviewed journals. A number of papers relating to these two RCTs were also identified.

Quality assessment of included trials

The methodological quality of the included trials is summarised in Table 9, using the criteria based on the guidance published by the CRD, which include key aspects of RCT design and quality. 67

Overall, the methodological quality of the included trials was adequate. Both papers state that participants were randomised to treatment and describe the method of randomisation used, but neither trial reported whether, or how, allocation was concealed. The baseline characteristics of patients were reported for both trials and comparability was considered to be partially achieved in STOP31 and achieved in STOP 2. 35 Both trials fully reported their inclusion criteria. In trials of this type, blinding of participants and administrators would be difficult or unethical; however, the administrators of the TCD ultrasonography were blinded to treatment group and the adjudication of suspected strokes was conducted by blinded assessment in both trials. An intention-to-treat (ITT) analysis was reported in STOP31 but not in STOP 2. 35 Both trials reported outcomes for more than 80% of participants originally randomised and patient dropouts were accounted for. There was no evidence that data for any of the outcomes stated at the outset were not reported in the final analyses.

Trial characteristics

The key trial characteristics for the two included RCTs31,35 are presented in Table 10. Both trials31,35 were multicentred and open label. The intervention in both trials31,35 was blood transfusion and the comparator was standard care. Standard care at the time of the trials was defined as no blood transfusion for primary stroke prevention. Both arms also received penicillin prophylaxis, pneumococcal vaccination, folic acid supplementation, surgery and treatment of acute illness, including the use of transfusion when needed for transient episodes but excluding the use of hydroxycarbamide or antisickling agents.

The purpose of STOP31 was to evaluate the use of blood transfusion to prevent a first stroke; the purpose of STOP 235 was to determine whether or not the time on prophylactic transfusion could be limited so that children did not receive blood transfusions continually until the age of 18 years. Patients in STOP31 had not previously received blood transfusions for primary stroke prevention and all participants had at least two abnormal TCD readings prior to entering into the trial. STOP 235 was an extension of STOP31 and a number of STOP31 patients, whose TCD readings had normalised after ≥ 30 months of transfusion, participated in the trial. In addition to STOP31 patients, other children who did not participate in STOP31 but whose condition met the criteria for eligibility (TCD normalised after ≥ 30 months of transfusion) participated in STOP 2. 35 Thus, the patients in STOP 235 were all receiving regular blood transfusion for primary stroke prevention and were required to have had at least two normal TCD readings prior to entry into the trial. It is not clear how many patients in STOP31 were included in STOP 2. 35 It is worth noting that patients aged 2–16 years were eligible to participate in STOP,31 whereas patients aged 5–20 years were eligible to participate in STOP 2. 35 STOP31 was published in 1998 and STOP 235 was published in 2005. Both STOP31 and STOP 235 were small in terms of participant numbers (n = 130 and n = 79, respectively).

The primary outcome measure in STOP31 was stroke (cerebral infarction or intracerebral haematoma/haemorrhage). Focal symptoms consistent with the occurrence of a cerebral infarction or an intracerebral haemorrhage were required unless the presentation suggested a diagnosis of subarachnoid haemorrhage. In the absence of supporting magnetic resonance imaging (MRI) findings, clear and compelling clinical evidence of a stroke was required. Transient symptoms were included if changes consistent with the occurrence of stroke were evident on MRI.

The primary outcome measure in STOP 235 was a composite of stroke (cerebral infarction or intracranial haemorrhage) and/or reversion to an abnormal TCD velocity. Stroke was defined as persistent neurological abnormalities or transient symptoms accompanied by a new cerebral lesion appropriate to the patients’ clinical presentations. Suspected strokes were adjudicated by experts who were blinded as to treatment assignment. Abnormal velocity on TCD scans was defined as two consecutive studies with abnormal velocities, three consecutive scans with an average velocity of ≥ 200 cm/second or three consecutive inadequate studies plus evidence of severe stenosis on magnetic resonance angiography (MRA).

Both STOP31 and STOP 235 were halted prematurely according to a priori criteria; STOP31 was halted due to increased rate of stroke in the non-treatment arm and STOP 235 owing to increased rate of stroke and/or reversion to abnormal TCD scan results following discontinuation of blood transfusion. The mean duration of follow-up was 19.6 months for STOP. 31 The mean duration of follow-up for STOP 235 was not reported.

Participant characteristics

The key characteristics of the patients in STOP31 and STOP 235 are described in Table 11. In STOP31 the majority of the baseline patient characteristics appear to be well balanced between the two arms of the trial. In the published paper31 it is noted that baseline haemoglobin and haematocrit levels were slightly lower in the transfusion arm. Approximately half of the patients were male, with a mean age of 8.2 years (transfusion) and 8.4 years (standard care).

In STOP 235 the majority of the baseline patient characteristics appear to be well balanced between the two arms of the trial, except that there was a greater percentage of male participants in the continued-transfusion arm (53% vs 32%). No significant differences between the two arms of the trial with regard to any of the baseline patient characteristics were noted in the published paper. 35 The mean age of the patients was 12.5 years (transfusion) and 12 years (standard care).

It is clear from Table 11 that there are differences between the two trials. The differences in many of the variables are largely explained by the different trial selection criteria; STOP 235 is a partial follow-on from STOP31 and consists of patients with a history of regular blood transfusion (≥ 30 months) prior to their entry into the trial, and who have no significant cardiovascular disease on MRA. As a result, patients in STOP 235 are older; owing to their history of transfusion, they have greater mean haemoglobin and haematocrit levels, lower mean HbS and fetal haemoglobin levels, and vastly higher levels of serum ferritin than the patients in STOP. 31

Results

Quality of life

Quality-of-life outcomes were neither collected nor reported on in either STOP31 or STOP 2. 35

Adverse events relating to transfusion

In the STOP trial,31 10 patients in the transfusion group developed alloimmunisation to red blood cells. There were 16 mild reactions in 12 patients to blood products.

In the STOP 2 trial,35 one new patient (continued-transfusion arm) was identified with alloimmunisation. Nine transfusion reactions in seven patients were noted. One of these was serious enough to require hospitalisation. Chelation therapy was received by 93% (n = 35) of patients in the continued-transfusion arm and 76% (n = 31) in the transfusion-halted arm.

Degree of disability from stroke

It is reported in the STOP31 trial that the 11 patients with cerebral infarction presented with hemiparesis (six left-sided, five right-sided) but weakness had resolved by the time of the neurological examination. All infarctions were in the carotid circulation and the MRI scan showed new or larger lesions in the affected hemisphere in all but one patient. Of the 11 patients, 10 were hospitalised; at the time of discharge from hospital, two patients were rated as having major disability, five had moderate disability, two had symptoms but no disability, and one was asymptomatic.

In the STOP 235 trial, one of the two patients who had a stroke presented with a right hemisphere infarction. No further details of strokes are reported for this trial.

Non-randomised studies identified

Seven non-randomised studies41, 43–45, 68–70 were identified. The majority were retrospective cohort studies. Of these seven studies,41, 43–45, 68–70 data were able to be extracted from only one68 (Table 13). The remainder considered children with high, conditional and low TCD velocities and did not separately report stroke rates or other data for only the high-risk children. The single-arm study described in Table 13 was based in France (patients treated with transfusion) and included 17 patients aged between 2 and 16 years. The mean length of follow-up was 32.4 months. None of the 17 patients suffered a stroke while receiving blood transfusion.

Key characteristics

The key characteristics of the cost-effectiveness analysis are shown in Table 15.

The model includes a starting population of 2-year-old children with SCD who have not experienced a stroke. The model is built in such a way that a hypothetical cohort of 1000 children is run for two scenarios: those receiving the blood transfusion intervention and those not receiving the intervention when their blood velocity on TCD ultrasonography is > 200 cm/second. Starting at the age of 2 years, in each scenario all 1000 patients are modelled for their lifetime, through both pre- and post-stroke states.

On entering the model, children’s blood flow velocity is measured using TCD ultrasonography. Patients are then screened yearly with TCD ultrasonography within the model in order to reflect the changing probability of developing an abnormal TCD velocity in childhood (as not all children develop abnormal TCD velocities at the age of 2 years). The model reflects real-life practice and the costs and benefits associated with this group of children as a whole (those with normal and abnormal TCD scans) are captured in the analysis.

In both arms of the model children receive TCD scans, in line with UK clinical guidelines for current best practice. 65 However, children who are identified to be at high risk of stroke after TCD ultrasonography receive chronic blood transfusion for primary stroke prevention in the intervention arm of the model only. Children in the non-intervention arm all receive TCD scans (reflecting current clinical practice) but do not receive blood transfusion for primary stroke prevention. Including the costs of scanning in the non-intervention arm may overestimate the costs in this group but it is important to note that the cost of a TCD scan is relatively small in relation to the overall cost of blood transfusion, stroke treatment and treatment of AEs, and is thus unlikely to influence the result of the analysis. This economic evaluation does not consider the added costs associated with implementing and running a TCD screening programme for children with SCD, as it is assumed that most centres across the UK already provide TCD ultrasonography to children with SCD. The cost-effectiveness analysis was undertaken in line with the National Institute for Health and Clinical Excellence (NICE) guidelines. 78 The analysis focuses on the health-care system costs and QoL impact of blood transfusions over the lifetime of the patients modelled.

Data collection

The parameters and sources used in the model are summarised in Appendix 4. Given the multiplicity of data required to populate the model, a range of approaches were used to identify parameter estimates.

First, a systematic review was undertaken to identify parameter estimates reported in the existing literature. Details of the search strategy can be found in Chapter 3 (see Identification of evidence: cost-effectiveness). However, as described in Chapter 3 (see Identification of evidence: clinical effectiveness and Identification of evidence: cost-effectiveness), the clinical data available were limited and no relevant published economic evaluations were identified. Specifically, a paucity of relevant data were found for utility weights associated with paediatric SCD and stroke, cost of paediatric stroke care in the UK, probability of patients’ health state improvement post stroke, and frequency of SCD complications, such as pain crises and acute chest syndrome.

Second, the gaps in the evidence were filled using clinical expert opinion. This opinion was collected by means of electronic questionnaires and telephone interviews. A questionnaire was developed and sent to clinicians. A template of the questionnaire can be found in Appendix 5. From the 17 clinicians who were contacted, four questionnaire responses were received. The responses from clinicians were incorporated into the economic model by taking the average value of responses across the questionnaires. These responses are summarised in Appendix 6, Table 51.

A separate set of questions regarding utility weights for SCD health states was developed for use in a telephone interview. Of the five clinicians contacted for a telephone interview, two experts agreed to participate and provide data. The telephone interview was not formally structured and focused on the discussion of stroke management, stroke outcomes and QoL. Topics included interventions performed by clinicians to manage patients in mild, moderate and severe post-stroke states (see data in Appendix 4, Table 46). The probability of ending up in the different post-stroke states following a first, second or third stroke was estimated. Further details are presented in Appendix 4, Table 41. Discussion of patients’ QoL focused on the impact of blood transfusion, pain crises and stroke. Review of relevant published literature did not provide sufficient information to calculate the utility weights required for economic evaluation. Finally, calibration was performed on the death rates used in the model to ensure the model’s predicted death rates fitted with those from validated sources. 16,40 The death rates were adjusted while ensuring that death rates were kept within the bounds of the ranges specified by experts in the surveys and interviews conducted by the assessment group, clinical experts in their responses to the surveys and interviews developed by the Liverpool Reviews and Implementation Group (LRiG) and relativities between other transition probabilities were maintained. Once death rates were updated, alternative routes through the model were also updated. These updates were made to ensure that the ratio between the original transition probabilities did not change.

Additionally, calibration was used to adjust rates of transition between second and third post-stroke health states in order to obtain survival estimates predicted by the published literature. 16,40

Model structure

Following best practice, a Markov structure was adopted to represent the progression of patients with SCD and stroke. 55,79 Markov structures are used when the probability of an event occurring changes over time, when the timing of events is important and when important events may happen more than once, as is the case with changes in blood velocity and stroke. The ability of Markov models to represent such repetitive events and the time dependence of both costs and utilities allows for more accurate representation of cost-effectiveness when modelling conditions that evolve over time or occur at various points in time. 80

A Markov model was built and run separately for two scenarios: scenario 1 (intervention), in a population of 1000 2-year-olds, children whose blood velocity measured by TCD ultrasonography is > 200 cm/second are treated with blood transfusions and children whose blood velocity is < 200 cm/second are not treated, and scenario 2 (comparator), in a population of 1000 2-year-olds, children whose blood velocity measured by TCD ultrasonography is > 200 cm/second or < 200 cm/second are not treated.

The difference between the costs, incidence of health outcomes and QoL outcomes between these two scenarios was used to estimate the incremental costs and effects of blood transfusions. A description of the Markov structure used to model the costs and effects of blood transfusion (intervention arm) is presented in Figure 1. In each scenario the model was run for the lifetime of a hypothetical cohort of 1000 children aged 2 years who have SCD but have no history of stroke. In each instance – with and without blood transfusion – the starting cohort has the same blood transfusion profile. The only difference between the scenarios is the probability of receiving blood transfusion pre-stroke when blood velocity is > 200 cm/second.

FIGURE 1.

Markov model used to estimate the efficiency of blood transfusion for patients with SCD: pre-stroke (intervention arm only).

Figures 1 and 2 show that the model has been divided into two distinct components:

1. Pre-stroke component, which simulates the change in blood velocity and receipt of blood transfusion prior to having a stroke. As the cohort of interest is children aged 2 years with SCD, who have not yet experienced a stroke, all participants start in this part of the model.

2. Post-stroke component, which simulates the transition of patients with SCD who experience a stroke between post-stroke health states – i.e. mild post-stroke health state, moderate post-stroke health state, severe post-stroke health state and death. Once patients have a stroke, they move from the pre-stroke part to the post-stroke part of the model.

FIGURE 2.

Markov model used to estimate the efficiency of blood transfusion for patients with SCD: post stroke (intervention arm only).

Pre-stroke component of the Markov model

This section provides an overview of the pre-stroke component of the model. Details on the data used to specify transition probabilities, costs, and utilities are available in Appendix 4.

The pre-stroke part of the model comprises four different health states, which are shown in Table 16.

The population – 1000 patients aged 2 years, diagnosed with SCD – all start in this part of the model as none of the children has suffered a stroke. At the start of the model, 89.3% of patients are in the ‘no transfusion < 200 cm/second’ health state and 10.7% are in the ‘no transfusion > 200 cm/second’ health state. The population is distributed in this way for both the intervention and non-intervention scenarios. 31 Between the ages of 2 and 18 years, patients can move to the > 200 cm/second state, based on TCD scan result. In the intervention cohort, the population starting in the ‘no transfusion < 200 cm/second’ state can either stay in this state or move to ‘no transfusion > 200 cm/second’ if their blood velocity increases. They do not immediately receive blood transfusion, as a second TCD scan confirming a velocity of > 200 cm/second is required prior to starting blood transfusion (D Rees, 2011, personal communication). The population starting in the ‘no transfusion > 200 cm/second’ state will move into the ‘transfusion > 200 cm/second’ state in the second cycle of the model. Patients can die as a result of either general mortality or non-stroke-related SCD mortality. Table 17 outlines the transition probabilities used to model these dynamics.

It is evident from Table 17 that the number of patients with SCD with high TCD scores in the model increases gradually. That is, the majority of patients (99.8%) who start a cycle with a blood velocity of < 200 cm/second end the cycle with a blood velocity of < 200 cm/second (A Streetly and D Rees, 2011, personal communication). These parameters are based on D Rees’ estimation that between the ages of 2 and 18 years approximately 15% of the starting cohort would have a TCD scan of > 200 cm/second and receive transfusions. Transition probabilities were calibrated to reflect this. Assuming the transition probabilities were constant between the ages of 2 and 18 years, the calibration was performed by adjusting the probability of moving from a blood velocity of < 200 cm/second to a blood velocity of > 200 cm/second until 15% of the alive cohort was in the > 200 cm/second group at the age of 18 years. Given the uncertainty in these estimates, these transition probabilities have been extensively tested through sensitivity analysis (see Appendix 8).

Table 17 also demonstrates that the death rate doubles when a child moves from having a blood velocity of < 200 cm/second to having a blood velocity of > 200 cm/second (from 0.1% per 3-month cycle to 0.2% per 3-month cycle) without receiving blood transfusion. If blood transfusions are administered, the death rate among those with a blood velocity of > 200 cm/second is reduced back to that of those with a blood velocity of < 200 cm/second.

Children whose blood velocity increases to a level of > 200 cm/second remain in this state until the age of 18 years. The assumption was based on clinical opinion stating that, in current UK clinical practice, most patients with SCD who receive transfusions will remain on transfusion until they are adults (D Rees, 2011, personal communication). Appendix 3 contains the full list of assumptions used in the model. Currently, clinical guidelines do not state for how many years patients should stay on transfusion6 but recommend transfusion until at least 16 years in the UK65 and 18 years in the USA. 62 The pre-stroke Markov submodel is run until the age of 18 years. At that point, 75% of those on blood transfusion at 18 years are assumed to continue on blood transfusion for the remainder of their lifetime, and the remainder of the population do not continue on blood transfusion. It is assumed that once the population moves to either the ‘no transfusion’ or ‘transfusion’ health state at 18 years old, they remain in that health state until death.

In the comparator arm, in which blood transfusion is not available, the Markov model is the same, except that the probability of moving to ‘transfusion > 200 cm/second’ is considered to be zero. Therefore, in this scenario there are effectively three health states: (1) ‘no transfusion, TCD scan < 200 cm/second’; (2) ‘no transfusion, TCD scan > 200 cm/second’; and (3) death.

A 3-month cycle length was used to capture changes in all events. The cycle length was selected in order to reflect the dynamics of SCD. Several key events occur within short time frames, such as blood transfusions (every 4 weeks).

Probability of stroke

Each state in the pre-stroke element of the model is associated with a probability of having a stroke. The probability of having a stroke is a function of blood velocity, whether or not a patient is receiving blood transfusion and age. 35 The incidence of stroke for patients with SCD varies significantly with age. 27 Specifically, general clinical opinion is that stroke risk is higher between the ages of 2 and 7 years, and 19 and 30 years. However, published evidence on risk of stroke does not support this view. For the purpose of the model we have used published data and tested these probabilities in scenario analyses. In order to reflect this dynamic in the probability of stroke, the parameters used in the model distinguish between four distinct age groups: age 2–7, 8–18, 19–30 and ≥ 31 years.

Table 23 shows the varying likelihood (by health state and age) that patients with SCD experience a first stroke. In the model, between the ages of 2 and 18 years, individuals with a high blood velocity are assumed to have a significantly higher risk of stroke, and this risk is reduced by blood transfusions. After the age of 18 years, general stroke rates for patients with SCD are drawn from Ohene-Frempong et al. 27 and the probability of stroke is assumed to be the same across health states. However, this report acknowledges that the data supporting these assumptions are uncertain.

On having a stroke, a patient leaves the pre-stroke component of the Markov structure and enters the post-stroke component of the Markov structure.

Post-stroke component of the Markov model

This section describes the structure of the post-stroke components of the Markov model. Further detail on the data used to specify the parameters of the model is available in Appendix 4.

Model dynamics

As with the pre-stroke element of the model, a 3-month cycle length is adopted for the post-stroke element of the model. A number of different types of dynamics are included in this part of the model:

• disease progression following first stroke

• probability of having a second or third stroke

• health outcome immediately after the second or third stroke

• disease progression following second or third stroke.

Disease progression following first stroke

Table 25 describes the probability of being in different health states – mild, moderate, severe, and dead – following a first stroke. Table 26 summarises the probability of subsequent disease state deterioration. It is assumed that those who have had a first stroke will move through these states in sequence. For instance, the population can only move from ‘mild post stroke’ by first going through ‘moderate post stroke’, i.e. the population cannot directly move from ‘mild post stroke’ to ‘severe post stroke’ or vice versa.

TABLE 26 - Transition probabilities following first stroke per 3-month cycle

Parameter	Probability (%) by age group (years)
	2–7	8–18	19–30	31+
Probability of moving to a worse state post stroke	1.00	1.00	6.00	10.00

Probability of having a second or third stroke

After individuals have incurred a first stroke, there is a probability of incurring a second and/or third stroke (Table 27). The process of calibration involves adjustment of all of the second and third stroke probabilities, so that the model predicts death rates reported in the literature. 21,45 Specifically, calibration was conducted on the probability of second and third stroke and the probability of post-stroke health state following the second and third stroke. Originally, the calibration was done so that the relativities between the first stroke and the second/third stroke were maintained, for example the second stroke would be relative to the first stroke by 10%, 20%, 30%, etc., and then the third stroke would be relative to the second stroke by 10%, 20%, 30%, etc. The calibration was required due to the absence of published data on the subject. The decision to model up to three strokes was based on clinical opinion and the absence of data on further strokes. 16,40 Based on clinician opinion, both the probability of a recurrent stroke and the disability associated with post-stroke health state are dependent on the patient’s current stroke health state.

TABLE 27 - Probability of second and third stroke per 3-month cycle

Parameters	Probability (%) by age group (years)
	2–7	8–18	19–30	31+
Probability of second stroke based on first stroke health state
Mild state post first stroke	0.25	0.25	0.25	0.00
Moderate state post first stroke	0.50	0.38	0.38	0.00
Severe state post first stroke	1.25	0.63	0.63	0.63
Probability of third stroke based on second stroke health state
Mild state post second stroke	1.25	0.50	0.50	0.00
Moderate state post second stroke	1.25	1.00	1.00	0.00
Severe state post second stroke	2.50	1.25	1.25	0.00

Health outcome immediately after the second or third stroke

Table 28 presents the health outcomes post second and third strokes. It is assumed that following another stroke the population can move only to a health state the same as or worse than the one they were in before the stroke. For instance, the population in a moderate post-stroke state after an initial stroke can move only to a moderate or severe post-stroke state following a second stroke. This assumption is based on information from clinicians obtained via telephone interviews.

TABLE 28 - Health state post second and third strokes

Starting health state	Ending health state	Probability (%) by age group (years)
		2–7	8–18	19–30	31+
Mild post stroke	Mild	80	80	80	80
	Moderate	18	18	18	18
	Severe	2	2	2	2
	Dead	0	0	0	0
Moderate post stroke	Moderate	45	45	45	45
	Severe	50	50	47	47
	Dead	5	5	8	8
Severe post stroke	Severe	90	90	85	85
	Dead	10	10	15	15

Data for these parameters have been originally collected through the clinicians’ telephone interviews and then calibrated using the existing literature. 16,40

Disease progression following second or third stroke

As with the first stroke, once the second or third stroke has occurred and individuals are in one of the four post-stroke health states, there is a probability of staying within the same health state or moving to a worse health state over time. Table 29 summarises how health states progress following the second and third strokes.

TABLE 29 - Transition probabilities for second and third stroke Markov per 3-month cycle

Parameter	Probability (%) by age group (years)
	2–7	8–18	19–30	31+
Probability of moving to a worse state post second stroke	15	15	15	25
Probability of moving to a worse state post third stroke	20	20	20	30

Predicted population outcomes

Figures 3 and 4 show population outcomes under the intervention and the non-intervention arms. They demonstrate how the population in the intervention arm moves on to transfusion early in life owing to high blood flow velocities and how the population in the non-intervention arm moves on to transfusion only post stroke. Also it is clear that a larger percentage of the population lives longer in the intervention arm in comparison with patients in the non-intervention arm.

FIGURE 3.

Population outcomes with blood transfusion following TCD scans.

FIGURE 4.

Population outcomes without blood transfusion following TCD scans.

Blood transfusions

Patients receive blood transfusions following a stroke regardless of post-stroke health state. These transfusions are associated with a number of health benefits. Patients also receive relevant treatments for stroke-related disabilities. Based on communication with D Rees, 2011, it was identified that, in the first 3 months following a stroke, patients receive an initial large transfusion, which is six times the strength of a regular transfusion, and then five further regular transfusions. Following the first 3 months post stroke, patients will return to the regular transfusion schedule. Table 30 summarises the transfusion costs used in the model for stroke patients. Detail on the calculation of the cost of blood transfusion is presented in Appendix 4, Table 44. Six transfusions are included in the cost of the first cycle post stroke. From the second cycle post stroke there are three transfusions per cycle as in the pre-stroke model. Account was not taken of the large transfusion received immediately post stroke, as no cost data could be found for this. In order to reflect the uncertainty around the initial cost of stroke, sensitivity of the results to the cost of treatment in the first cycle post-stroke was tested (see Appendix 8).

TABLE 30 - Transfusion costs for stroke patients per 3-month cycle

Cost values incorporate the proportion receiving simple, exchange, and combined, as explained in Table 44.

Cost of the very first transfusion post stroke is included in immediate post-stroke treatment cost (see Table 31).

Time from stroke	No. of transfusions	Cost (£) by age group (years)a
		2–7	8–18	19–30	31+
First cycle post stroke	5b	3667	4917	5691	5730
Second cycle post stroke to death	3	2206	2950	3415	3438

Stroke treatment

The cost of post-stroke treatment varies depending on the effect that stroke has on the patient. Data on the costs of paediatric stroke care in the UK were not available, and the published US data are not applicable. Thus, specific stroke treatments were determined via telephone interviews with clinicians. All post-stroke patients require hospitalisation. Depending on severity of stroke, patients may be put on a ventilator and spend additional days in an intensive care unit (ICU) or high-dependency unit (HDU). Patients who suffer moderate/severe strokes require physical and psychological rehabilitation which may last from a few months to a lifetime. Severely impacted patients would require part or full-time care after discharge. Detail of the calculation of the cost of treatment is presented in Appendix 4. Table 31 presents the cost of treatment per 3-month cycle in various states.

TABLE 31 - Immediate and ongoing (lifetime) post-stroke care costs per 3-month cycle

The disutility value for a moderate stroke was assumed to be the midpoint between the disutility values for a mild and severe stroke.

Post-stroke state	Immediate (first 3 months) post-stroke care costs (£)	Ongoing post-stroke care costs (£)	Disutility
Mild	3737	327	0.03
Moderate	8161	1649	0.08a
Severe	18,417	6618	0.13

Analysis

Modelling and the calculation of model parameters were undertaken in Microsoft Excel, version 2007 (Microsoft Corporation, Redmond, WA, USA). All costs were uplifted to 2010 GBP using gross domestic product (GDP) deflators available from the Treasury. 88 All costs and effects occurring ≥ 1 year after the start of the model were discounted at 3.5%. The analysis was designed to generate two measures of efficiency: incremental cost per quality-adjusted life-year (QALY) gained and incremental cost per stroke avoided (CPSA). More detail on this analysis is presented in Appendix 7. The analysis was validated in a number of ways. First, a number of checks were run to ensure the internal consistency of the models. For instance, the size of the population in each part and cycle of the model was checked to ensure that the cohort size remained 1000 throughout the model. Second, model outcomes were compared with data in the published literature. Specifically, the predicted cost savings produced by the model (post-stroke costs and costs associated with the management of patients with SCD) were compared with those published in the literature. The estimated overall survival was validated against two US studies following patients with SCD from birth,16,40 and the estimated decrease in risk of stroke due to transfusions was validated against data from the STOP trial. 31

The validation should have been undertaken across a broader range of parameters; however, the majority of data available had already been used to specify the model parameters. In addition to validating the outcomes against published literature, the model outcomes were presented to D Rees, P Telfer and A Yardumian to ensure that the model was producing outcomes consistent with their clinical opinions.

Third, sensitivity analysis was performed to determine whether or not the model results are sensitive to any of the uncertainties identified in the evidence. The variables listed in Table 32 were tested as part of one-way sensitivity analyses, as they were considered to be the variables subject to greatest uncertainty.

Efficacy of blood transfusion

This section presents the results of the cost-effectiveness analysis. Table 33 shows the lifetime incremental costs and incremental effects of providing blood transfusions when blood velocity is > 200 cm/second compared with no blood transfusions when blood velocity is > 200 cm/second for a cohort of 2-year-old children with SCD.

The analysis suggests that over the lifetime of patients the intervention could be considered cost-effective with a cost per QALY gained of £24,075 and a CPSA of £203,099. The overall incremental cost-effectiveness ratio (ICER) is within the £20,000–30,000/QALY gained threshold that NICE uses to assess cost-effectiveness.

Over a lifetime, the intervention costs an additional £13.8M – or £13,751 per patient – generates an additional 571 QALYs (an additional 0.6-QALY gain per patient) and helps avoid 68 strokes (0.07 strokes avoided per patient).

Table 33 shows that patients in the intervention arm suffer fewer strokes than patients in the non-intervention arm up until the age of 30 years. After the age of 30 years, four additional strokes are observed in an intervention arm compared with the non-intervention arm. This trend can be explained by the impact of blood transfusion on the death rate. That is, as fewer people are dying as a result of stroke in the early parts of the model, they are living longer and thus increasing the number of strokes later in the model. Although the average likelihood of a stroke later in the model is still lower with blood transfusion than without blood transfusion, the greater number of the cohort alive later in the model causes the number of strokes to increase slightly, and thus increases costs.

The impact of blood transfusion on survival

Figure 5 shows the survival rate of the cohort of patients with SCD who receive blood transfusion.

FIGURE 5.

Overall survival rate of individuals with SCD identified to be at high risk of stroke following TCD and receiving blood transfusion.

Figure 5 also shows a comparison of the results of the model with previous studies of patients with SCD. Quinn et al. 40 studied patients with SCD on transfusion in the USA. They followed the population from birth to the age of 18 years. The survival rate observed by Quinn et al. 40 is indicated on Figure 5 by the dashed grey line. Platt et al. 16 also studied a population with SCD in the USA. Although this population was on treatment, the precise nature of the treatment is not clear from the paper. The dotted line on Figure 5 represents the proportion of the cohort observed by Platt et al. ,16 who were alive at the age of 20 years.

The model was calibrated to the death rates estimated in Platt et al. 16 and Quinn et al. 40 The calibration was performed by adjusting the probabilities of having a second and third stroke. The resulting survival rate is higher than that in the Platt et al. 16 and Quinn et al. 40 studies. This is the result of limits to the number of people having first strokes, and thus able to have second and third strokes. The number of people with a blood velocity of > 200 cm/second by age 18 years was limited to 15% in accordance with clinical opinion (see Pre-stroke component of the Markov model, above).

Impact of blood transfusion on the incidence of stroke

The impact of blood transfusion on the incidence of strokes is summarised in Figure 6. It demonstrates the impact of blood transfusion on the number of first strokes. Without blood transfusion, nearly 13% of patients experience first stroke by the age of 30 years. The equivalent rate with blood transfusion is around 6%.

FIGURE 6.

Number of patients experiencing at least one stroke, with and without blood transfusion, following TCD (cohort size = 1000).

The existing literature on the incidence of first stroke in SCD children was used to populate the model and therefore it was not possible to validate these outputs against published literature. However, D Rees confirmed that incidence of first stroke predicted by the model reflected his clinical opinion.

Costs incurred by patients with sickle cell disease

The model estimates that the lifetime cost (until age 82 years) of treating a patient with SCD in the intervention arm is £52,472. This estimate compares well with those in the existing literature. 81,89 However, Karnon et al. 81 estimate the lifetime costs to be £185,614 (£248,300 in 2010 GBP). Karnon’s work includes the cost of a number of SCD complications, such as renal failure, hip replacement, leg ulcers, acute anaemia, chronic lung disease, retinopathy and other operations, which are not included in our analysis. This extensive list of complications explains the difference between estimates by Karnon et al. 81 and those produced by the model. Saka et al. 89 estimate direct costs of stroke in the UK to be £34,011 per year (in 2010 GBP); this is higher than the average direct annual cost per first stroke at £18,862 per patient per year produced by the model. The difference in these estimates can be explained by the fact that the model costs paediatric stroke. Paediatric patients who have mild or moderate post-stroke outcomes (61% of all strokes) incur fewer costs than the general stroke population, for whom additional comorbidities may prolong post-stroke recovery. For calculation of this comparison, see Appendix 4, Table 50.

Figures 7 and 8 show the breakdown of the costs of treating patients with SCD both with and without blood transfusion following TCD scans. In both arms of the model post-stroke treatment costs account for a sizeable proportion of all costs, and post-stroke treatment costs increase with age. However, there is a clear difference between the intervention and the non-intervention groups. In the latter, post-stroke treatment costs constitute the greatest cost, whereas in the intervention arm, cost of transfusion and chelation are significantly greater than the cost of post-stroke treatment. It is unsurprising that the costs of blood transfusion and chelation and associated treatments are higher with blood transfusion following TCD scans than without. Without blood transfusion following TCD scans, transfusions are provided only post stroke. Costs associated with complications of SCD (pain crisis, acute chest syndrome) are higher in the comparator arm than in the intervention arm and these costs are especially high before the age of 10 years. Once again, this is an expected outcome, as blood transfusion lowers the probability of complications and, consequently, the costs associated with these complications.

FIGURE 7.

Breakdown of cost types with blood transfusion following TCD scans.

FIGURE 8.

Breakdown of cost types without blood transfusion following TCD scans.

Impact of blood transfusion on sickle cell disease-related complications and adverse events

Figure 9 shows incidence of complications and AEs with and without the intervention. As expected, cases of SCD complications (pain crisis and acute chest syndrome) decrease as a result of the intervention.

FIGURE 9.

Number of complications, with and without blood transfusion, following TCD (cohort = 1000).

Figures 10–12 show the incidence of complications and AE over the lifetime of the intervention and comparator arms of the model.

FIGURE 10.

Number of alloimmunised patients, with and without blood transfusion, following TCD.

FIGURE 11.

Number of episodes of acute chest syndrome, with and without blood transfusion, following TCD.

FIGURE 12.

Number of episodes experiencing serious pain crisis, with and without blood transfusion, following TCD.

Cases of alloimmunisation (see Figure 10) are higher in the intervention arm as blood transfusions increase the probability of alloimmunisation. Rates of alloimmunisation become constant from about the age of 30 years, as by this time most patients who are on chronic transfusion have become alloimmunised or have been taken off transfusion. In the model no differentiation is made between different types of alloimmunisation, for example if a patient developed a reaction to at least one of the antibodies he/she is already considered to be alloimmunised and we do not account for further reactions developed to other antibodies.

Cases of complications of SCD (see Figures 11 and 12) are lower in the intervention arm. Rates of acute chest syndrome become constant at about the age of 30 years, as these are more prevalent in younger age groups.

The trends presented in Figure 12 are only for serious cases of pain crisis, defined as cases that require hospital admission or at least an Accident and Emergency department visit.

Clinical review

• No RCTs were identified which evaluated the efficacy of BMT for primary stroke prevention.

• One RCT60 was identified which evaluated the efficacy of hydroxycarbamide for primary stroke prevention; this trial is currently recruiting participants.

• Two RCTs were identified for inclusion in the review (STOP31 and STOP 235); both studied children with SCD identified to be at high risk of stroke using TCD ultrasonography. One study considered the effectiveness of blood transfusions compared with standard care in preventing primary stroke and the other compared discontinuation of prophylactic blood transfusion with continued transfusion.

• Neither of these STOP studies was carried out in the UK.

• The STOP trial31 demonstrated the efficacy of initiating blood transfusions in children with SCD, who were identified to be at high risk of stroke, using TCD ultrasonography in reducing stroke risk.

• The STOP 2 trial35 demonstrated the importance of continued blood transfusion in reducing risk of primary stroke in children with SCD, identified to be at high risk of stroke, using TCD ultrasonography.

• Both trials were terminated early owing to risks associated with the control arm.

• No studies were identified which reported QoL data.

Economic review

• No published economic evaluations relevant to the decision problem were identified from searching the published literature.

• The assessment group developed a de novo economic model to compare TCD scans plus prophylactic blood transfusion in children at high risk, with TCD scans only in patients with SCD aged between 2 and 18 years.

• The economic analysis suggests that blood transfusions post TCD scans for patients with SCD ≥ 2 years (compared with TCD scans alone) may be good value for money. The intervention has an ICER of £24,075 per QALY gained, within the range of the NICE threshold of £20,000–30,000 per QALY gained.

• The intervention leads to improvement in health-related QoL, by helping to avoid 68 strokes over the lifetime of a population of 1000 patients. The intervention costs an additional £13,751 per patient and generates 0.6 extra years of life in full health per patient.

• These estimates are subject to significant uncertainty, given the limitations in the published data. The sensitivity analysis and validation against existing data and expert opinion generally provide some reassurance that the conclusion that blood transfusions for patients with SCD may be cost-effective compared with no transfusion is reasonable. However, it is possible that the conclusion that blood transfusions are cost-effective may be influenced by uncertainty in a small number of model parameters. Further research is thus required to verify the results obtained here.

• Given that the population of interest is patients with SCD of 2 years of age, estimates of costs and benefits later in life – age 18 years and onwards – are subject to substantial uncertainty. Just focusing on the short-term estimates of costs and benefits reinforces the conclusion that blood transfusions are cost-effective. For instance, the ICER is £19,738 if the model is just run between the ages of 2 and 18 years.

Clinical research recommendations

The published literature on the use of TCD ultrasonography to identify children at high risk of stroke and implementation of primary stroke prevention strategies is limited. Most of the RCT data come from the two seminal STOP trials,31,35 both of which considered the efficacy of blood transfusion for primary stroke prevention, and were terminated early due to large numbers of harmful events in the comparator arms. Data on the effectiveness of blood transfusion are available from a number of cohort studies but there are still gaps for which further research would be welcomed.

For example, published data on stroke incidence, morbidity and mortality from stroke in older children and adults would be valuable, particularly if these data were to come from follow-up studies of existing published cohorts, such as the East London cohort30 and the Dallas cohort. 40

Additionally, clinicians still do not know the optimal duration of continuation of blood transfusion into adulthood. There are some data to suggest that patients with abnormal TCD velocities during childhood may not have velocities of > 200 cm/second in adulthood. The majority of patients exhibit no neurological deficit, absent transient ischaemic attack, no or minor cerebral ischaemia, and no or mild cerebrovascular disease, owing to the implementation of early TCD screening and treatment with blood transfusion for stroke prophylaxis before there is any evidence of neurological damage on MRI scanning.

Treatment with hydroxycarbamide is associated with reduced complications and costs compared with blood transfusion. More research is required to assess the potential effects of hydroxycarbamide in reducing risk of primary stroke in children with SCD. Two trials are suggested.

First, it is likely that there is a ‘window period’ of high risk of stroke in children with abnormal TCD velocities, up to age of about 10 years; velocities decrease with age (in adolescence) and an individual’s stroke risk also diminishes after the age of 10 years. It would therefore be useful to assess whether or when transfusions can be discontinued or replaced with hydroxycarbamide therapy in these children owing to the risks associated with long-term blood transfusion. Those children who have been fully protected by transfusion (i.e. those children who have had no stroke, minimal cerebral ischaemia and no, or minimal, cerebrovascular disease) could be randomised to either continued transfusions, discontinuation of transfusion or replacement of transfusion with hydroxycarbamide between the ages of 12 and 15 years. This trial may help to identify the optimal length of continuation of transfusion, and the potential role of hydroxycarbamide in this process. This has important implications for the QoL and cost of treating children with SCD and the subsequent implications for policy and practice within the NHS in the future.

Second, a trial that randomises patients to blood transfusion or hydroxycarbamide for TCD velocities in the borderline to low abnormal category (185–210 cm/second), and assesses the impact of this on stroke rate or HbS levels would be useful. Children with TCD velocities of between 185 and 200 cm/second do not currently receive transfusion for primary stroke prevention, thus this trial will assess the impact of hydroxycarbamide on borderline to low abnormal TCD velocities before they become a greater risk.

Although the clinical benefits of TCD scanning in children have been demonstrated, there are few data relating to long-term impact. Reports from more mature TCD screening programmes may be able to provide information about the long-term benefits associated with TCD scanning from the age of 2 years. There are also no published data relating to QoL in children with SCD who are identified to be at high risk of stroke using TCD ultrasonography and who are undergoing any primary stroke prevention strategy. Further research is required to identify QoL deficits in these children and to establish the impact of various primary stroke prevention strategies on children’s QoL.

In addition, data are needed on the outcomes of implementing TCD scans and blood transfusion in children with SCD (routine clinical practice in the UK). In particular, reports are needed of children’s uptake of TCD screening at the age of 2 years, uptake of transfusion therapy in children with high TCD velocities, and incidence of stroke. Reports on stroke incidence should focus not only on transfused children with high blood flow velocities, but also on children with velocities of < 200 cm/second and who therefore do not receive transfusion.

Economic research recommendations

The published literature on SCD contains a number of gaps in relation to the cost and clinical outcome data required to assess the cost-effectiveness of blood transfusion for patients with SCD. Further research is required to generate more robust data on which to base estimates of cost-effectiveness, or against which model outputs can be calibrated. Specifically, further research should be undertaken looking at utility weights associated with SCD and variation with age, transfusions, strokes and SCD complications – such as pain crisis and transfusion-related treatments (such as alloimmunisation and chelation). This call for further research follows Mazumdar et al. 55 and Nietert et al. ,79 who point out that utility weights associated with SCD have not been established and report QoL data sourced from expert clinical opinion.

Research is also needed around the cost of paediatric stroke care in the UK. In the UK there is a body of literature on costs associated with adult stroke. 89,90 Cost of paediatric stroke care has been estimated in the USA. 91 In the UK, an economic analysis would be helpful to collect relevant resource and cost data associated with paediatric strokes of various severity. Given that it is unlikely that any budget data will accurately reflect the costs of paediatric stroke, it is proposed that this analysis adopts a bottom-up approach. 92

In accordance with NICE guidance,78 data collection should focus on the incurred costs of the health-care system. However, anticipating greater interest in a broader set of costs, it is proposed that the research should also consider indirect costs and the cost of informal care. 93 For example, Saka et al. 89 concluded that indirect costs of post-stroke care constitute 24% and informal care is 27% of total post-stroke costs in the UK. These costs have not been established for post-paediatric stroke care.

In addition, research around post-stroke outcome data would be welcomed. No data were identified relating to the distribution of severity of paediatric stroke in order to identify what proportion of children with SCD end up in mild, moderate, severe or dead post-stroke health states. An ongoing study is currently attempting to develop and validate a paediatric stroke severity scale and to compare these scores with infarct volume and with functional outcomes at 3 and 12 months. 94 Further work would be useful to build on this study to elicit post-stroke outcomes in patients with SCD.

More information is needed about death rates (survival in SCD). Karnon et al. 81 state that there remains considerable uncertainty about the survival of newborn cohorts of patients with sickle cell disorders and the possible development of new complications of new interventions. The present review used survival data reported in Karnon et al. 81 validated by data reported by Platt et al. 16 in 1994. However, as Karnon et al. 81 point out, current management of SCD may have improved survival rates dramatically, so it is important that more recent data on the survival of patients with SCD are collected.

Finally, the frequency of complications and AEs in patients with SCD when on and off transfusion requires further research. The literature states that pain crisis and acute chest syndrome are more frequent in patients with SCD who are not receiving transfusion. However, no data were found regarding rate of these events. Data should be collected longitudinally from patients with SCD who are on and off transfusions to identify the frequency of complications they experience.

Incremental cost

Incremental cost (iC) = CtI – CtC

where:

• CtI = cost of treatment (intervention)

• CtC = cost of treatment (non-intervention)

and:

where:

• SPI_bd = number of patients who have had at least one stroke in state b in cycle d

• NSI_d = number of patients with no previous strokes in cycle d

• SC_b = cost of stroke treatment

• T_f = probability of transfusion of type f

• TC_fg = cost of transfusion, type f (simple, exchange, combined), depending on whether old or new stroke (g)

• G_hf = probability of complication h when on transfusion type f

• GC_h = cost of complication h

• SS_id = cost of TCD scan in cycle d, depending on previous TCD result (i).

Strokes avoided

Strokes avoided (SA) = nCC – nCI

where:

• nCC = number of strokes in non-intervention [nCC = (∑PC_abcd × S_abce)]

• nCI = number of strokes in intervention [nCI = (∑PI_abcd × S_abce)]

where:

• PC_abcd = number of patients in the age group a and state (mild/moderate/bad/dead) b who have had c strokes in cycle d (PI_abcd = patients in intervention group)

• S_abce = probability of strokes for patients in the age group a and state (mild/moderate/bad/dead) b who have had c strokes, post-stroke state e.

Cost per stroke avoided

Cost per stroke avoided (CPSA) = SA_d/iC_d

where:

• SA_d = the difference in the number of strokes avoided in the non-intervention scenario compared with the intervention scenario in period d

• iC_d = the difference in cost in the intervention scenario compared with the non-intervention scenario in period d.

Incremental quality-adjusted life-year

Incremental quality-adjusted life-year (iQALY) = QALYlc – QALYli

where:

where:

• SPI_bd = number of patients in the intervention arm who have had at least one stroke in state b in cycle d

• SPC_bd = number of patients in the non-intervention arm who have had at least one stroke in state b in cycle d

• NSI_d = number of patients with no previous strokes in cycle d in the intervention arm

• NSC_d = number of patients with no previous strokes in cycle d in the non-intervention arm

• sQALY_b = QALY loss from stroke treatment, state b

• T_f = probability of transfusion of type f

• tQALY_f = QALY loss from transfusion, type f (simple, exchange, combined)

• G_hf = probability of complication h when on transfusion type f

• gQALY_h = QALY loss for complication h.

Incremental cost-effectiveness ratio

ICER = iQALY_d/iCost_d

where:

• iQALY_d = the difference in QALY loss in the intervention scenario compared with the non-intervention scenario in period d

• iCost_d = the difference in cost in the intervention scenario compared with the non-intervention scenario in period d.

TAR team

Liverpool Reviews and Implementation Group (LRiG), University of Liverpool

Correspondence to:

Rumona Dickson, Dr

Director, LRiG

University of Liverpool

Whelan Building

The Quadrangle

Brownlow Hill

Liverpool

L69 3GB

Tel: +44 (0) 151 794 5682/5067

Fax: +44 (0)151 794 5821

Email: R.Dickson@liv.ac.uk

For details of expertise within the TAR team, see Expertise in this TAR team and competing interests of authors, below.

Plain English summary

Sickle cell disease (SCD) is an inherited disorder characterised by unpredictable episodes of acute illness and chronic organ damage. Sickle cell disease is now the most common genetic condition in the UK, affecting approximately 1 in every 2000. Sickle cell anaemia (SCA) is the most common type of sickle cell disease, which occurs when the sickle mutation is inherited from both parents. Nearly all studies of stroke and SCD have involved people with SCA. Rates of stroke in individuals with SCA are higher than the general population, with risk of stroke estimated at 0.7 per 100 patient years at age 5 years, 2.7 at age 10 years, 4.3 at age 15 years and 12.8 at age 20 years.

Transcranial Doppler (TCD) ultrasonography can be used to identify children at high risk of stroke by measuring blood velocity in the brain. Approximately 10% of children who have blood velocity of > 200/second go on to suffer a stroke within a year. In order to identify those at high risk of stroke, TCD ultrasound screening is now recommended for children between the ages of 2 and 16 years who have SCA and no previous history of stroke. Treatment to prevent stroke, which in the UK is regular blood transfusions, may then be offered to children considered to be at high risk.

This review aims to assess whether or not primary stroke prevention strategies for use in children with SCA are clinically useful. The review will compare treatment with blood transfusions and treatment with hydroxycarbamide with no intervention and/or with each other. If suitable data are available, the review will also consider the cost-effectiveness of TCD ultrasound screening to identify children at high risk of stroke, and their subsequent treatment.

Decision problem

Current guidelines

A clinical alert issued in 1997 by the National Heart, Lung and Blood Institute (NHLBI) in the USA,23 recommends that children with SCA and HbS/ β⁰ thalassaemia SCD, with no previous history of stroke and who are between the ages of 2 and 16 be screened using TCD ultrasound to identify those with high cerebral blood velocity rates, and therefore increased risk of stroke. This recommendation was based on the results of the STOP16 trial. The NHLBI further advocates considering transfusion for children who have received two abnormal TCD ultrasound results, as a preventative measure for stroke. 23

A second alert, based on the findings of STOP 2,17 was issued in 2004. 23 This recommended that blood transfusions be continued to reduce the rate of strokes in children with SCD until at least 3 years. The Special Writing Group of the American Heart Association Stroke Council and the Council on Cardiovascular Disease in the Young have since advised that transfusion be continued for at least 5 years or until the child is 18 years old. 24

In March 2009, the NHS Antenatal and Newborn Screening Programme produced guidelines on the management of stroke in children with SCA and HbS/β⁰ thalassaemia. 18 These guidelines (based on a combination of a review of the literature and clinical expert opinion) state that children and young adults with SCA and HbS/⁰thalassaemia should be offered annual TCD ultrasound scans from the age of 2 years until at least the age of 16 years. Children should be classified as either ‘high risk’, ‘conditional’ or ‘standard risk’ in line with the results of the STOP16 study (Table 3).

The NHS Antenatal and Newborn Screening Programme guidelines further state that children with ‘high risk’ and ‘conditional’ TCD scans should have them repeated within 2 months and the benefits of receiving regular blood transfusions should be discussed with parents for those children remaining at high risk. Primary stroke prevention treatment following a second high velocity reading is transfusion continued throughout childhood. 16

Recent guidelines published by the Royal College of Physicians in the UK endorse annual imaging using TCD ultrasound scanning of children with SCD from the age of 3 years. 12

Objectives of the HTA project

The aim of this review is to assess the clinical effectiveness and cost-effectiveness of primary stroke prevention treatments in children with SCD, identified to be at high risk of stroke following TCD ultrasonography. The review will consider the effectiveness of primary stroke prevention treatment, treatment with blood transfusion and treatment with hydroxycarbamide, for the prevention of stroke in children with SCD who are identified to be at high risk of stroke. The review will also examine the existing health economic evidence and identify the key economic issues related to primary stroke prevention treatment in clinical care for this group of patients. If suitable data are available, an economic model will be developed and populated to evaluate the cost-effectiveness of primary stroke prevention treatments within the NHS.

Search strategy

The major electronic databases including MEDLINE, EMBASE and The Cochrane Library will be searched for relevant published literature. Information on studies in progress, unpublished research or research reported in the grey literature will be sought by searching a range of relevant databases including National Research Register and Controlled Clinical Trials.

Bibliographies of previous systematic reviews and retrieved articles will also be examined. A database of published and unpublished literature will be assembled from systematic searches of electronic sources, hand searching, and consultation with experts in the field. The database will be held in the EndNote X4 software package.

Systematic review of published economic literature

The literature review of economic evidence will identify any relevant published cost-minimisation, cost-effectiveness, cost-utility and/or cost-benefit analyses.

Methods of analysis/synthesis

Prioritisation Group

1. Director, NIHR HTA programme, Professor of Clinical Pharmacology, Department of Pharmacology and Therapeutics, University of Liverpool

2. Professor Imti Choonara, Professor in Child Health, Academic Division of Child Health, University of Nottingham

   Chair – Pharmaceuticals Panel

3. Dr Bob Coates, Consultant Advisor – Disease Prevention Panel

4. Dr Andrew Cook, Consultant Advisor – Intervention Procedures Panel

5. Dr Peter Davidson, Director of NETSCC, Health Technology Assessment

6. Dr Nick Hicks, Consultant Adviser – Diagnostic Technologies and Screening Panel, Consultant Advisor–Psychological and Community Therapies Panel

7. Ms Susan Hird, Consultant Advisor, External Devices and Physical Therapies Panel

8. Professor Sallie Lamb, Director, Warwick Clinical Trials Unit, Warwick Medical School, University of Warwick

   Chair – HTA Clinical Evaluation and Trials Board

9. Professor Jonathan Michaels, Professor of Vascular Surgery, Sheffield Vascular Institute, University of Sheffield

   Chair – Interventional Procedures Panel

10. Professor Ruairidh Milne, Director – External Relations

11. Dr John Pounsford, Consultant Physician, Directorate of Medical Services, North Bristol NHS Trust

    Chair – External Devices and Physical Therapies Panel

12. Dr Vaughan Thomas, Consultant Advisor – Pharmaceuticals Panel, Clinical

    Lead – Clinical Evaluation Trials Prioritisation Group

13. Professor Margaret Thorogood, Professor of Epidemiology, Health Sciences Research Institute, University of Warwick

    Chair – Disease Prevention Panel

14. Professor Lindsay Turnbull, Professor of Radiology, Centre for the MR Investigations, University of Hull

    Chair – Diagnostic Technologies and Screening Panel

15. Professor Scott Weich, Professor of Psychiatry, Health Sciences Research Institute, University of Warwick

    Chair – Psychological and Community Therapies Panel

16. Professor Hywel Williams, Director of Nottingham Clinical Trials Unit, Centre of Evidence-Based Dermatology, University of Nottingham

    Chair – HTA Commissioning Board

    Deputy HTA Programme Director

HTA Commissioning Board

1. Professor of Dermato-Epidemiology, Centre of Evidence-Based Dermatology, University of Nottingham

2. Professor of Bio-Statistics, Department of Public Health and Epidemiology, University of Birmingham

3. Professor of Clinical Pharmacology, Director, NIHR HTA programme, Department of Pharmacology and Therapeutics, University of Liverpool

1. Professor Zarko Alfirevic, Head of Department for Women’s and Children’s Health, Institute of Translational Medicine, University of Liverpool

2. Professor Judith Bliss, Director of ICR-Clinical Trials and Statistics Unit, The Institute of Cancer Research

3. Professor David Fitzmaurice, Professor of Primary Care Research, Department of Primary Care Clinical Sciences, University of Birmingham

4. Professor John W Gregory, Professor in Paediatric Endocrinology, Department of Child Health, Wales School of Medicine, Cardiff University

5. Professor Steve Halligan, Professor of Gastrointestinal Radiology, Department of Specialist Radiology, University College Hospital, London

6. Professor Angela Harden, Professor of Community and Family Health, Institute for Health and Human Development, University of East London

7. Dr Joanne Lord, Reader, Health Economics Research Group, Brunel University

8. Professor Stephen Morris, Professor of Health Economics, University College London, Research Department of Epidemiology and Public Health, University College London

9. Professor Dion Morton, Professor of Surgery, Academic Department of Surgery, University of Birmingham

10. Professor Gail Mountain, Professor of Health Services Research, Rehabilitation and Assistive Technologies Group, University of Sheffield

11. Professor Irwin Nazareth, Professor of Primary Care and Head of Department, Department of Primary Care and Population Sciences, University College London

12. Professor E Andrea Nelson, Professor of Wound Healing and Director of Research, School of Healthcare, University of Leeds

13. Professor John David Norrie, Director, Centre for Healthcare Randomised Trials, Health Services Research Unit, University of Aberdeen

14. Professor Barney Reeves, Professorial Research Fellow in Health Services Research, Department of Clinical Science, University of Bristol

15. Professor Peter Tyrer, Professor of Community Psychiatry, Centre for Mental Health, Imperial College London

16. Professor Martin Underwood, Professor of Primary Care Research, Warwick Medical School, University of Warwick

17. Professor Caroline Watkins, Professor of Stroke and Older People’s Care, Chair of UK Forum for Stroke Training, Stroke Practice Research Unit, University of Central Lancashire

18. Dr Duncan Young, Senior Clinical Lecturer and Consultant, Nuffield Department of Anaesthetics, University of Oxford

1. Dr Tom Foulks, Medical Research Council

2. Dr Kay Pattison, Senior NIHR Programme Manager, Department of Health

HTA Clinical Evaluation and Trials Board

1. Director, Warwick Clinical Trials Unit, Warwick Medical School, University of Warwick and Professor of Rehabilitation, Nuffield Department of Orthopaedic, Rheumatology and Musculoskeletal Sciences, University of Oxford

2. Professor of the Psychology of Health Care, Leeds Institute of Health Sciences, University of Leeds

3. Director, NIHR HTA programme, Professor of Clinical Pharmacology, University of Liverpool

1. Professor Keith Abrams, Professor of Medical Statistics, Department of Health Sciences, University of Leicester

2. Professor Martin Bland, Professor of Health Statistics, Department of Health Sciences, University of York

3. Professor Jane Blazeby, Professor of Surgery and Consultant Upper GI Surgeon, Department of Social Medicine, University of Bristol

4. Professor Julia M Brown, Director, Clinical Trials Research Unit, University of Leeds

5. Professor Alistair Burns, Professor of Old Age Psychiatry, Psychiatry Research Group, School of Community-Based Medicine, The University of Manchester & National Clinical Director for Dementia, Department of Health

6. Dr Jennifer Burr, Director, Centre for Healthcare Randomised trials (CHART), University of Aberdeen

7. Professor Linda Davies, Professor of Health Economics, Health Sciences Research Group, University of Manchester

8. Professor Simon Gilbody, Prof of Psych Medicine and Health Services Research, Department of Health Sciences, University of York

9. Professor Steven Goodacre, Professor and Consultant in Emergency Medicine, School of Health and Related Research, University of Sheffield

10. Professor Dyfrig Hughes, Professor of Pharmacoeconomics, Centre for Economics and Policy in Health, Institute of Medical and Social Care Research, Bangor University

11. Professor Paul Jones, Professor of Respiratory Medicine, Department of Cardiac and Vascular Science, St George‘s Hospital Medical School, University of London

12. Professor Khalid Khan, Professor of Women’s Health and Clinical Epidemiology, Barts and the London School of Medicine, Queen Mary, University of London

13. Professor Richard J McManus, Professor of Primary Care Cardiovascular Research, Primary Care Clinical Sciences Building, University of Birmingham

14. Professor Helen Rodgers, Professor of Stroke Care, Institute for Ageing and Health, Newcastle University

15. Professor Ken Stein, Professor of Public Health, Peninsula Technology Assessment Group, Peninsula College of Medicine and Dentistry, Universities of Exeter and Plymouth

16. Professor Jonathan Sterne, Professor of Medical Statistics and Epidemiology, Department of Social Medicine, University of Bristol

17. Mr Andy Vail, Senior Lecturer, Health Sciences Research Group, University of Manchester

18. Professor Clare Wilkinson, Professor of General Practice and Director of Research North Wales Clinical School, Department of Primary Care and Public Health, Cardiff University

19. Dr Ian B Wilkinson, Senior Lecturer and Honorary Consultant, Clinical Pharmacology Unit, Department of Medicine, University of Cambridge

1. Ms Kate Law, Director of Clinical Trials, Cancer Research UK

2. Dr Morven Roberts, Clinical Trials Manager, Health Services and Public Health Services Board, Medical Research Council

Diagnostic Technologies and Screening Panel

1. Scientific Director of the Centre for Magnetic Resonance Investigations and YCR Professor of Radiology, Hull Royal Infirmary

2. Professor Judith E Adams, Consultant Radiologist, Manchester Royal Infirmary, Central Manchester & Manchester Children’s University Hospitals NHS Trust, and Professor of Diagnostic Radiology, University of Manchester

3. Mr Angus S Arunkalaivanan, Honorary Senior Lecturer, University of Birmingham and Consultant Urogynaecologist and Obstetrician, City Hospital, Birmingham

4. Dr Diana Baralle, Consultant and Senior Lecturer in Clinical Genetics, University of Southampton

5. Dr Stephanie Dancer, Consultant Microbiologist, Hairmyres Hospital, East Kilbride

6. Dr Diane Eccles, Professor of Cancer Genetics, Wessex Clinical Genetics Service, Princess Anne Hospital

7. Dr Trevor Friedman, Consultant Liason Psychiatrist, Brandon Unit, Leicester General Hospital

8. Dr Ron Gray, Consultant, National Perinatal Epidemiology Unit, Institute of Health Sciences, University of Oxford

9. Professor Paul D Griffiths, Professor of Radiology, Academic Unit of Radiology, University of Sheffield

10. Mr Martin Hooper, Public contributor

11. Professor Anthony Robert Kendrick, Associate Dean for Clinical Research and Professor of Primary Medical Care, University of Southampton

12. Dr Nicola Lennard, Senior Medical Officer, MHRA

13. Dr Anne Mackie, Director of Programmes, UK National Screening Committee, London

14. Mr David Mathew, Public contributor

15. Dr Michael Millar, Consultant Senior Lecturer in Microbiology, Department of Pathology & Microbiology, Barts and The London NHS Trust, Royal London Hospital

16. Mrs Una Rennard, Public contributor

17. Dr Stuart Smellie, Consultant in Clinical Pathology, Bishop Auckland General Hospital

18. Ms Jane Smith, Consultant Ultrasound Practitioner, Leeds Teaching Hospital NHS Trust, Leeds

19. Dr Allison Streetly, Programme Director, NHS Sickle Cell and Thalassaemia Screening Programme, King’s College School of Medicine

20. Dr Matthew Thompson, Senior Clinical Scientist and GP, Department of Primary Health Care, University of Oxford

21. Dr Alan J Williams, Consultant Physician, General and Respiratory Medicine, The Royal Bournemouth Hospital

1. Dr Tim Elliott, Team Leader, Cancer Screening, Department of Health

2. Dr Joanna Jenkinson, Board Secretary, Neurosciences and Mental Health Board (NMHB), Medical Research Council

3. Professor Julietta Patrick, Director, NHS Cancer Screening Programme, Sheffield

4. Dr Kay Pattison, Senior NIHR Programme Manager, Department of Health

5. Professor Tom Walley, CBE, Director, NIHR HTA programme, Professor of Clinical Pharmacology, University of Liverpool

6. Dr Ursula Wells, Principal Research Officer, Policy Research Programme, Department of Health