Defining stroke risk in children with sickle cell anaemia

Transcription

1 review Defining stroke risk in children with sickle cell anaemia Carolyn Hoppe Department of Hematology/Oncology, Children s Hospital and Research Center at Oakland, Oakland, CA, USA Summary Sickle cell anaemia (SCA) is the most common cause of childhood stroke, occurring with the highest frequency before the age of 6 years. Despite the relative frequency of stroke in SCA, few predictors of risk exist. Anaemia, leucocytosis, hypertension, silent infarction, and history of acute chest syndrome are well-documented risk factors for ischaemic stroke in SCA. Recent data suggest that other environmental and genetic factors, many unrelated to SCA, influence the development of cerebrovascular disease. Non-invasive assessment of individual stroke risk using transcranial Doppler ultrasonography has provided a means of selecting and prophylactically treating SCA children at highest risk. With the ultimate goal of preventing stroke, the information gained from the studies reviewed here may lead to improved prediction of stroke so that clinical trials to assess risk-based therapy may be carried out on selected children with SCA. Keywords: stroke, sickle cell anaemia, risk, children, review. Children with sickle cell anaemia (SCA) carry a 300-fold increased risk for stroke, making SCA the most common cause of childhood stroke (Ohene-Frempong et al, 1998). By 20 years of age, 11% of children with SCA will have experienced a clinical stroke syndrome, and a further 17 22% will have subclinical evidence of cerebral infarction on brain magnetic resonance imaging (MRI). These subclinical lesions on MRI correlate with significant neuropsychological deficits, and almost half of the children with silent infarcts will eventually require life-long support or custodial care (Schatz et al, 2002). Even with adequate chronic transfusion therapy, the child who has had a stroke may be left with significant physical and/or cognitive deficits that limit quality of life (Armstrong et al, 1996). As a result, there has been considerable effort to identify patients at risk for stroke and intervene before a stroke has occurred. This review will discuss the clinical, neuroimaging and genetic factors that are presently used to define stroke risk in children with SCA. Epidemiology of stroke in SCD Powars et al (1978) first reported on the natural history of stroke in sickle cell disease (SCD) 25 years ago, documenting a peak incidence in young children. Moreover, strokes recurred in 67% of untreated patients, with the highest risk during the first 3 years following the initial stroke. Since then, the most extensive findings on stroke incidence have come from the Cooperative Study of Sickle Cell Disease (CSSCD), a national, multicentre study designed to define the natural history of SCD in the USA. This study determined prevalence and incidence rates of stroke, based on data from over 4000 SCD patients followed for up to 10 years from 1978 to 1988 (Ohene-Frempong et al, 1998). As part of this study, the CSSCD initiated a prospective infant cohort trial in 1978 in which almost 700 infants were enrolled at birth and followed prospectively. In this cohort, all children had brain MRI scanning performed when aged 6 years and every 2 years thereafter. This study confirmed earlier reports of a high rate of symptomatic and asymptomatic cerebral infarction in children with SCD, documenting an 11% cumulative incidence of clinical stroke and a 17% prevalence of unsuspected cerebral infarction on MRI (Sarnaik & Lusher, 1982; Balkaran et al, 1992). The majority of strokes occurred in SCA (homozygous HbS) patients, and much less frequently in the other common genotypes of SCD. Slightly lower rates of stroke have been observed in other SCD populations. The French Study Group on SCD documented a 6Æ7% prevalencse of clinically overt stroke and a 15% prevalence of silent infarcts in SCA patients of primarily African descent (Bernaudin et al, 2000; Neonato et al, 2000). The incidence of stroke in a birth cohort of 310 Jamaican children with SCD was found to be 7Æ8% by age 14 years (Balkaran et al, 1992). Stroke appears to be rare in Saudi- Arabian and Nigerian children with SCD (Izuora et al, 1989, 2003; Al-Rajeh et al, 1991, Obama et al, 1994). Silent infarcts are also uncommon in the Kuwaiti SCD population and are found primarily in adults, with an overall prevalence of 3Æ3% (Marouf et al, 2003). Stroke phenotypes in SCD Correspondence: Carolyn Hoppe, Department of Hematology/ Oncology, Children s Hospital Oakland, nd Street, Oakland, CA 94609, USA. choppe@mail.cho.org The cerebrovascular manifestations of SCA vary from extensive, large vessel distribution infarcts to more subtle lacunar ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128, doi: /j x

2 infarcts, and differ in their epidemiologic, clinical and pathologic features (Table I). The majority of strokes in children with SCA are ischaemic, with intracranial haemorrhage becoming more common later in life (Ohene-Frempong et al, 1998). In most cases of symptomatic stroke, infarction results from large vessel occlusion usually in the distribution of the distal internal carotid (dica), proximal middle cerebral (MCA), and anterior cerebral (ACA) arteries (Stockman et al, 1972; Russell et al, 1976) (Fig 1). Moyamoya syndrome, resulting from progressive narrowing of these vessels with compensatory collateral vessel development, occurs in 30% of stroke patients and indicates severe cerebrovascular disease (Seeler et al, 1978; Dobson et al, 2002). Children with neuroimaging evidence of moyamoya syndrome show greater neuropsychological impairments than stroke patients lacking moyamoya changes (Dobson et al, 2002). These children are also at higher risk of recurrent stroke, even with transfusion therapy (Dobson et al, 2002). Haemorrhagic stroke, fatal in 26% of cases, is usually caused by rupture of moyamoya vessels or aneurysms (Oyesiku et al, 1991; Ohene-Frempong et al, 1998). Asymptomatic, or silent infarcts appear on MRI as punctate hyperintensities in the deep white matter of the brain. These ischaemic lesions are a result of small vessel occlusion and usually occur in arterial borderzones. Children with silent infarcts identified by MRI may appear asymptomatic, but perform significantly lower on neuropsychological tests than their counterparts with a normal MRI (Armstrong et al, 1996). While stroke in the patient with SCA is usually attributable to SCD itself, one cannot overlook other diagnostic possibilities, such as endocarditis, arterial dissection, antiphospholipid antibody syndrome, inherited prothrombotic states, and cerebral venous thrombosis, that may represent the primary aetiology or contribute to the genesis of stroke (de Veber, 2003). Pathogenesis of stroke in SCD Although the natural history of stroke in SCD has been wellcharacterized, the pathogenesis and progression of cerebrovascular disease in affected children are not fully understood. Histopathological interpretation of large vessel involvement in children with SCD and stroke varies, but, in general, shows a pattern of smooth muscle proliferation with overlying endothelial damage and fibrosis (Rothman et al, 1986; Koshy et al, 1990). Smaller arterioles and capillaries are marked by distention, thrombosis and vessel wall necrosis (Tuohy et al, 1997). Sickle cell disease may contribute to the development of stroke by providing a source of persistent endothelial injury via the effects of hypoxia, increased shear stress, abnormal endothelial adherence of sickled red blood cells (RBCs), and inflammation induced by reperfusion injury (Fig 2) (Francis, 1991; Hebbel & Vercellotti, 1997; Solovey et al, 1997). These effects trigger the production of cytokines, leading to Table I. Comparison of stroke phenotypes*. Ischaemic stroke Haemorrhagic stroke Silent infarction Peak incidence (age) 2 5 years years Unknown; prevalence ¼ 22% in 6 12 year olds; number of lesions increase with age Presentation Hemiparesis; aphasia; hemisensory or visual Severe headache; altered consciousness; Absence of neurologic symptoms; neurocognitive dysfunction deficits; focal seizures generalized seizures; syncope Neuroimaging MRI: infarcts in dica, MCA territory; MRA: stenosis/ SAH or ICH due to ruptured Arterial borderzone infarcts; confined to deep white matter occlusion of large intracranial arteries; moyamoya disease aneurysm or moyamoya collaterals Histopathology LV: intimal hyperplasia, thrombosis, smooth muscle Aneurysmal dilatation in hypertrophy; aneurysms regions of intimal hyperplasia Location of lesions Frontal, parietal, temporal lobes; Frontal, parietal, temporal lobes; <1Æ5 cm basal ganglia, thalamus; >1Æ5 cm Screening/primary prevention TCD Unknown Unknown (?neuropsychometric testing; functional imaging studies) Intervention/treatment Chronic transfusion therapy; HCT Surgical revascularization;?chronic Unknown (?chronic transfusion therapy) transfusion therapy dica, distal internal carotid artery; MCA, middle cerebral artery; HCT, haematopoietic stem cell transplantion; LV, large vessels; TCD, transcranial Doppler ultrasound; HCT, haematopoietic stem cell transplantation; SAH, subarachnoid haemorrhage; ICH, intracerebral haemorrhage. *For simplicity, stroke has been broken down into three categories; however, individual patients may fall into more than one category. 752 ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128,

3 Fig 1. FLAIR MRI, MRA. (A) FLAIR MRI of the brain in an 11-year old child with SCA demonstrating a left ACA-left MCA borderzone frontal infarct and a right MCA territory fronto-parietal infarct, extending towards the right MCA-ACA deep borderzone (arrows). (B) The corresponding MRA shows complete absence of flow within the terminal right ICA, with reduced flow in the right MCA (large arrows). Narrowing of the proximal left MCA is also present (small arrow). The patient experienced their first stroke at age 4 years. Chronic transfusions were discontinued after 5 years, and hydroxyurea was initiated. Nine months later, the patient experienced a recurrent stroke. Abbreviations: ACA, anterior cerebral artery; MCA, middle cerebral artery; ICA, internal carotid artery. endothelial dysregulation, upregulation of proinflammatory cells, and activation of prothrombotic (e.g. antiphospholid antibodies, protein C, S), and proadhesive molecules [e.g. vascular cell adhesion molecule (VCAM-1), intracellular adhesion molecule (ICAM-1), E-selectin], while inhibiting production of cytoprotective mediators (e.g. nitric oxide) (Shreeniwas et al, 1992; Phelan & Faller, 1996). Microvascular responses characteristic of SCD, including phospholipid exposure on sickle RBCs and increased platelet procoagulant activity, may further contribute to endothelial injury (Key et al, 1998; Solovey et al, 1998; de Jong et al, 2001). Vascular tone instability and vulnerability to fluctuations in nitric oxide (NO) levels have also been implicated in the processes leading to cerebrovascular damage (French et al, 1997). Nitric oxide, an endogenous antioxidant, vasodilator, and inhibitor of platelet aggregation, is decreased in SCD. Depletion of NO has been shown to occur via haemoglobin (Hb)-mediated scavenging, consumption by reactive oxygen species and arginase-mediated substrate depletion (Aslan et al, 2001; Reiter et al, 2002; Gladwin et al, 2004; Morris et al, 2004). Although studies specific to stroke have not been performed, NO depletion has been observed in SCA patients at steady-state, and in association with vaso-occlusive pain events and acute chest syndrome (ACS) (Morris et al, 2000; Schnog et al, 2004). Related studies have also shown that plasma Hb-mediated consumption of NO and the uninhibited production of oxidants also activate inflammatory responses leading to further vascular injury (Osarogiagbon et al, 2000; Klings & Farber, 2001). Given the strong association between ACS and stroke, as well as the remarkable histologic and pathophysiologic similarities between pulmonary injury and cerebrovascular disease in SCD, it is tempting to speculate that the systemic effects of chronic haemolysis, including decreased NO bioavailability, may play a central role in the development of stroke. The age-specific progression of stroke risk in SCA during the first decade of life may be partly explained by the fact that cerebral blood flow (CBF) is developmentally regulated and highest during early childhood. The even higher CBF observed in SCD using xenon inhalation techniques is related to chronic anaemia and presumably due to hypoxia-induced vasodilation (Huttenlocher et al, 1984; Prohovnik et al, 1989). The relationship between anaemia and abnormally dilated cerebral arteries on magnetic resonance angiography (MRA) has been substantiated by demonstrating partial correction of the MRA abnormalities with transfusion (Steen et al, 2001). The end result of this vasodilatory response is a decrease in cerebral reserve capacity. Reduced vascular reserve in the presence of large vessel stenosis or systemic effects of sickled cells may further compromise perfusion to vulnerable areas of the brain resulting in distal-field infarctions. Silent infarcts are thought to result from microvascular vaso-occlusion or thrombosis, but have also been attributed to chronic hypoxia in the small vessels stemming from progressive large vessel disease (Kugler et al, 1993; Moser et al, 1996; Wang et al, 1998). SCD-Related risk factors for stroke Using data from the CSSCD, several observational studies have documented SCD-related risk factors for cerebrovascular disease in children (Table II). Although most of these risk factors showed weak associations and could not sufficiently predict stroke to justify an intervention trial, the CSSCD ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128,

4 Fig 2. Pathophysiologic mechanisms of cerebrovascular disease. Abbreviations: RBCs, red blood cells; PS, phosphatidyl serine; CAMs, cell adhesion molecules; NO, nitric oxide; ROS, reactive oxygen species. provided invaluable data on the epidemiology, natural history and untreated risk of stroke in children with SCD. Symptomatic stroke The CSSCD identified several clinical and laboratory risk factors for both ischaemic and haemorrhagic stroke (Ohene- Frempong et al, 1998). On multivariate analysis, prior transient ischaemic attack (TIA), anaemia, a past history or increased frequency of ACS, and hypertension, were independently associated with ischaemic stroke, while anaemia and leucocytosis were independent risk factors for haemorrhagic stroke. In addition to the CSSCD report of an elevated stroke risk in patients with low (steady-state) Hb levels, other studies have shown that acute anaemia with ensuing hypoxia is also associated with increased stroke risk (Ohene-Frempong, 1991; Steen et al, 1999; Setty et al, 2003). Interestingly, fetal haemoglobin (Hb F) levels were not found to confer protection from stroke, suggesting that cerebrovascular disease may differ pathophysiologically from most other systemic complications of SCD. The positive correlation between stroke and ACS found in the CSSCD also suggests a hypoxic trigger for cerebral insults. Neurologic complications, including symptomatic stroke and silent infarction, have also been reported in association with Table II. SCD-related risk factors for stroke. Stroke subtype* Risk factor Risk estimate (RR, HR or OR) 95% CI P-value Reference Ischaemic stroke Prior TIA RR ¼ 56Æ0 (12Æ0, 285) <0Æ001 Ohene-Frempong et al (1998) Anaemia (per 1 g/dl in Hb) RR ¼ 1Æ85 (1Æ32, 2Æ59) <0Æ001 Ohene-Frempong et al (1998) Recent ACS RR ¼ 7Æ03 (1Æ85, 26Æ7) 0Æ001 Ohene-Frempong et al (1998) ACS rate (per event/year) RR ¼ 2Æ39 (1Æ27, 4Æ48) 0Æ005 Ohene-Frempong et al (1998) Hypertension (per 10 mmhg ) RR ¼ 1Æ31 (1Æ03, 1Æ67) 0Æ033 Ohene-Frempong et al (1998) Silent infarcts RR ¼ 14Æ0 (1Æ95, 336) 0Æ006 Miller et al (2001) HR ¼ 7Æ2 0Æ027 Nocturnal hypoxemia HR ¼ 0Æ85 (0Æ77, 0Æ95) 0Æ003 Kirkham et al (2001a) (per 1% in SaO 2 ) Haemorrhagic Anaemia (per 1 g/dl Hb) RR ¼ 1Æ61 (1Æ11, 2Æ35) 0Æ013 Ohene-Frempong et al (1998) stroke Leucocytosis (per /l WBC) RR ¼ 1Æ94 (1Æ73, 2Æ18) 0Æ026 Ohene-Frempong et al (1998) Silent infarcts Prior seizures OR ¼ 14Æ4 (1Æ5, 141) 0Æ023 Kinney et al (1999) Painful event rate OR ¼ 0Æ53 (0Æ30, 0Æ95) 0Æ034 Kinney et al (1999) (per one event /year) Leucocytosis (WBC >11Æ /l) OR ¼ 3Æ23 (1Æ24, 14Æ37) 0Æ016 Kinney et al (1999) SEN b S globin haplotype OR ¼ 2Æ53 (1Æ03, 6Æ23) 0Æ044 Kinney et al (1999) Prior silent infarcts OR ¼ 13Æ0 <0Æ001 Miller et al (2001); Pegelow et al (2002) TIA, transient ischaemic attack; ACS, acute chest syndrome; SBP, systolic blood pressure; SaO 2, mean overnight oxygen saturation; RBC, red blood cell count; Hb, haemoglobin; RR, relative risk; HR, relative hazard; OR, odds ratio;, increase;, decrease. *CSSCD definition of stroke: acute neurologic syndrome due to vascular occlusion or haemorrhage with neurologic symptoms lasting >24 h. Silent infarcts are defined as neuroimaging evidence of cerebral infarction in the absence of neurologic symptoms. With time-dependent multivariate model. 754 ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128,

5 severe (Vichinsky et al, 2000; Henderson et al, 2003). Although the mechanisms for stroke in this setting are unclear, these findings lend credence to the role of hypoxemia, acute hypertension and cerebrovascular dysregulation in the development of stroke. The association between ACS and stroke also suggests a possible infectious aetiology for stroke. Chlamydia pneumoniae, a chief cause of ACS, has been implicated in the development of stroke in the general population. Studies investigating the association of C. pneumoniae and stroke in SCD are conflicting (Styles et al, 2001; Goyal et al, 2004). Whether persistent infection with C. pneumoniae in young children with SCD predisposes to cerebrovascular disease is currently under investigation. Miller et al (2000) assessed risk factors for disease severity in the CSSCD infant cohort. Dactylitis, anaemia, and leucocytosis were associated with an increased risk of developing severe complications, including stroke. In a separate analysis, loss of splenic function during infancy was also prognostically significant. The CSSCD data were also analysed to determine whether silent infarcts increased the risk of subsequent symptomatic stroke (Miller et al, 2001; Pegelow et al, 2002). Among the various clinical and laboratory parameters examined, the presence of silent infarcts on MRI was the strongest independent risk factor for stroke and was associated with a 14-fold increased risk of developing overt stroke. These findings are consistent with a previous report showing subclinical MRI lesions as a risk factor for clinically apparent stroke in SCD (Kugler et al, 1993). Similar results were also found among untransfused children with SCA who were enrolled in the Stroke Prevention Trial in Sickle Cell anaemia (STOP) (Pegelow et al, 2001). Taken together, these data imply that silent infarcts are progressive. However, the mechanism by which small vessel disease leads to large vessel disease is not known. It has been postulated that progression to stroke may begin with sickling and occlusion of the small vessels that supply the large arteries (Stockman et al, 1972). Alternatively, the same pathologic processes occurring in small vessels may also be active in the large vessels, but, because of the larger size of these vessels, the injury does not become clinically apparent until much later. Recent studies in other SCD patient populations have documented an independent effect of nocturnal hypoxemia on stroke risk. Kirkham et al (2001a) found that children with SCD and nocturnal hypoxemia, documented by overnight pulse oximetry, had a greater chance of developing an acute neurological event (stroke, TIA, or seizure). In combination with an increased MCA flow velocity and elevated Hb level, nocturnal hypoxemia had an additive effect on risk. A subsequent study by the same group showed that the mean overnight oxygen saturation correlated directly with the severity of intracranial vasculopathy, as determined by flow turbulence on MRA (Kirkham et al, 2001b). These findings are supported by previous reports of an association between obstructive sleep apnea and stroke in SCD (Robertson et al, 1988; Davies et al, 1989; Wali et al, 2000). Adenotonsillectomy has also been shown to reduce the risk of stroke in children with documented obstructive sleep apnea (Wali et al, 2000). A controlled trial is presently testing whether intervention in SCD patients with nocturnal hypoxemia is effective in preventing stroke and other neurologic events. Silent infarction Children enrolled in the CSSCD who showed MRI-documented evidence of silent infarcts were more likely to have a previous history of seizure and a lower rate of vaso-occlusive pain events as compared with children who had a normal MRI (Kinney et al, 1999). Leucocytosis and the Senegal (SEN) b S globin haplotype, a cluster of coinherited alleles spanning the b globin locus, were also identified as risk factors for silent infarcts in this study. In the same CSSCD cohort of children, Armstrong et al (1996) found a strong association between low steady-state Hb level and silent infarcts. Neuropsychometric performance, particularly on tests of arithmetic, vocabulary, and visualmotor speed and coordination, was also significantly worse in children with silent infarcts as compared with SCD children with a normal MRI. These lesions occur as early as 9 months of age and appear to be progressive, as those children with silent infarcts at the time of study enrollment were 13 times more likely to develop new or progressive lesions compared with children with a normal MRI (Wang et al, 1998; Pegelow et al, 2002). Risk stratification: transcranial Doppler ultrasound To date, transcranial Doppler ultrasound (TCD), a noninvasive method of measuring flow velocities in the intracranial arteries, is the single most important predictor of stroke in children with SCA (Adams et al, 1992). Due to their anaemia, children with SCD generally have higher TCD flow velocities ( cm/s) than children without SCD (90 cm/s) (Nichols et al, 2001). A time-averaged mean of the maximum (TAMM) flow velocity exceeding 140 cm/s in the dica or the proximal MCA indicates either increased CBF (appearing as widespread increases in velocity), or arterial stenosis (focal increase in velocity). Because children have thinner cranial bones and larger transtemporal acoustic windows than adults, optimal flow velocities in the dica or MCA are relatively easily detected in children. Among the 2324 children screened in the STOP study, 6% had a TCD study that could not provide a reading of the flow velocities in these vessels, in agreement with a 5% prevalence of inadequate TCD studies within our own institution (Adams et al, 2004). Three independent studies have shown that increased flow velocities in the segments most likely to reveal vasculopathy in SCA predict future stroke (Adams et al, 1992, 1997; Seibert ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128,

6 et al, 1998). Adams et al (1992) were the first to demonstrate the effectiveness of non-duplex Doppler in screening for cerebrovascular abnormalities in SCA. In 190 asymptomatic sickle cell patients prospectively screened with TCD, a TAMM velocity 170 cm/s in the dica or MCA indicated an increased risk for the development of stroke. A follow-up study confirmed that stroke risk increased as a function of TCD velocity, with the highest risk in children who had a TAMM velocity >240 cm/s (Fig 3) (Adams et al, 1997). A low (<70 cm/s) or undetectable TAMM velocity also predicted stroke. To determine the predictive value of TCD abnormalities in SCA, Seibert et al (1998) evaluated 90 children without stroke and 27 children with a history of clinical stroke. In this study, TCD screening had a sensitivity of 94%, specificity of 51%, and a positive predictive value of 36% for the subsequent development stroke. Comparison of TCD with cerebral angiography in 34 SCD patients showed that arterial narrowing 50% corresponded with TAMM velocities in the range of cm/s (sensitivity ¼ 90%; specificity ¼ 100%) (Adams, 2000a). Using TCD to select children at high stroke risk for preventive treatment with transfusions, the STOP study Fig 3. Stroke-free survival among children with SCA classified by TCD velocity. TCD classification was based on the highest time-averaged means of the maximum (TAMM) velocity in the internal carotid or middle cerebral arteries. In cases where multiple TCD evaluations were performed, follow-up was computed from the time the TCD first reached the cutoff point. Abbreviation: TCD, transcranial Doppler ultrasound. From Adams et al (1997) with permission. screened 2000 children aged 2 16 years old with SCD without a clinical history of stroke (Adams et al, 1998). In this population of children, a TAMM velocity >170 cm/s [1Æ5 standard deviations (SD) above mean] was associated with increased stroke risk. About 9% of the children screened had a TAMM velocity >200 cm/s in either the dica or MCA, indicating an increased risk of primary stroke from the baseline rate of 0Æ5 1% per year to 10 13% per year. Bilateral high TAMM velocities, in the cm/s range, were found in 40% of these high risk patients. Based on these screening results, children with a TAMM velocity >200 cm/s were randomized to receive either chronic transfusion therapy to maintain HbS levels <30% or standard care. The trial was prematurely stopped when 11 strokes were observed in the standard care arm and only one in the transfusion arm, indicating a 90% reduction in stroke risk with transfusion. Based on the seminal findings from the STOP study, the National Heart, Lung, and Blood Institute (NHLBI) issued a Clinical Alert, recommending TCD screening and consideration of chronic transfusion to prevent first clinical stroke in children with SCA who have two consecutive TAMM velocities 200 cm/s (Adams, 2000b). Follow-up data from patients screened as part of the STOP study, with normal (TAMM <170 cm/s) or conditional (TAMM cm/s) velocities, showed that the probability of conversion to an abnormal TCD study was highest in those children who were younger and whose initial TCD results were at the high end of the conditional range (Adams et al, 2004). These patients require the closest surveillance with more frequent TCD screening. Although the optimal frequency of screening has not been numerically defined, it is recommended that children be rescreened annually if the TCD study is normal, within 3 6 months if the study is conditional, and within a few weeks if the study is abnormal (Adams et al, 1997; Nichols et al, 2001). Although abnormal flow velocities in the dica and MCA detected by TCD are now established risk factors for stroke in SCA, it is unknown whether an abnormal flow velocity in other intracranial vessels predicts an increased stroke risk. Case reports of two SCA children with elevated flow velocities limited to the ACA on TCD suggest that an increased flow velocity in this vessel may also confer an increased stroke risk (Kwiatkowski et al, 2004). Bulbar conjunctival vessel velocity has also been correlated with MCA flow velocity on TCD using computer-assisted intravital microscopy (CAIM), suggesting that small vessel vasculopathy may be an indicator of large vessel disease (Cheung et al, 2002). A larger study correlating bulbar conjunctival vessel velocities with MRA or angiographic evidence of large vessel disease is needed to substantiate these findings. Kral et al recently reported an association between TCD velocity and neurocognitive functioning in 60 SCA children without a history of stroke (Kral et al, 2003; Kral & Brown, 2004). These children had a prior TCD performed as part of the STOP study and underwent neuropsychological evaluation 756 ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128,

7 to assess intellectual abilities, academic achievement, attention, and executive function. Children with the highest TAMM flow velocities (>200 cm/s) demonstrated greater executive dysfunction, particularly in the areas of inhibitory control, problem-solving flexibility and modulation of emotions, than those with intermediate ( cm/s) and normal (<170 cm/s) TAMM velocities. Although this association between TCD and neurocognitive impairment lends support to the growing evidence that links deficits in attention and executive function to progressive cerebrovascular disease in children with SCA, this relatively small exploratory study was limited by the lack of neuroimaging data needed to detect children with coexisting silent infarcts (DeBaun et al, 1998; Brown et al, 2000; Schatz et al, 2001). Future investigations will need to corroborate TCD and neurocognitive data with neuroimaging studies to determine whether TCD independently predicts cognitive impairment, particularly executive dysfunction. In another study of 84 SCD children without a history of stroke, verbal and full-scale intelligence quotients (IQ) were related to abnormal TCD, MRA and MRI studies (Kirkham et al, 2000). On univariate analysis, an abnormal TCD was associated with a below average IQ (<80). However, in the multivariate model, only height at age 3 years remained as an independent risk factor for low IQ, suggesting that early nutrition may play a significant role in the development of neuropsychological deficits. Although this pilot study included comprehensive neuroimaging data, behavioural correlates, such as executive function, were not examined. The specific abnormalities in executive function found by Kral et al (Kral et al, 2003; Kral & Brown, 2004) suggest that the frontal systems of the brain may be the most vulnerable to cerebrovascular injury in SCA. Previous MRI-based studies, showing localization of lesions to the frontal lobes in 93% of SCA children with a history of overt stroke or silent infarcts, along with data in adult SCD patients showing pervasive frontal cortex blood flow dysfunction in the absence of MRI abnormalities and altered metabolism in the frontal lobes on PET scans, support the hypothesis that the vasculopathic changes occurring in the developing brain of children with SCA are localized and progressive (Pavlakis et al, 1988; Rodgers et al, 1988; Prohovnik, 1995; Brown et al, 2000). Neuroimaging modalities for stroke surveillance Several other non-invasive imaging techniques have been investigated in an attempt to augment the predictive value of TCD screening and improve the selection of children with high stroke risk for treatment. Although silent infarction on MRI may confer an increased risk of symptomatic stroke, it does not appear to add to the effect of abnormal TCD results in predicting stroke (Adams, 2000a). However, this conclusion is based on subset analyses from the STOP study that were confounded by the high correlation between MRI and TCD abnormalities. Further studies are needed to determine if MRI can improve prediction of stroke with TCD ultrasound. Using neuropsychometric data from the CSSCD and STOP studies, Pegelow et al (2001) compared TCD and MRI as predictors of cognitive dysfunction in 78 school-aged children with SCD who had no history of overt stroke. Discordant results were found in 29% of patients, indicating that the pathogenesis of most silent infarcts differs from strokes due to large vessel disease. Although TCD was not found to accurately predict silent infarcts, this study revealed strong correlations between MRI-documented borderzone hyperintensity lesions and neuropsychologic abnormalities. In another study of 90 unaffected and 27 stroke patients with SCD, the combination of an abnormal MRA together with abnormal TCD best predicted the development of stroke (Seibert et al, 1998). Data from the STOP study confirmed a higher risk of stroke in patients having both abnormal MRA and TCD results (Abboud et al, 2004). Among the 35 patients who had an MRA performed prior to randomization in the STOP study, TCD TAMM flow velocities were significantly higher in those with severe stenosis on MRA (TAMM ¼ 308Æ3 ± 18Æ5 cm/s) compared with patients with mild stenosis or a normal MRA (TAMM ¼ 220Æ4 ±20Æ2 cm/s). However, criteria for estimating the degree of vessel stenosis based on TCD flow velocity have not been published. Nonetheless, these findings highlight the clinical value of additional imaging with MRA in all patients with abnormal or uninterpretable TCD results. Anatomical neuroimaging Cerebral computed tomography (CT) and MRI are the imaging modalities presently used to confirm a clinical diagnosis of stroke in SCA. Although CT is useful for the immediate diagnosis of intracranial haemorrhage, it may not detect an acute infarct for hours to days. Conventional MRI (cmri), using routine T1, T2-weighted imaging, is sensitive and specific for white matter and borderzone lesions, but diffusion-weighted MRI (DWI) has become the chosen modality in most centres, as it is capable of detecting acute ischaemia earlier than cmri (Steen et al, 1998). The addition of MRA to MRI studies provides critical information on the status of the cerebral vasculature in symptomatic or at-risk SCD patients, most often revealing stenotic or occlusive lesions in the dica or MCA. MRA is an accurate, non-invasive technique that, with high quality images, obviates the need for intra-arterial catheter angiography (Kandeel et al, 1996). CT angiography appears to provide greater detail than MRA, but requires intravenous contrast (Sameshima et al, 1999) and is therefore not routinely performed. MRA abnormalities may appear in SCD children as young as 7 months of age (Wang et al, 1998). As well as correlating with TCD to predict symptomatic stroke, MRA abnormalities are also associated with the development of silent infarcts (Kandeel et al, 1996; Gillams et al, 1998). ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128,

8 Functional neuroimaging In an effort to identify ischaemic damage prior to the development of lesions on conventional imaging studies, more sensitive imaging techniques have recently been developed. Several studies have assessed brain physiology in SCD with positron emission tomography (PET). Using radioactive tracers, PET indirectly measures the intactness of cortical neurons by quantifying aminobutyric acid (GABA) receptor loss, as well as glucose metabolism and microvascular blood flow. PET studies in SCD patients have shown metabolic abnormalities, particularly in the frontal lobes, and areas of low perfusion undetectable by conventional MRI (Rodgers et al, 1988; Powars et al, 1999; Reed et al, 1999). The application of PET to patients with SCD has confirmed that both CBF and cerebral blood volume are increased, while oxygen extraction is unaffected (Herold et al, 1986). These abnormalities are partially reversed by transfusion, through correction of anaemia and the effects of sickle cells. Powars et al (1999) assessed the combined use of PET and MRI in detecting cerebral dysfunction in a group of 49 children with SCA. The addition of PET to MRI identified a greater proportion of SCA children with neuroimaging abnormalities than MRI alone (90% vs. 61%). In addition, chronic transfusion therapy was shown to be beneficial in reversing impaired cerebral metabolic activity in a subset of patients. A magnetic resonancecompatible PET scanner, permitting simultaneous anatomic and functional assessment, is presently being tested in animal studies (Marsden et al, 2002). Using perfusion magnetic resonance (dynamic susceptibility contrast MRI) in children with SCD, Kirkham et al (2001b) detected perfusion abnormalities in association with various neurological symptoms, including headache, seizures, TIA and stroke. Conventional MRI and DWI failed to identify the abnormalities detected by perfusion MR in a subset of patients studied. Moreover, abnormal perfusion studies and neurological symptoms often persisted despite transfusion therapy. This study suggested the adjunctive use of perfusion MR to guide individual risk in children known to be at high risk of stroke by other measures. However, further investigations are needed to clarify how perfusion or blood flow measurements should be used to alter management in these patients. Blood oxygen level-dependent (BOLD) MRI is a new oxygen-sensitive imaging technique that measures CBF and perfusion without radioisotopes. BOLD MRI is currently under investigation for the evaluation of stroke risk in SCD. In a transgenic mouse model of SCD, BOLD results were compared with microvascular CBF (Kennan et al, 2004). Sickle mice demonstrated an abnormally increased BOLD hyperoxia response, indicating an excess of deoxyhaemoglobin, along with a concomitant decrease in CBF. The BOLD hyperoxia response complemented perfusion studies in assessing cerebrovascular function in this study. However, translation of this method to human studies is still needed to determine its clinical utility. Genetic modifiers of stroke in SCD Sickle cell anaemia is an autosomal recessive disease resulting from a single nucleotide substitution in the b globin gene (b Glu6Val,orb S ) that leads to a conformational change in the structure of the Hb molecule. Polymerization and precipitation of this abnormal Hb in the deoxygenated state causes the characteristic sickling of RBCs and ultimate clinical manifestations of SCD. Homozygosity for the b S mutation results in the most common form of SCD (HbSS, or SCA), but interaction of b S with other b globin variants or thalassaemia also leads to SCD (e.g. HbSC, HbS/b 0 thalassaemia). Even in individuals with SCA, who all share the same genotype, there is considerable heterogeneity in the phenotypic expression of this disease. In addition, a chain genotypes and b globin haplotypes may influence the clinical phenotype of SCD. SCD is the most common genetic disorder predisposing to stroke in childhood, accounting for 39% of childhood strokes in one population study (Earley et al, 1998). Although the increased risk of sudden death in military recruits with sickle cell trait is well described, stroke in individuals with sickle cell trait (HbAS) has only been sporadically reported, usually in association with dehydration, anaemia, or the perioperative period (Partington et al, 1994; Kerle & Nishimura, 1996; Drehner et al, 1999). In a recent study, arterial tortuosity was more frequently demonstrated on MRA in children with sickle cell trait compared with normal controls (Steen et al, 2003). Among children with sickle cell trait, the percentage of HbS was greater in those with this mild vasculopathy than in those who had normal vasculature on MRA. These MRA findings may partially explain the excess risk of stroke in individuals with sickle cell trait. As with other clinical complications of SCD, the aetiology of stroke is complex and undoubtedly involves the combined effects of SCD-related factors, environmental variables, and modifier genes. A familial predisposition to stroke was first suggested by the observation of an increased prevalence of stroke among siblings with SCA (Russell et al, 1984). A subsequent analysis of 42 sibships with SCA in which at least one sibling had a stroke revealed a greater than expected number of families having at least two children with stroke (Driscoll et al, 2003). In another study, SCA children having a sibling with an abnormal TCD result were 50 times more likely to also have an abnormally elevated flow velocity on TCD than those having a sibling with a normal TCD result (Kwiatkowski et al, 2003). To date, relatively few studies have identified specific genetic associations with stroke in SCA (Table III). Positive associations with specific b S globin haplotypes have been found in some studies, but not others (Powars, 1991; Miller et al, 2000; Sarnaik & Ballas, 2001). The CSSCD reported a higher prevalence of the SEN b S globin gene haplotype in SCD children with silent infarcts (Kinney et al, 1999). Although increased Hb F levels have been shown to improve the overall 758 ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128,

9 Table III. Genetic associations with ischaemic stroke risk in SCA. Risk factor Study population Risk estimate Reference Ischaemic Stroke b S -Globin haplotype CSSCD (n ¼ 2675) ns Ohene-Frempong et al (1998) FSG (n ¼ 299) ns Neonato et al (2000) Powars (1991) Hb F MCG (n ¼ 300) P <0Æ001 Adams et al (1994) CSSCD (n ¼ 2675) ns Ohene-Frempong et al (1998) FSG (n ¼ 299) ns Neonato et al (2000) Jamaica (n ¼ 310) ns Balkaran et al (1992) Los Angeles (n ¼ 537) ns Powars et al (1978) a-thalassaemia* CSSCD (n ¼ 2675) ns Ohene-Frempong et al (1998) FSG (n ¼ 299) P <0Æ001 Neonato et al (2000) MCG (n ¼ 300) OR ¼ 5Æ04 Adams et al (1994) Factor V Leiden Brazil (n ¼ 73); New Orleans (n ¼ 82) ns Kahn et al (1997); Andrade et al (1998) Prothrombin G20210A Brazil (n ¼ 73) ns Andrade et al (1998) MTHFR (C677T) Cincinnati (n ¼ 40); Jamaica (n ¼ 96) ns Balasa et al (1999); Cumming et al (1999) Homocysteine level Washington, DC OR ¼ 3Æ5à Houston et al (1997) (n ¼ 100) Cincinnati (n ¼ 40) ns Balasa et al (1999) Protein C, S activity Richmond, VA (n ¼ 13) P ¼ 0Æ049, 0Æ044 Tam (1997) London, UK (n ¼ 54) ns Liesner et al (1998) AGT (A3/A4 GT-repeats) Washington, DC (n ¼ 63) OR ¼ 4Æ0 Tang et al (2001) FCGR (2A, 3A, 3B) Jamaica (n ¼ 102) ns Taylor et al (2002a) VCAM-1 (G1238C) Jamaica (n ¼ 102) OR ¼ 0Æ35 Taylor et al (2002b) TNF (G-308A) CSSCD (n ¼ 230) OR ¼ 0Æ52 Hoppe et al (2004) ADRB2 (Q27E) OR ¼ 0Æ53 IL-4R (S503P) OR ¼ 2Æ50 HLA-A (0102, 2612, 3301) CSSCD (n ¼ 230) OR ¼ 4Æ7, 6Æ3, 0Æ25 Hoppe et al (2003) Silent Infarction b S -globin haplotype CSSCD (n ¼ 230) OR ¼ 2Æ53 Kinney et al (1999) HLA-DPB1 (0401, 1701) CSSCD (n ¼ 230) OR ¼ 3Æ0, 0Æ27 Hoppe et al (2003) VCAM1 (-T1594C) CSSCD (n ¼ 230) OR ¼ 1Æ98 Hoppe et al (2004) LDLR (NcoI-) CSSCD (n ¼ 230) OR ¼ 0Æ53 FSG, French Study Group; LAC-USC, Los Angeles County-University of Southern California Comprehensive Sickle Cell Center; MCG, Medical College of Georgia Pediatric Sickle Cell Clinic; Howard University College of Medicine Pediatric and Adult Sickle Cell Clinics; Haemoglobinopathy Clinic, Queen Elizabeth Hospital for Children, London; MCV, Medical College of Virginia, Richmond, VA, USA; CNMC, Children s National Medical Center (Washington, DC, USA); CHMC, Children s Hospital Medical Center, Cincinnati, OH, USA; Haematology- Haemotherapy Centre, State University of Campinas-UNICAMP, Campinas-SP, Brazil; Tulane University Medical Center, and the Southeastern Louisiana Sickle Cell Center, New Orleans, LA, USA. OR, odds ratio; ns, not significant. *a-thalassaemia is defined as presence of <4 a-globin genes. Stroke association with absence of a-thalassaemia. àor based on homocysteine level >10Æ1 lmol/l. clinical course of SCD, the CSSCD did not find an association between low Hb F levels and stroke or silent infarction (Balkaran et al, 1992; Ohene-Frempong et al, 1998; Kinney et al, 1999). Given that stroke occurs in only a fraction of children with SCA at an early age of onset, and is, itself, phenotypically heterogeneous, it is likely that genes outside of the b globin locus are involved. Coinherited a thalassaemia appears to ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128,

10 confer protection from stroke (Adams et al, 1994; Neonato et al, 2000; Hsu et al, 2003) possibly by improving RBC deformability and decreasing haemolysis. Previous studies, however, found that a thalassaemia was not a significant risk factor after adjustment for Hb concentration (Miller et al, 1988; Balkaran et al, 1992; Ohene-Frempong et al, 1998). Given the established human leucocyte antigen (HLA)- associated predisposition to Moyamoya disease in Japanese children, the HLA locus was studied for a potential stroke association in 230 SCA children previously enrolled in the CSSCD (Hoppe et al, 2003). Distinct HLA alleles were identified as risk factors for large vessel stroke and small vessel stroke, indicating separate pathogenetic mechanisms for these two stroke subtypes. However, these findings still need confirmation in other large cohorts or family-based studies. Small studies investigating functionally significant polymorphisms in other genes regulating immune function and inflammation, such as low-affinity Fc receptor and tumour necrosis factor-alpha (TNF-a), have failed to find an association with stroke (Carroll et al, 1998; Taylor et al, 2002a). Several case reports and case control studies have been published regarding the association of coagulation abnormalities and stroke in SCA. Prothrombotic gene polymorphisms, including Factor V Leiden, 5,10-methylenetetrahydrofolate reductase (MTHFR), glycoprotein IIIA, and prothrombin G20210A, do not appear to contribute to an increased stroke risk in SCA (Kahn et al, 1997; Andrade et al, 1998; Zimmerman & Ware, 1998; Cumming et al, 1999). Higher homocysteine levels have been associated with stroke in one report (Houston et al, 1997), but other studies found no differences in homocysteine levels, or in the MTHFR mutation frequency among controls and SCD patients with and without stroke (Balasa et al, 1999; Cumming et al, 1999). Protein C and S activity levels, already low in asymptomatic children with SCD, appear to be even further decreased in children with stroke (Tam, 1997). Given the influence of relative hypertension on stroke in SCD, angiotensinogen gene polymorphisms were examined in a small association study revealing an increased stroke risk with specific GT repeat alleles (Tang et al, 2001). The lack of positive findings from many of these limited studies has cast doubt on the importance of genetic modifiers of stroke in SCA. However, because stroke is most probably influenced by many genes, each with only modest effects, the power of even very large studies may be low. Investigation of the combined effect of multiple candidate genes may prove more successful than evaluation of single gene loci. A multilocus approach was used to identify several risk-associated genes, including VCAM-1, in a large study of children enrolled in the CSSCD (Hoppe et al, 2004). The interaction between TNF and interleukin 4 receptor (IL-4R) alleles also revealed a greater than additive risk of large vessel stroke in this population. These results highlight the combined influence of several candidate stroke susceptibility genes and suggest biological differences between stroke subtypes. Particular VCAM-1 associations with stroke risk in SCA have been reported, suggesting that several polymorphisms within the VCAM gene may jointly influence stroke risk (Taylor et al, 2002b; Hoppe et al, 2004). Recent in vitro studies have further demonstrated a functional correlation between variants in VCAM1 promoter haplotypes and specific inflammatory responses (unpublished observations). This study underscores how the identification of risk genes may help define the functional differences in pathways leading to stroke in SCA. The investigation of intermediate phenotypes to assess genetic risk factors may yield a more straightforward genetic analysis of stroke, by capturing variation in genes involved in earlier pathways leading to stroke (Gottesman & Gould, 2003). Genetic risk factors for presymptomatic vasculopathy, as assessed by TCD, are currently under investigation in a large candidate gene association study and genomewide screen (Adams et al, 2003). Preventive therapies for stroke in SCD Chronic transfusion therapy aimed at maintaining the level of HbS below 30% is the recommended treatment for primary and secondary stroke prevention in at-risk children aged 2 16 years (Adams et al, 1998). Hospital-based prevalence data have already indicated a decreased rate of symptomatic stroke in Californian children with SCD, presumably because of the implementation of TCD screening and prophylactic intervention with transfusion therapy in those identified to be at high risk (Fullerton et al, 2004). Despite its demonstrated efficacy, chronic transfusion for stroke prophylaxis is not without significant associated risks. Long-term complications of therapy include infection, alloimmunization, and iron overload requiring chelation therapy (Styles & Vichinsky, 1994; Vichinsky, 2001). For many, the potential morbidity associated with chronic transfusion therapy outweighs the 10% annual risk of stroke. Case series showing reductions in transfusion intensity, or treatment with hydroxyurea (HU) alone, have suggested that transfusion therapy may be discontinued or modified, but this needs systematic study (Cohen et al, 1992; Ware et al, 1995). An ongoing randomized controlled trial (STOP II) is presently addressing whether transfusion can be safely halted after a period of treatment during which TCD velocity reverts to normal. If this study shows that risk of stroke remains low with discontinuation of transfusion therapy after a finite period, much of the present transfusion-associated morbidity may be prevented. Haematopoietic stem cell transplantation (HCT) has proven curative in young patients with severe SCD and has resulted in stabilization of cerebral vasculopathy as documented by MRI (Walters et al, 2000). Currently, the event-free survival rate after allogeneic-matched sibling HCT for SCD is 82%. However, lack of an eligible HLA compatible sibling donor and potential transplant-related complications remain substantial barriers to HCT in SCD. Novel conditioning regimens 760 ª 2004 Blackwell Publishing Ltd, British Journal of Haematology, 128,