Detecting white matter injury in sickle cell disease using voxel-based morphometry.код для вставкиСкачать
Detecting White Matter Injury in Sickle Cell Disease Using Voxel-Based Morphometry Torsten Baldeweg, MD,1,2 Alexandra M. Hogan, PhD,1,2 Dawn E. Saunders, MD,3 Paul Telfer, MD,4 David G. Gadian, DPhil,2,5 Faraneh Vargha-Khadem, PhD,1,2 and Fenella J. Kirkham, MBBChir2,6 Objective: Sickle cell disease (SCD) is associated with cerebrovascular disease, cerebral infarction, and cognitive dysfunction. This study aimed to detect the presence and extent of white matter abnormalities in individuals with SCD using voxel-based morphometry (VBM). Methods: Thirty-six children and adolescents with SCD (age range, 9 –24 years) and 31 controls (8 –25 years) underwent magnetic resonance investigations using T1- and T2-weighted protocols. White and gray matter density maps were obtained from three-dimensional magnetic resonance imaging (MRI) data sets. Using VBM, we compared the maps between controls and SCD individuals with silent white matter infarct lesions (SCDⴙL; n ⴝ 16), and those without visible abnormality (SCDⴚL; n ⴝ 20). Results: In comparison with controls, intelligence quotients (IQs) were lower in both SCD groups irrespective of presence of visible lesions. VBM showed widespread bilateral white matter abnormalities in the SCDⴙL group, extending beyond the regions of focal infarction in the deep anterior and posterior white matter borderzones. Bilateral white matter abnormalities were also observed in the SCDⴚL group, in locations similar to those in the SCDⴙL group. Interpretation: VBM is sensitive to detection of widespread white matter injury in SCD patients in borderzones between arterial territories even in the absence of evidence of infarction. Those changes may contribute to cognitive deficits in this population. Ann Neurol 2006;59:662– 672 As mortality for sickle cell disease has decreased,1 attention has increasingly focused on the chronic adverse effects that significantly impair quality of life. In addition to a high incidence of overt stroke with a peak in childhood2 and a high recurrence rate in those who are not chronically transfused,3 covert infarction, in the absence of neurological symptoms and signs, affects up to a quarter of children with sickle cell disease (SCD) screened with magnetic resonance imaging (MRI).4,5 It has also become increasingly clear that some patients show cognitive deficits and are at risk of intellectual decline.6 Compared with patients without infarcts, those with overt or covert infarcts are at greater risk for such decline.6,7 Lesions are often unilateral, small, and focal, yet the cognitive deficits are widespread, involving different functional domains, including verbal and nonverbal intelligence. In view of these deficits, we hypothesize that there may well be additional, relatively extensive damage that is not seen on conventional imaging. We further suspect that such damage may extend to both hemispheres; otherwise, given that the le- sions are acquired early in life when the reorganizational capacity of the brain is high, we might expect rescue of function by the contralateral hemisphere.8,9 Cognitive deficits have also been found in children without obvious lesions,6,10 but whether some have brain abnormality beyond the resolution of current T2-weighted imaging remains to be addressed. Previous reports have documented subtle abnormalities in gray matter T1 on quantitative MRI,11 but the distribution across the whole brain has not yet been reported, and there are few quantitative data on white matter abnormality, although this is the site of most of the visible covert infarcts. This study was conducted to examine the distribution of covert infarcts in children with SCD using voxel-based morphometry (VBM) analysis of MRI scans. This technique can identify group differences in white and gray matter across the whole brain.12 The VBM technique has been successfully applied to MRI data sets to reveal subtle abnormalities of white and gray matter density not visible on conventional imaging in pediatric patient groups, such 1 Developmental Cognitive Neuroscience Unit, Institute of Child Health, University College; 2Great Ormond Street Hospital for Children and 3Department of Radiology, Great Ormond Street Hospital for Children; 4Department of Paediatric Haematology and Oncology, Queen Elizabeth Children’s Service, The Royal London Hospital; 5Radiology and Physics Unit, Institute of Child Health and 6Neurosciences Unit, Institute of Child Health, University College, London, United Kingdom. Current address for Dr Hogan: Developmental Brain-Behaviour Unit, University of Southampton, Southampton, United Kindgom. Received Oct 12, 2005, and in revised form Nov 16. Accepted for publication Dec 2, 2005. 662 Published online Jan 31, 2006, in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/ana.20790 Address correspondence to Dr Baldeweg, Institute of Child Health, 30 Guilford Street, London WC1N 1EH, UK. E-mail: email@example.com © 2006 American Neurological Association Published by Wiley-Liss, Inc., through Wiley Subscription Services as those with hypoxia,13 autism,14 and temporal lobe epilepsy15 and in children born very prematurely.16 Hence, we investigated the efficacy of this method in identifying white and gray matter abnormalities in patients with SCD, with or without covert infarction, and explored the possible relationship of this abnormality to intellectual abilities. Subjects and Methods Participants The patients with SCD were recruited as part of a longitudinal follow-up study of the East London cohort.7,17 Ethical permission was granted by the Great Ormond Street Hospital Research and Ethics Committee and fully informed consent was obtained from each participant. SCD was diagnosed by hemoglobin electrophoresis either on cord-blood screening at birth (n ⫽ 14) or after clinical symptoms emerged later in childhood. Patients with SCD and controls originally recruited to the study at the Institute of Child Health were invited to return for 1 day of neuropsychological and MRI assessments. A total of 108 patients (89 with hemoglobin [Hb] SS, 14 with HbSC, and 5 with Hb S␤ thalassemia) had undergone a baseline neuropsychology assessment between 1992 and 2000,7 of whom 51 (22 female, 44 HbSS, 7 HbSC) agreed to return for follow-up assessments between 2001 and 2003. Of the 24 sibling controls originally recruited, 12 agreed to undergo MRI examination during the follow-up assessment. This group was supplemented by the recruitment of 19 additional healthy nonsibling, white controls, so that the total was 31. Neuropsychology All subjects were assessed by a single researcher (A.M.H.) on the day of the MR investigation. Measures of intelligence were obtained using the Wechsler Intelligence Scale for Children–3rd edition UK (WISC-III18), for those children aged 6 years and older (12 patients, 16 controls), and the Wechsler Adult Intelligence Scales (WAIS19) for adults aged from 16 years (24 patients, 15 controls). Magnetic Resonance Imaging STRUCTURAL MAGNETIC RESONANCE IMAGING. The MRI investigation was performed on a 1.5-Tesla Siemens Vision system (Siemens AG, Erlangen, Germany) and included sagittal and coronal T1-weighted (TR, 570 milliseconds; TE, 14 milliseconds), axial TSE T2-weighted (TR, 3,458 milliseconds; TE ⫽ 96 milliseconds), and coronal turbo-fluid attenuated inversion recovery (FLAIR) T2-weighted imaging (TR, 9,999 milliseconds; TE, 119 milliseconds; inversion time TI, 2,210 milliseconds). Volume T1-weighted scans were acquired using a three-dimensional FLASH sequence (TR, 16.8 milliseconds; TE, 5.7 milliseconds; flip angle, 21 degrees; voxel size, 0.8 ⫻ 0.8 ⫻ 1.0mm). All images were evaluated by a consultant pediatric neuroradiologist, who was unaware of the status of each participant. MRI abnormalities then were classified by the neurologist (F.J.K.) as overt or covert infarction, respectively, if there was an area of increased signal intensity on T2weighted MRI with or without a history of a neurological event lasting more than 24 hours. All MRIs of SCD cases without infarct lesions (SCD⫺L) were reviewed a second time, jointly with a second consultant pediatric neuroradiologist, to confirm the accuracy of the initial evaluation. For the illustration of lesion sites across the group, lesion overlap maps were created using MRIcro software (C. Rorden, www.mricro.com). For this purpose, location and size of lesions (as volume of interest) in each case was first manually transcribed from individual axial and coronal T2-weighted images onto the same set of 12 coronal MRI slices. These slices were obtained from the Montreal Neurological Institute (MNI) single-subject T1 image (available as a stereotactic template image within SPM99 software [SPMs]) by sectioning this three-dimensional image in the coronal plane at an angulation of ⫺20 degrees with a slice thickness of 1cm. The white matter lesions of each individual patient then were transcribed into these 12 slices, and region of interest images were created for each individual. Finally, a group lesion overlap map was created by averaging across all 16 individual patient region of interest images and by color-coding areas of lesion overlap. Because the same template brain was used for both the lesion overlap maps and SPM analysis, the lesion coordinates correspond approximately to those used in SPM. LESION OVERLAP MAPS. VBM analysis was performed using SPM99 software (Wellcome Department of Imaging Neuroscience, www.fil.ion.ucl.ac.uk). Scans were normalized to the MNI template using a 16-parameter affine transformation and 4 ⫻ 5 ⫻ 4 nonlinear bias function. Images then were classified into gray matter, white matter, and nonbrain tissue, including cerebrospinal fluid, and three other background classes. A modified mixture model cluster analysis technique was used to identify voxel intensities matching particular tissue types, combined with an a priori knowledge of the spatial distribution of these tissues in normal subjects.20 This information is in the form of prior probability images, provided by the Montreal Neurological Institute, which have been derived from MRIs of 152 healthy subjects. The segmentation requires an iterative algorithm that begins with assigning starting estimates for the belonging probabilities of tissue classes based on these prior probability images. The segmentation step also incorporates an image intensity nonuniformity correction to address image intensity variations that arises for various reasons in MRI. Segmented gray and white matter images were smoothed using a 12mm full-width half-maximum Gaussian kernel, which renders data more normally distributed12 for subsequent statistical analysis on a voxel-by-voxel basis. The voxel values represent the amount of tissue per unit volume that is classified as white or gray matter. We refer to this as white or gray matter density. Data were normalized by global white or gray matter to account for any differences among the participants that are simply caused by brain size. Therefore, the resultant density measure reflects group differences in local but not global white or gray matter. Data were presented on glass brain projections in the form of SPMs showing regions where the local amounts of white or gray matter differ between two groups. SPMs were also overlaid onto VOXEL-BASED MORPHOMETRY ANALYSIS. Baldeweg et al: White Matter Injury in SCD 663 white or gray matter segments obtained by averaging across all control subjects. The statistical threshold of p ⬍ 0.05, family-wise errorcorrected for multiple comparisons across the whole brain,21 was used for regional differences where there was no prior hypothesis. Regions where there was an a priori hypothesis that white matter reductions would be present (ie, in periventricular regions) were evaluated at threshold of p ⬍ 0.001, uncorrected for multiple comparisons. No threshold for cluster size was applied. SPMs were displayed in all figures at the threshold of p ⬍ 0.001, uncorrected for multiple comparisons (details on the SPM coordinates of significant group differences at a threshold of p ⬍ 0.05, corrected for multiple comparisons, are available upon request). To visualize the distribution of individual subject data points across groups, we extracted mean parameter estimates (in arbitrary units corresponding to the likelihood that a voxel belongs to a particular tissue compartment, ie, white matter) in selected clusters of maximal z values after extracting the first eigenvariate of all voxels belonging to that cluster using a standard SPM99 function. Results Clinical Characteristics The total sample size available for the VBM analysis was 67 children divided into three groups (31 controls, including 12 siblings, 20 SCD⫺L, and 16 SCD⫹L). Of the 51 SCD patients recruited for this study, 13 did not complete the MRI examination or there were movement artifacts in the three-dimensional data sets. Another two children with overt stroke and large corticosubcortical infarcts were excluded from further analysis. Of the remaining 36 patients with SCD, 3 had had anterior transient ischemic attacks (TIAs), 7 had had posterior TIAs, 2 seizures, 16 headaches, and/or learning and behavioral difficulties, and 9 had been neurologically asymptomatic. Table 1. Demographic Characteristics of the Sample Included in the VBM Analysis SCD Silent Infarct (⫹L) Characteristic Controls No Infarct (⫺L) N Age (SD) Sex (No. female) Hb status: SC/SS Neuropsychology (SD) VIQ PIQ FSIQ 31a 15.7 (3.6) 13 — 20 17.1 (4.1) 6 6/20 16 18.2 (4.4) 8 0/16 102 (11) 101 (14) 101 (11) 93 (15)b 92 (14) 92 (14)b 84 (14)b 82 (13)b 82 (13)b a Including 12 siblings. Post hoc difference from control value significant at p⫽0.05. b VBM ⫽ voxel-based morphometry; SCD ⫽ sickle cell disease; SD ⫽ standard deviation;VIQ ⫽ verbal; PIQ ⫽ performance; FSIQ ⫽ full scale IQ. derzone areas of the anterior and posterior cerebral circulation. Neuropsychological Data Intelligence quotients (IQs) were significantly lower in both groups with SCD compared with controls (see Table 1), with effects on verbal (VIQ) (F[2,64] ⫽ 10.6; p ⬍ 0.001), performance (PIQ) (F[2,64] ⫽ 10.3; p ⬍ 0.001), and full-scale IQ (FSIQ) (F[2,69] ⫽ 13.7; p ⬍ 0.001). Significant post hoc differences ( p ⬍ 0.05) were found for VIQ and FSIQ between the conTable 2. Magnetic Resonance Imaging Findings in the SCD Groups SCD Identification of Lesions on Conventional Magnetic Resonance Imaging The 31 controls (“Control” group) and 20 of the patients with SCD had normal MRI scans on conventional neuroradiological assessment (SCD⫺L group). The “Lesion” group (SCD⫹L) consisted of 16 participants with covert lesions confined to white matter only (Table 1). There were no significant differences between the three groups in mean age (F[2,64] ⫽ 2.2, p ⫽ 0.12) and gender (2 ⫽ 1.542, p ⫽ 0.463). The distribution of lesions and presence of other MRI findings are shown in Table 2. In most cases, the silent infarct lesions clustered in the deep frontal and parietal periventricular white matter, often bilaterally, as illustrated in the lesion overlap maps (Fig 1). The MRIs from one of the cases with overt stroke are shown in Figure 2. It is of note that the white matter destruction in this case mirrors the distribution of covert infarct lesions shown in Figure 1, which cluster along the bor- 664 Annals of Neurology Vol 59 No 4 April 2006 Finding No Infarct (⫺L) N 20 MRA abnormalities present 5 Cortical atrophy None 19 Mild 1 Moderate 0 Severe 0 Distribution of silent infarct lesions Frontal Unilateral Bilateral Posterior Parietal Occipital Frontal and posterior Basal ganglia Silent Infarct (⫹L) 16 5 12 2 2 0 11 4 6 23 6 2 SCD ⫽ sickle cell disease; MRA ⫽ magnetic resonance angiography. Fig 1. Lesion overlap maps show the average location of white matter lesions in the sickle cell disease with silent white matter infarct lesions (SCD⫹L) group. Color coding indicates the number of cases overlapping in one location and slice. In each case, the location and size of the lesion was manually transcribed from individual axial and coronal T2-weighted images onto coronal slices of a single-subject T1 image. Sections are shown as indicated schematically on the right. trol and SCD⫺L group as well as on all three measures between the SCD⫺L, and SCD⫹L groups. Furthermore, intellectual abilities of the sibling control group (n ⫽ 12) were not different from those of the nonsibling control participants (FSIQ: 103 [standard deviation (SD) 8], 101  respectively; t[df ⫽ 29] ⫽ 0.56; p ⫽ 0.583), ensuring that group differences cannot be attributed to factors such as ethnicity and family environment. The same group differences in IQ scores, as described above, were observed when the statistical comparison of cases with SCD was restricted to the group of sibling controls only. IQ scores were significantly lower in both groups with SCD compared with sibling controls, affecting both VIQ (F[2,45] ⫽ 8.4; Fig 2. T2-weighted images of a 15-year-old boy who had suffered an extensive bilateral stroke involving the white matter and the overlying cortical tissue of the frontal and parietal lobes. It is of note that the white matter destruction in this case mirrors the distribution of silent infarct lesions shown in Figure 1, which cluster (indicated by arrows) along the borderzone areas of the anterior and posterior cerebral circulation. Baldeweg et al: White Matter Injury in SCD 665 Fig 3. Voxel-based morphometry comparison of white matter density between controls and the sickle cell disease with silent white matter infarct lesions (SCD⫹L) group. Regions of reduced white matter density in SCD⫹L are displayed on the mean white matter segment (A) and in a glass brain view (B). Standardized parameter estimates for a cluster of maximal T value (6.50) in the right frontal white matter (SPM coordinates indicated by crosshair in (A): x ⫽ 22, y ⫽ 36, z ⫽ 12) are shown in boxplots and as individual data points (C). SCD⫹L patients were grouped according to location of lesions in either frontal or posterior regions. “F⫹P” indicates that lesions were found in both frontal and posterior areas. p ⬍ 0.001) and PIQ (F[2,59] ⫽ 6.2; p ⫽ 0.004). Hemoglobinopathy did not significantly influence IQ scores in the SCD⫺L group (VIQ: 91 [SD 12], 96 ; PIQ: 90 , 96 ; p ⫽ 0.588 and p ⫽ 0.363, for individuals with HbSS and HbSC, respectively). Voxel-Based Morphometry First, VBM was used to evaluate the extent of white matter abnormalities in the group with identified covert infarct lesions compared with controls. Significant decreases in white matter density (ie, the amount of tissue per unit volume classified as white matter; see Subjects and Methods) were found extending along the ventricles bilaterally from the anterior frontal to parietooccipital white matter (Fig 3). The anterior frontal white matter density decreases were in a similar locations to those indicated by the lesion overlap maps (slices 3–5 in Fig 1). The decrease of anterior frontal white matter density did not appear to be dependent on the presence of lesions in frontal white matter, because similar changes were seen in those cases with exclusively posterior lesions (see boxplot in Fig 3C). Furthermore, white matter density was also decreased along WHITE MATTER ABNORMALITIES. 666 Annals of Neurology Vol 59 No 4 April 2006 the whole extent of the corpus callosum, where no overt lesions had been detected. Nevertheless, the frontal white matter changes were more pronounced in those patients with presence of both frontal and posterior lesion (see Fig 3C). Second, to explore the possibility that similar white matter changes could be detected in the absence of MRI evidence of infarction, an additional VBM analysis was conducted comparing the SCD⫺L with the control group. Indeed, similar bilateral white matter density decreases ( p ⬍ 0.001 for each hemisphere, uncorrected) were also observed in this group (Fig 4). Although the spatial extent of white matter change was less compared with that found in the SCD⫹L group, there was considerable similarity in the location of change, showing an anterior to posterior extension along the ventricles bilaterally. In contrast with the lesion group, there was much less change in the deep frontal white matter, likely because of the absence of infarcts in those regions. An additional VBM analysis was performed using only the data available from 12 sibling controls and 12 age-matched SCD⫺L cases. In agreement with the previous analysis (see Fig 4), significant white matter density reductions were found in similarly extended re- Fig 4. Voxel-based morphometry comparison of white matter density between controls and the sickle cell disease without visible abnormality (SCD⫺L) group. Regions of reduced white matter density in SCD⫺L are displayed on the mean white matter segment (A) and in a glass brain view (B). Standardized parameter estimates for a cluster of maximal T value (7.80) in the right central white matter (SPM coordinates indicated by crosshair in (A): x ⫽ 26, y ⫽ ⫺20, z ⫽ 42) are shown in boxplots and as individual data points (C). gions along the ventricles, suggesting that the observed changes are not related to factors other than SCD (data available upon request). Furthermore, to explore if the degree of white matter abnormality is directly related to the level of intellectual disability in this cohort, we computed a correlation analysis between IQ scores and white matter density (Fig 5). Although these correlations did not reach corrected levels of significance, their distribution is indeed very similar to the regions identified in the group comparisons (see Figs 3 and 4). Similar but less extensive clusters of positive correlations were also found when the analysis was restricted to the SCD groups only (see Fig 5B). GRAY MATTER ABNORMALITIES. Because the focal distribution of gray matter abnormalities11 across the whole brain has not yet been reported, changes in gray matter density were evaluated here using VBM (Fig 6). In the lesion group, areas of extensive gray matter density decrease were found along the medial wall of the frontal and parietal lobes, surrounding the cingulate sulcus and extending posteriorly into the precuneus (see Fig 6A). At a threshold of p ⬍ 0.001, uncorrected, these changes also extended onto the lateral surface of the frontal and parietal lobes. In the SCD⫺L group, some gray matter density reductions were also visible at p ⬍ 0.001 uncorrected along the medial frontal surface, with changes in small regions of the right lateral frontal lobe reaching a threshold of p ⫽ 0.05, after correction for the whole-brain volume. Discussion This study examined the distribution of brain abnormalities in children and adolescents with SCD using VBM. Compared with controls, we have shown focal abnormality in white matter density (ie, the amount of tissue per unit volume classified as white matter) in a distribution compatible with the borderzones between arterial territories in patients with sickle cell disease. Together with changes in gray matter density in focal regions of the frontal lobes, these MRI changes may contribute to the spectrum of cognitive difficulties in this population. Previous applications have demonstrated that VBM is a powerful tool to demonstrate the severity and localization of atrophy of gray matter and white matter in neurodegenerative disorders as well as of hypoplasia in developmental conditions13–16,22,23 There are, however, few data in populations at risk of chronic ischemic brain damage, in part, because VBM is currently not the most appropriate technique for characterizing Baldeweg et al: White Matter Injury in SCD 667 Fig 5. Correlation analysis between verbal (VIQ, left side) and performance intelligence quotients (PIQ, right side) and white matter density using SPM. Clusters of significant positive correlations (threshold at p value less than 0.05, two-tailed, uncorrected for multiple comparisons) are seen in distributed regions of the frontoparietal white matter: (A) when all subjects in the study (n ⫽ 67) were included as well as when (in B) the analysis was restricted to sickle cell disease individuals (n ⫽ 36). Peak correlations in the anterior deep frontal regions reached p ⬍0.0001 in both analyses. focal infarcts.24 Nevertheless, VBM showed focal gray matter loss in children with hypoxic brain injury.13 Lesion Location The distribution of white matter changes on VBM is compatible with the anatomical location of borderzone areas of the cerebral circulation between the anterior and middle cerebral arteries, and to some degree is also indicative of changes in the posterior borderzone areas between the middle and posterior cerebral arteries. Neuropathological25 and imaging studies26,27 have shown that most overt and covert ischemic strokes occur in those borderzone areas. This finding was replicated here, as shown in the lesion overlap maps in Figure 1, and the cerebral distribution of those at risk areas is amply demonstrated in the case of severe frontoparietal stroke shown in Figure 2. It is sometimes assumed that the borderzone has a uniform, wedge-like shape between the anterior and middle artery distributions (eg, Mantyla and colleagues28). However, our data are in agreement with the view that this wedgelike volume can be differentiated into two separate borderzones, a deep (medullary) and a superficial (leptomeningeal) borderzone.29 The superficial borderzone is composed of the relatively strong anastomoses between the capillaries of the cortical branches of the anterior, middle, and posterior cerebral arteries. In contrast, the deep borderzone includes the deep ganglionic branches arising from the proximal portions of the vessels and 668 Annals of Neurology Vol 59 No 4 April 2006 the associated communicating arteries that supply the deep central portion of the thalamus and cerebral hemispheres and these anastomoses are weak.29 This vascular weakness may account for the predominant clustering of covert infarct lesions as well as VBM changes along the deep borderzone. The incidence of covert infarction was 22% in the cooperative study of children with SCD aged 6 to 19 years30; the majority have abnormal MRI by the age of 6 years, but some children sustain injury later and recurrent covert and overt stroke lesions are common. Previous single-center studies have reported a prevalence of MRI abnormalities as high as 46%,31 and in the current study of a population selected through clinic attendance in 1992 to 2000 and their willingness to return for a follow-up covert infarction was found in 44% (16/36). Cognitive Impairments Since the early 1990s, several studies have associated cognitive impairment with the presence of infarction in children with SCD.32–34 Although some of these studies have reported the results of large batteries of neuropsychological assessments7,34,35 or looked specifically for deficits in attention33,36 and working memory,37 most have focused on intelligence (IQ). Covert infarcts in the frontal lobes are associated with lowered IQ, but deficits have also been found in children without infarction10,38 (for review see Schatz and colleagues39). Fig 6. Voxel-based morphometry comparison of gray matter density between controls and the sickle cell disease with silent white matter infarct lesions (SCD⫹L) (A) and sickle cell disease without visible abnormality (SCD⫺L) (B) groups. Regions of reduced gray matter density in the SCD groups are displayed on the mean gray matter segment (top) and in a glass brain view (bottom). SPM coordinates (x, y, z) of the displayed sections as indicated by arrow in the glass brains are for A: 5, ⫺5, 53, and for B: 1, ⫺13, 57. Our findings confirm those reports, including the lowered intellectual abilities in children without infarction, and also suggest that environmental factors are less important than sickle cell disease. QUANTITATIVE MAGNETIC RESONANCE IMAGING STUDIES. Our findings of more extensive cerebral abnor- mality than is visualized on conventional T2-weighted imaging complements previous studies using quantitative MRI. Steen and colleagues40,41 used a precise and accurate inversion recovery (PAIR) method for parametric T1 mapping in a single transverse slice acquired at the level of the basal ganglia. Lower T1 values were found in the gray matter of the basal ganglia, thalamus, and cortex in a cohort of children and adolescents with SCD. The position of the acquisition slice, however, was more ventral than the level at which white matter changes were seen on VBM in the current study, that is, in the centrum semiovale and corona radiata. This may explain why no such white matter deficits were observed in the studies of Steen and colleagues. The severity and extent of VBM changes is broadly commensurate with the degree of cognitive impairment in this cohort. The extensive bilateral, frontoparietal changes in both white and gray matter in the lesion group could be caused by secondary (Wallerian) degeneration of white matter tracts after infarction. However, white matter density was also decreased in the corpus callosum, where no lesions were found. In addition, the presence of deep frontal white matter changes after exclusively posterior lesions (see Fig 3C) and the white matter changes observed in the contralateral hemisphere in cases with unilateral infarcts (data not shown) suggest that a more diffuse process is involved. Such diffuse brain injury was previously suspected on the basis of cognitive data alone.31,38 Indeed, the more modest cognitive impairments and VBM changes in the group without infarct lesions may be related to chronic processes, such as sustained or intermittent ischemia or hypoxemia that may be present from early infancy. Cognitive deficits previously have been associated with subclinical cerebral vasculopathy40 and low hemoglobin level.11,42 If so, this suggests that infarcts in children with SCD might be more appropriately viewed as an additional burden on an already compromised system. VBM demonstrates an anatomical basis for the functional abnormalities which might allow cause to be in- Baldeweg et al: White Matter Injury in SCD 669 vestigated. The observed distribution of white and gray matter changes as well as of correlations with IQ scores is compatible with the functional anatomy of distributed medial and lateral frontal as well as posterior regions implicated in supporting fluid intelligence.43,44 On VBM, the density of gray and white matter in bilateral frontal, temporal, and parietal areas is associated with higher IQ in healthy adults.45 In children, who were born preterm, the decline in verbal IQ was correlated to the density in bilateral deep frontal white matter.16 In contrast with the increase in understanding of overt arterial ischemic stroke, the pathophysiology of covert infarction remains unclear. Clinical risk factors associated with silent infarction are a low pain event rate, history of seizures, high leukocyte count, and the SEN ␤s globin gene haplotype.46 Possible mechanisms include (1) large vessel disease and relative hypotension47 leading to perfusion failure in the borderzones, as for overt stroke, (2) small vessel disease or (3) microembolization, from diseased cerebral vessels or perhaps paradoxically via a patent foramen ovale, as well as the residua of (4) sinovenous thrombosis48 or posterior leukencephalopathy.49 Our data suggest that ischemic injury occurs in a borderzone distribution even in patients with no visible covert infarcts. The distribution of injury demonstrated on VBM is consistent with a hemodynamic mechanism, with critically reduced blood flow in the borderzones (for further discussion of the borderzone concept see Pavlakis and colleagues26). The neuropathological basis for the white matter changes in the SCD⫺L group is currently not known. A likely cause of lower white matter density could be a low T1-weighted signal intensity, as observed in focal areas of leukomalacia (eg, Fig. 1 in Steen and colleauges40), resulting in a reduced likelihood of voxels being classified as white matter. PATHOPHYSIOLOGICAL CONSIDERATIONS. LIMITATIONS. Although the age of our study participants spanned a range in which maturational changes are seen,50,51 when age was included as a covariate in the analyses, the current findings were not altered. However, it is likely that such analysis lacked the power to detect differences in maturation between the healthy and children with SCD, as suggested by other studies.52,53 Furthermore, the presence of small lesions, although poorly visible on T1-weighted images, may have affected the spatial normalization and segmentation of scans. Therefore, great care was taken to visually check the resulting images and to exclude all scans with subtle artifacts, largely caused by subject motion. How- 670 Annals of Neurology Vol 59 No 4 April 2006 ever, these factors cannot account for the strikingly similar findings in the SCD group without lesions. The present VBM findings are independent of possible differences in brain size and global gray matter and white matter volume. A previous investigation observed no differences in white matter volume but a significant reduction in cortical and subcortical gray matter volume in SCD children compared with controls.54 Together with the current data both studies suggest that there are global as well as local gray matter changes in the frontal lobes, whereas white matter changes appear of focal nature. Finally, we emphasize that our VBM analysis is based on group comparisons, and attempts at characterizing lesions in individual data sets may lack the sensitivity and accuracy24 required to make it a diagnostically useful tool. In summary, VBM was found sensitive to widespread white matter abnormalities in the borderzone territories in children and adolescents with SCD. This is in keeping with their neuropsychological profile which also suggests more widespread dysfunction of both verbal and nonverbal cognition. The distribution of white matter changes is in agreement with the prediction that such bilateral changes severely constrain the reorganizational capacity of the developing brain. Even more importantly, our data also suggest that the VBM method is able to detect changes in white matter density in the absence of any MRI evidence of infarction. The effects of covert pathology (anemia, silent infarction, relative systemic hypotension, pulmonary hypertension) on cognition is an important issue for the appropriate management of children with SCD, and VBM has considerable potential to address these. Further research should identify the age range during which VBM changes emerge in childhood, whether they are predictive of silent infarct or stroke and if they can be modified by other risk factors or reversed with treatment. The identification of children at risk of stroke and silent infarction using noninvasive techniques is a major goal of current research in sickle cell disease and a validated technique might be of considerable use as a surrogate marker in clinical trials. This study was funded by Action Medical Research (UK) and the Wellcome Trust and was undertaken at Great Ormond Street Hospital for Children NHS Trust, which received a proportion of its funding from the NHS Executive. We thank Drs K. Chong and T. Cox for neuroradiological expertise, J. Ho and H. Ducie for conducting the MRI investigations, Dr J. Evans for referring her patients for this study. We are grateful to all participants for their time and cooperation. References 1. Platt OS, Brambilla DJ, Rosse WF, et al. Mortality in sickle cell disease. Life expectancy and risk factors for early death. N Engl J Med 1994;330:1639 –1644. 2. Ohene-Frempong K, Weiner SJ, Sleeper LA, et al. Cerebrovascular accidents in sickle cell disease: rates and risk factors. Blood 1998;91:288 –294. 3. Powars D, Wilson B, Imbus C, et al. The natural history of stroke in sickle cell disease. Am J Med 1978;65:461– 471. 4. Pegelow CH, Macklin EA, Moser FG, et al. Longitudinal changes in brain magnetic resonance imaging findings in children with sickle cell disease. Blood 2002;99:3014 –3018. 5. Bernaudin F, Verlhac S, Freard F, et al. Multicenter prospective study of children with sickle cell disease: radiographic and psychometric correlation. J Child Neurol 2000;15:333–343. 6. Wang W, Enos L, Gallagher D, et al. Neuropsychologic performance in school-aged children with sickle cell disease: a report from the Cooperative Study of Sickle Cell Disease. J Pediatr 2001;139:391–397. 7. Watkins KE, Hewes DK, Connelly A, et al. Cognitive deficits associated with frontal-lobe infarction in children with sickle cell disease. Dev Med Child Neurol 1998;40:536 –543. 8. Vargha-Khadem F, Isaacs E, Muter V. A review of cognitive outcome after unilateral lesions sustained during childhood. J Child Neurol 1994;9(suppl 2):67–73. 9. Hogan AM, Kirkham FJ, Isaacs E. Intelligence after Stroke in Childhood: a review of the literature and suggestions for future research. J Child Psychol 2000;15:325–332. 10. Steen RG, Miles MA, Helton KJ, et al. Cognitive impairment in children with hemoglobin SS sickle cell disease: relationship to MR imaging findings and hematocrit. AJNR Am J Neuroradiol 2003;24:382–389. 11. Steen RG, Xiong X, Mulhern RK, et al. Subtle brain abnormalities in children with sickle cell disease: relationship to blood hematocrit. Ann Neurol 1999;45:279 –286. 12. Ashburner J, Friston KJ. Voxel-based morphometry—the methods. Neuroimage 2000;11:805– 821. 13. Gadian DG, Aicardi J, Watkins KE, et al. Developmental amnesia associated with early hypoxic-ischaemic injury. Brain 2000;123:499 –507. 14. Salmond CH, Ashburner J, Connelly A, et al. The role of the medial temporal lobe in autistic spectrum disorders. Eur J Neurosci 2005;22:764 –772. 15. Cormack F, Gadian DG, Vargha-Khadem F, et al. Extrahippocampal grey matter density abnormalities in paediatric mesial temporal sclerosis. Neuroimage 2005;27:635– 643. 16. Isaacs EB, Edmonds CJ, Chong WK, et al. Brain morphometry and IQ measurements in preterm children. Brain 2004;127: 2595–2607. 17. Kirkham FJ, Hewes DK, Prengler M, et al. Nocturnal hypoxaemia and central-nervous-system events in sickle-cell disease. Lancet 2001;357:1656 –1659. 18. Wechsler D. Wechsler Intelligence Scale for Children. 3rd ed. Sidcup, Kent: The Psychological Corporation, 1991. 19. Wechsler, D. Wechsler Adult Intelligence Scale. 3rd ed. London: The Psychological Corporation, 1997. 20. Ashburner J, Friston K. Multimodal image coregistration and partitioning—a unified framework. Neuroimage 1997;6: 209 –217. 21. Friston KJ, Holmes A, Poline JB, et al. Detecting activations in PET and fMRI: levels of inference and power. Neuroimage 1996;4:223–235. 22. Belton E, Salmond CH, Watkins KE, et al. Bilateral brain abnormalities associated with dominantly inherited verbal and orofacial dyspraxia. Hum Brain Mapp 2003;18:194 –200. 23. Chung MK, Dalton KM, Alexander AL, et al. Less white matter concentration in autism: 2D voxel-based morphometry. Neuroimage 2004;23:242–251. 24. Mehta S, Grabowski TJ, Trivedi Y, et al. Evaluation of voxelbased morphometry for focal lesion detection in individuals. Neuroimage 2003;20:1438 –1454. 25. Rothman SM, Fulling KH, Nelson JS. Sickle cell anemia and central nervous system infarction: a neuropathological study. Ann Neurol 1986;20:684 – 690. 26. Pavlakis SG, Bello J, Prohovnik I, et al. Brain infarction in sickle cell anemia: magnetic resonance imaging correlates. Ann Neurol 1988;23:125–130. 27. Adams RJ, Nichols FT, McKie V, et al. Cerebral infarction in sickle cell anemia: mechanism based on CT and MRI. Neurology 1988;38:1012–1017. 28. Mantyla R, Aronen HJ, Salonen O, et al. The prevalence and distribution of white-matter changes on different MRI pulse sequences in a post-stroke cohort. Neuroradiology 1999;41: 657– 665. 29. Bryan RN, Whitlow WD, Levy LM. Cerebral infarction and ischemic disease. In: Atlas SW, ed. Magnetic resonance imaging of the brain and spine. New York: Raven Press, 1996:411– 437. 30. Miller ST, Macklin EA, Pegelow CH, et al. Silent infarction as a risk factor for overt stroke in children with sickle cell anemia: a report from the Cooperative Study of Sickle Cell Disease. J Pediatr 2001;139:385–390. 31. Steen RG, Xiong X, Langston JW, et al. Brain injury in children with sickle cell disease: prevalence and etiology. Ann Neurol 2003;54:564 –572. 32. Hariman LM, Griffith ER, Hurtig AL, et al. Functional outcomes of children with sickle-cell disease affected by stroke. Arch Phys Med Rehabil 1991;72:498 –502. 33. Craft S, Schatz J, Glauser TA, et al. Neuropsychologic effects of stroke in children with sickle cell anemia. J Pediatr 1993;123: 712–717. 34. Cohen MJ, Branch WB, McKie VC, et al. Neuropsychological impairment in children with sickle cell anemia and cerebrovascular accidents. Clin Pediatr 1994;33:517–524. 35. DeBaun MR, Schatz J, Siegel MJ, et al. Cognitive screening examinations for silent cerebral infarcts in sickle cell disease. Neurology 1998;50:1678 –1682. 36. Brown RT, Davis PC, Lambert R, et al. Neurocognitive functioning and magnetic resonance imaging in children with sickle cell disease. J Pediatr Psychol 2000;25:503–513. 37. White D, Salorio C, Schatz J, DeBaun MR. Preliminary study of working memory in children with stroke related to sickle cell disease. J Clin Exp Neuropsychol 2000;22:257–264. 38. Steen RG, Fineberg-Buchner C, Hankins G, et al. Cognitive deficits in children with sickle cell disease. J Child Neurol 2005;20:102–107. 39. Schatz J, Finke RL, Kellett JM, et al. Cognitive functioning in children with sickle cell disease: a meta-analysis. J Pediatr Psychol 2002;27:739 –748. 40. Steen RG, Reddick WE, Mulhern RK, et al. Quantitative MRI of the brain in children with sickle cell disease reveals abnormalities unseen by conventional MRI. J Magn Reson Imaging 1998;8:535–543. 41. Steen RG, Langston JW, Ogg RJ, et al. Diffuse T1 reduction in gray matter of sickle cell disease patients: evidence of selective vulnerability to damage? Magn Reson Imaging 1999;17: 503–515. 42. Kirkham FJ, Hogan AM. Risk factors for arterial ischemic stroke in childhood. CNS Spectrums 2004;9:451– 464. 43. Gray JR, Chabris CF, Braver TS. Neural mechanisms of general fluid intelligence. Nat Neurosci 2003;6:316 –322. 44. Duncan J, Seitz RJ, Kolodny J, et al. A neural basis for general intelligence. Science 2000;289:457– 460. Baldeweg et al: White Matter Injury in SCD 671 45. Haier RJ, Jung RE, Yeo RA, et al. Structural brain variation and general intelligence. Neuroimage 2004;23:425– 433. 46. Kinney TR, Sleeper LA, Wang WC, et al. Silent cerebral infarcts in sickle cell anemia: a risk factor analysis. The Cooperative Study of Sickle Cell Disease. Pediatrics 1999;103: 640 – 645. 47. Pegelow CH, Colangelo L, Steinberg M, et al. Natural history of blood pressure in sickle cell disease: risks for stroke and death associated with relative hypertension in sickle cell anemia. Am J Med 1997;102:171–177. 48. Sebire G, Tabarki B, Saunders DE, et al. Cerebral venous sinus thrombosis in children: risk factors, presentation, diagnosis and outcome. Brain 2005;128:477– 489. 49. Henderson JN, Noetzel MJ, McKinstry RC, et al. Reversible posterior leukoencephalopathy syndrome and silent cerebral infarcts are associated with severe acute chest syndrome in children with sickle cell disease. Blood 2003;101:415– 419. 672 Annals of Neurology Vol 59 No 4 April 2006 50. Steen RG, Ogg RJ, Reddick WE, et al. Age-related changes in the pediatric brain: quantitative MR evidence of maturational changes during adolescence. AJNR Am J Neuroradiol 1997;18: 819 – 828. 51. Paus T. Mapping brain maturation and cognitive development during adolescence. Trends Cogn Sci 2005;9:60 – 68. 52. Steen RG, Schroeder J. Age-related changes in the pediatric brain: proton T1 in healthy children and in children with sickle cell disease. Magn Reson Imaging 2003;21:9 –15. 53. Steen RG, Hunte M, Traipe E, et al. Brain T1 in young children with sickle cell disease: evidence of early abnormalities in brain development. Magn Reson Imaging 2004;22: 299 –306. 54. Steen RG, Emudianughe T, Hunte M, et al. Brain volume in pediatric patients with sickle cell disease: evidence of volumetric growth delay? AJNR Am J Neuroradiol 2005;26:455– 462.