NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Cherry MG, Greenhalgh J, Osipenko L, et al. The Clinical Effectiveness and Cost-Effectiveness of Primary Stroke Prevention in Children with Sickle Cell Disease: A Systematic Review and Economic Evaluation. Southampton (UK): NIHR Evaluation, Trials and Studies Coordinating Centre (UK); 2012 Nov. (Health Technology Assessment, No. 16.43.)

Cover of The Clinical Effectiveness and Cost-Effectiveness of Primary Stroke Prevention in Children with Sickle Cell Disease: A Systematic Review and Economic Evaluation

The Clinical Effectiveness and Cost-Effectiveness of Primary Stroke Prevention in Children with Sickle Cell Disease: A Systematic Review and Economic Evaluation.

Show details

4Assessment of clinical effectiveness


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

TABLE 9. Methodological quality of included trials.


Methodological quality of included trials.

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.

TABLE 10. Trial characteristics.


Trial characteristics.

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).

TABLE 11. Participant characteristics.


Participant characteristics.

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


Number of and type of transfusion

In STOP,31 the 63 patients in the transfusion group received a total of 1521 transfusions. Of these, 63% were simple transfusions, 12% were exchange transfusions, and 25% were a combination of simple and exchange transfusions. The mean interval between transfusions was 25 days [standard deviation (SD) = 8]. The 143 episodes in which the target level of HbS of < 30% was exceeded were ‘usually isolated and minor’. Ten patients dropped out of the transfusion group – four due to compliance issues, one due to multiple antibodies, one due to ineligibility and four for unspecified reasons. Two patients crossed over to the transfusion group.

In STOP 2,35 the 38 patients in the transfusion group received 1070 transfusions. Of these, 19 patients received transfusion without phlebotomy, four received manual exchanges and seven received automated erythrocytapheresis (a method for administering exchange transfusion); eight patients received transfusion by two or more methods. Measures of HbS were reported as 76% meeting the stated target level of < 30%, 19% were above the target level but < 40%, and 5% were > 40%. Five patients discontinued participation in the transfusion group. Chelation therapy was received by 93% of patients in the transfusion arm and 76% in the transfusion-halted arm.

Primary outcomes

The primary outcome of STOP31 was stroke (cerebral infarction or intracranial haemorrhage). Table 12 shows that in STOP31 one patient in the transfusion group had a stroke compared with 11 patients in the standard care group (p < 0.001). This equates to a 92% lower risk of stroke for patients in the transfusion group.

TABLE 12. Primary outcomes of STOP and STOP 2 trials.


Primary outcomes of STOP and STOP 2 trials.

In STOP 2,35 the primary composite end point was stroke (cerebral infarction or intracranial haemorrhage) or reversion to abnormal velocity on TCD scans. In Table 12, 16 patients in the transfusion-halted group experienced an event, whereas there were no events in the continued-transfusion group (p < 0.001). In the transfusion-halted group, two of the patients had a stroke, whereas 14 other patients reverted to abnormal velocities measured by TCD scan.

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, 4345, 6870 were identified. The majority were retrospective cohort studies. Of these seven studies,41, 4345, 6870 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.

TABLE 13. Characteristics of non-randomised studies.


Characteristics of non-randomised studies.

Clinical discussion

The purpose of this clinical review was to evaluate the clinical effectiveness of primary stroke prevention treatments for children with SCD identified by TCD to be at high risk of stroke. The review considered the effectiveness of blood transfusion, hydroxycarbamide and BMT compared with standard care. Two relevant RCTs – STOP31 and STOP 235 – were identified for inclusion in the review; STOP31 compared blood transfusion with standard care, whereas STOP 235 compared continued blood transfusion with halted blood transfusion in previously transfused patients.

The STOP31 and STOP 235 trials both utilised stroke (and, in the case of STOP 2,35 reversion to abnormal TCD velocity) as their primary end point; however, we considered that the patient populations of these two trials were too different to synthesise their data using a standard meta-analytic approach. Patients in STOP 235 had all previously received blood transfusion for primary stroke prevention, whereas patients in STOP31 had not.

The results of STOP31 demonstrated a statistically significant difference in the number of strokes between the transfusion and standard-care arms (1 vs 11; p < 0.001). The trial was halted early due to the higher number of stroke events in the standard-care group compared with the blood transfusion group.

The results of STOP 235 demonstrated a statistically significant difference in the number of end point events (stroke or reversion to abnormal TCD velocity) between the two arms of the trial; no events occurred in the continued-transfusion arm compared with 16 patients in the halted-transfusion arm (p < 0.001). Two of the events were strokes; the remainder were reversions to abnormal TCD velocities. The STOP 235 trial was also closed early owing to the higher number of events in the halted-transfusion group compared with the continued-transfusion group.

The STOP31,35 trials were (relatively) small trials in terms of patient numbers. Both trials were terminated early due to the number of events that occurred in the comparator groups. Early closure is of concern given the findings of a recent meta-analysis71 that compared the results of 91 trials that were closed prematurely for benefit, with 424 similar trials that ran to full term. The authors reported large differences in treatment effect size between trials that were stopped early and similar trials that ran their full course. This was true regardless of the methodological quality of the trial or the presence of statistical stopping rules. One implication of this finding is that early closure of trials can lead to exaggerated treatment effects that would not be borne out in the longer term. Although it would clearly be unethical to have continued the STOP trials31,35 it is unclear what the full-term outcomes might have been.

In Chapter 2 (see Current guidelines for stroke prevention) it was reported that clinical guidelines in the UK65 and the USA64 for the management of children with SCD are based on the results of the STOP trials,31,35 i.e. children with SCD aged ≥ 2 years should receive annual TCD scans. Children who are identified as being at high risk of stroke (those with abnormal scan results) should be considered for prophylactic blood transfusions that will continue throughout their childhood. Studies of patient cohorts in the UK,30 the USA42 and France45 attest to the benefits of implementing the STOP31 protocol in clinical practice in terms of reduced rates of stroke per patient-year.

Following the publication of the seminal STOP31,35 trials, which clearly show the benefit of initiating and continuing chronic transfusion therapy in children with high TCD velocities, a number of key issues remain unresolved. Treatment with monthly blood transfusion carries a number of serious risks (such as iron overload, alloimmunisation, unknown future infection) and disbenefits (such as iron chelation treatment, regular hospital visits for transfusions), and yet it is not known for how long prophylactic blood transfusion should continue. At present, although guidelines recommend that blood transfusion be continued until the age of 18 years, most children are transfused at least until they are 16 years old, and approximately 75% of transfused children receive blood transfusion for life. This means that a number of children or adults may be receiving treatment beyond the time of its benefit. An estimated 60% of patients who have TCD scans that show abnormal velocities do not go on to have a stroke.45 This means that a considerable proportion of patients will receive treatment that carries a number of serious risks but for no benefit. There is currently no means of predicting which children will go on to have a stroke.

Iron overload is a significant side effect of treatment with regular blood transfusion. It was noted in Chapter 2 (see Regular blood transfusion) that the majority of data with respect to iron overload are derived from patients with thalassaemia major. The pattern of iron overload in patients with SCD is poorly understood. This is an important area for future research. Data on compliance with older iron chelation therapy47,72 suggest that adherence to treatment is generally poor but may be improved by the use of newer oral chelation methods.48,49 More recording of data that reflect the use of oral deferasirox is needed to enable the benefits of improved adherence to chelation treatment to be fully understood.

The effects of long-term blood transfusion on mortality rates are also unknown. The Mazumdar55 model discussed in Chapter 2 (see Regular blood transfusion) suggests that chronic transfusion, although decreasing stroke risk, may also impact negatively on mortality. This assumption has been criticised24 and the counter-argument proposed – that one would expect chronic transfusion to ameliorate the progression of chronic complications of SCD seen in adulthood and improve life expectancy. There are, as yet, no data to support any claims regarding mortality benefits or disbenefits.

The authors of this review experienced considerable difficulty in obtaining UK data regarding the numbers of children affected by SCD, their treatments and outcomes. A National Haemoglobinopathy Register (sponsored by the Department of Health) has recently been set up in order to obtain data on the prevalence of SCD across the country, and the frequency of specific treatment interventions and of specific severe complications, including mortality. In addition, the West Midland Quality Review Service very recently published an overview of services across the UK for children with haemoglobin disorders.73 It is likely that the registry and overview will, in time, prove to be of great value in future treatment planning for patients with SCD, providing that funding continues.

No data were identified which considered the efficacy of hydroxycarbamide for primary stroke prevention. The recently completed BABY HUG RCT59 compared hydroxycarbamide with placebo in reducing organ dysfunction and clinical complications in children with SCD, aged between 9 and 18 months. The average increase in TCD velocity was found to be ‘significantly less’ when receiving hydroxycarbamide compared with placebo. These findings may impact on the future treatment of children with SCD by reducing TCD velocity prior to the recommended initial TCD scan at the age of 2 years.

The findings of this review suggest that TCD ultrasonography is an effective method of identifying children with SCD who may be at high risk of stroke. However, it is important to note that, historically, not all children in the UK have been within geographically easy access to a TCD screening centre74 and therefore some potentially high-risk children may not have received recommended care. Implementation of the NHS Sickle Cell and Thalassaemia Screening Programme75 has partially addressed this issue but further work is needed in this area to ensure greater access. It is recommended that scanning using TCD for all children with SCD should be made routinely available at hospitals across the country to ensure that all infants have easy access to a testing centre.

© 2012, Crown Copyright.

Included under terms of UK Non-commercial Government License.

Bookshelf ID: NBK115677
PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (2.8M)
  • Disable Glossary Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...