Introduction
Multiple sclerosis (MS) is a chronic, autoimmune-mediated, inflammatory, and neurodegenerative disease of the central nervous system that affects both adults and children. Given the diffuse pathology to cortical and subcortical structures, cognitive impairment in adults with MS is a common outcome, with memory among the most frequently disrupted domains (Nocentini et al., Reference Nocentini, Pasqualetti, Bonavita, Buccafusca, De Caro, Farina and Caltagirone2006; Olivares, Sanchez, Wollmann, Hernandez, & Barroso, Reference Olivares, Sanchez, Wollmann, Hernandez and Barroso2005; Rao, Leo, & St. Aubin-Faubert, Reference Rao, Leo and St. Aubin-Faubert1989). The impact of the disease on memory function may be even more pronounced in children than in adults with MS given the immaturity of the brain at time of disease onset.
Studies in adults with MS have shown structural and functional abnormalities in medial temporal lobe structures and in particular the hippocampus (Roosendaal et al., Reference Roosendaal, Hulst, Vrenken, Feenstra, Castellins, Pouwels and Geurts2010; Sicotte et al., Reference Sicotte, Kern, Giesser, Arshanapalli, Schultz, Montag and Boohheimer2008). One study using high-resolution 3 Tesla (T) magnetic resonance imaging (MRI) (Sicotte et al., Reference Sicotte, Kern, Giesser, Arshanapalli, Schultz, Montag and Boohheimer2008) showed that patients with either secondary-progressive MS (SPMS) or relapsing-remitting MS (RRMS) exhibited reduced bilateral hippocampal volume, particularly in the cornu ammonis I of the hippocampus, that was in excess of reductions in global brain volume relative to controls. Post-mortem studies revealed that hippocampal demyelination is frequent and extensive in MS, with a high number (n = 37) of lesions found in or around the hippocampus in 15 of 19 MS patients (Geurts et al., Reference Geurts, Bo, Roosendaal, Hazes, Daniels, Barkhof and van der Valk2007). Hippocampal gliosis using MR spectroscopy (Geurts et al., Reference Geurts, Reuling, Vrenken, Uitdehaag, Polman, Castelijns and Pouwels2006) and reduced glucose metabolism using positron emission tomography (Paulesu et al., Reference Paulesu, Perani, Fazio, Comi, Pozzilli, Martinelli and Fieschi1996) have also been revealed.
Given the pivotal role of the hippocampus in episodic memory and retrieval processes (Squire, Stark, & Clark, Reference Squire, Stark and Clark2004), the extent to which hippocampal atrophy contributes to memory impairment has been a topic of recent investigation in adult MS. In a study of 23 patients with RRMS and 11 with SPMS, regional and total hippocampal volume was associated with verbal learning (encoding of a word-list), but not with processing speed (Sicotte et al., Reference Sicotte, Kern, Giesser, Arshanapalli, Schultz, Montag and Boohheimer2008). Another study in adult MS patients (Benedict et al., Reference Benedict, Ramsamy, Munshauer, Weinstock-Guttman and Zivadinov2009) showed that volume of the thalamus, caudate, and hippocampus was most robustly associated with learning of verbal material, whereas volume of the amygdala was associated with retention of recently learned visual and verbal information. While these findings demonstrate an association between hippocampal volume and verbal learning, other studies in adult RRMS (Anderson et al., Reference Anderson, Fisniku, Khaleeli, Summers, Penny, Altman and Miller2010) fail to show a relationship between hippocampal volume and memory performance. However, this lack of correlation may reflect the restricted power of using a composite score that combined tests purported to tap different memory processes subserved by different brain regions.
The cerebral areas involved in memory—most notably frontal cortex and its connections with medial temporal lobe structures via their interactions with frontal-subcortical networks (Zola-Morgan, & Squire, Reference Zola-Morgan and Squire1993)—are also brain regions that actively mature during childhood (de Haan, Mishkin, Baldeweg, & Vargha-Khadem, Reference de Haan, Mishkin, Baldeweg and Vargha-Khadem2006). MS onset during childhood could have a particularly negative impact on these neural networks. In a study of 37 children with RRMS (MacAllister et al., Reference MacAllister, Belman, Milazzo, Weisbrot, Christodoulou, Scherl and Krupp2005), only one child (2.7%) demonstrated impairment in immediate verbal recall, while seven patients (19%) demonstrated difficulties with delayed recall of information. In contrast, both learning and retrieval deficits were noted in a study of 63 children with MS, with impaired immediate and delayed verbal recall reported in 10 (16%) and 7 (11%) children, respectively (Amato et al., Reference Amato, Goretti, Ghezzi, Lori, Zipoli, Portaccio and Trojano2008). Both studies also identified deficits in recalling visual information in approximately 10–15% of their pediatric cohorts (Amato et al., Reference Amato, Goretti, Ghezzi, Lori, Zipoli, Portaccio and Trojano2008; MacAllister et al., Reference MacAllister, Belman, Milazzo, Weisbrot, Christodoulou, Scherl and Krupp2005). In a smaller study (Banwell & Anderson, Reference Banwell and Anderson2005) of 10 children with RRMS, reduced verbal and visual memory outcomes relative to age expectations were reported. Performance was lowest among the children with a longer duration of disease (mean of 5.1 years) relative to the children with a shorter duration of disease (mean of 0.9 years), suggesting that memory impairment may emerge with increasing duration of disease.
The current study aims to extend existing knowledge in pediatric-onset MS by (i) evaluating memory function in children and adolescents with MS using tasks purported to measure the ability to learn, store, and retrieve or recognize information; (ii) measuring brain volumes for structures associated with memory function, including the hippocampus, amygdala, and thalamus; and (iii) examining the relationship between memory function and brain volumetric measures. Based on the available pediatric literature, we hypothesized that youth with MS will show difficulties in learning and in retrieval of information. On the basis of prior structure-function relationships reported in the adult MS literature, we also hypothesized that both global brain metrics and regional brain abnormalities will associate with memory performance.
Methods
Participants
The current investigation was part of a larger serial study in children with MS conducted between 2007 and 2010 at The Hospital for Sick Children (HSC) in Toronto, Canada. Participants were 32 pediatric patients with MS (25 females) diagnosed with relapsing-remitting MS (Polman et al., Reference Polman, Reingold, Edan, Filippi, Hartung, Kappos and Wolinsky2005) before age 18, and 26 age- and sex-matched healthy controls (HCs) (21 females). Patients were recruited from the Pediatric MS Clinic at HSC and HCs were recruited via local advertisement. All participants were required to be fluent in English, and to have a negative history of prior head injury (defined as a physician-diagnosed concussion or post-traumatic loss of consciousness for greater than 5-min duration), or other pre-existing medical condition known to affect brain function (e.g., alcohol or illicit drug abuse, history of a psychiatric disorder, cerebrovascular disease). Patients were not preselected on the basis of cognitive complaints and all patients were evaluated more than four weeks from any recent relapse or corticosteroid therapy. We had originally recruited and tested 30 HCs, but subsequently excluded data from three females and one male who had Full Scale IQ scores exceeding the 91st percentile to have a more appropriate comparison group for our sample of MS patients. Institutional Research Ethics Board approval and written informed consent from each subject and/or their guardian were obtained before study initiation.
For patients with MS meeting study criteria, the participation rate was 89.5% (34/38). Four patients and their families declined to participate because of disinterest in the study (n = 2) or insufficient time to participate (n = 2). These patients did not differ from those who participated with regard to age at disease onset, number of relapses, or on the basis of any sociodemographic variables. Of the 34 who were evaluated, two patients were excluded from analyses because their MRI data failed quality control (one due to artifacts from dental hardware and one due to failed quality control resulting from excessive head motion during MRI acquisition). Hence, valid neuropsychological and MRI data were obtained for 32 of the original 38 patients who were approached about the study. None of the patients included in the study experienced a clinical relapse between the MRI and neurocognitive evaluation.
Neuropsychological Testing
Participants were evaluated by a clinical neuropsychologist or trained psychometrist. The Test of Memory and Learning–Second Edition (TOMAL-2) (Reynolds & Voress, Reference Reynolds and Voress2007) was used to assess verbal and nonverbal memory function. Subtests of the TOMAL-2 have been shown to be valid and reliable measures of memory function in a variety of populations with acquired brain injury and medical disorders (Reynolds & Voress, Reference Reynolds and Voress2007). The following subtests were administered:
Word Selective Reminding (WSR)
This subtest assesses verbal learning and immediate and delayed recall of verbal information. In the learning phase, the examinee listens to a list of 12 words read by the examiner (or 8 words for children less than 9 years of age) over a maximum of six trials or until all words from the list are recalled correctly. On trials 2 through 6, subjects are reminded of the words that were missed on the preceding trial and are then instructed to recall the entire word-list again. Repetition of the list is discontinued when a subject recalls all words on the word-list; full credit is given for any subsequent trials. Thirty minutes after the learning phase, a delayed recall trial is conducted. Subjects are not informed during the learning phase that a delayed recall trial will be administered. Total number of words recalled in the learning phase (WSR-L) and in the delayed recall (WSR-D) trial served as our dependent outcomes.
Memory for Stories (MFS)
This subtest assesses memory for meaningful verbal material. The examinee is asked to listen to a story read by the examiner and then recall as many details as possible both immediately after presentation and after a delay. Total number of story units recalled on the immediate (MFS-I) and delayed (MFS-D) condition is recorded.
Abstract Visual Memory (AVM)
This subtest evaluates visual recognition memory for abstract visual designs. The examinee is shown a design for 5 s and then asked to recognize it among six abstract designs displayed immediately after. Total number recognized correctly is measured.
Facial Memory (FM)
This subtest assesses visual recognition memory for human faces. The examinee is presented with a group of faces and then asked to recognize the previously presented faces from a larger group of faces immediately after. The number of faces to remember increases on each trial. The exposure time to the target faces ranges from 5 to 20 s. Total number of faces recognized correctly is measured.
In addition to the memory tests, general intelligence was evaluated as part of a comprehensive test battery (see Till et al., Reference Till, Ghassemi, Aubert-Broche, Narayanan, Arnold, Desrocher and Banwell2011) using the Wechsler Abbreviated Scale of Intelligence (WASI) (Wechsler, Reference Wechsler1999). Symptoms indicative of depression and anxiety were assessed using the Anxiety and Depression subscales of the Self-Report form of the Behavior Assessment for Children–2nd edition (BASC-2) (Reynolds & Kamphaus, Reference Reynolds and Kamphaus2004). Expanded Disability Status Scale (EDSS) (Kurtzke, Reference Kurtzke1983) score, as a measure of physical disability, was determined for each patient by a neurologist on the same day as the MRI. Finally, fatigue (coded qualitatively as yes vs. no) was determined during an interview with the patient at time of cognitive evaluation. Fatigue was defined as a sustained feeling of physical tiredness and lack of energy being present without major exertion and present for at least 2 months.
Neuroimaging
MRI scans for all participants were performed on the same 1.5 T GE TwinSpeed Excite 12.0 scanner, according to our standardized research protocol. Two-thirds of patients and controls were scanned on the same day as the neuropsychological evaluation; the remaining one-third of participants returned for scanning within a 90-day interval. The following sequences were acquired: (i) whole brain, three-dimensional (3D) T1-weighted radiofrequency-spoiled gradient-recalled echo sequence with 1.5-mm-thick sagittal partitions, repetition time (TR) 22 ms, echo time (TE) 8 ms, excitation pulse angle 30°, 250 mm field of view (FOV) and a 256 × 256 matrix, resulting in a voxel size of 0.98 × 0.98 × 1.5 mm3; (ii) 2D multislice proton density-weighted / T2-weighted fast spin-echo sequence with echo train length = 8, TR: 3500 ms, TE1/TE2 (effective) = 15/63 ms, 2-mm-thick slices without gaps oriented parallel to the line connecting the inferior aspects of the genu and splenium of the corpus callosum, covering from above the apex of the head to the bottom of the cerebellum; and (iii) multi-slice Fast Spin Echo Fluid-attenuated Inversion Recovery (TR = 9002 ms; TE = 105 ms) with 3 mm slices and no gaps, aligned parallel to the callosal line and covering from above the top of the head to the foramen magnum.
All images were evaluated for adequate signal-to-noise ratio, freedom from significant motion or other artifact, and consistency of the sequence parameters. MRI analyses were performed at the McConnell Brain Imaging Centre of the Montreal Neurological Institute by trained staff, blinded to clinical data. A pre-processing routine was run on all images, followed by automated lesion volume segmentation with manual review and correction in supratentorial cortex on T2-weighted images and manual segmentation of infratentorial lesions. Total brain lesion volume was determined.
Two model-based segmentation methods (Aubert-Broche et al., Reference Aubert-Broche, Fonov, Ghassemi, Narayanan, Arnold, Banwell and Collins2011; Collins & Pruessner, Reference Collins and Pruessner2010) were used to identify specific brain structures, including the hippocampus, amygdala, and thalamus, in each T1-weighted image. We used the MNI Talairach-like MNI ICBM152 stereotaxic 18.5-43.5 template as the standard space (Fonov et al., Reference Fonov, Evans, Botteron, Almli, McKinstry and Collins2011). The native T1-weighted images of all subjects were automatically processed including: intensity inhomogeneity correction (Sled, Zijdenbos, & Evans, Reference Sled, Zijdenbos and Evans1998), intensity normalization, and linear registration to the template space (Collins, Neelin, Peters, & Evans, Reference Collins, Neelin, Peters and Evans1994). A hierarchical multi-scale non-linear fitting algorithm (Collins, Holmes, Peters, & Evans, Reference Collins, Holmes, Peters and Evans1995) was then applied to obtain the 3D deformation vector field that maps the individual T1-weighted images onto the template. Segmentation of all structures (except hippocampus and amygdala) was achieved by mapping template labels through the inverse of the estimated deformation onto the subject. The hippocampus and amygdala were segmented using label fusion of multiple template segmentations as illustrated in Figure 1 (Collins & Pruessner, Reference Collins and Pruessner2010). Brain volume was normalized for skull size using the scaling factor computed by SIENAx (Smith et al., Reference Smith, Zhang, Jenkinson, Chen, Matthews, Federico and De Stefano2002).
Fig. 1 Automatic segmentation of left hippocampus (blue), right hippocampus (green), left amygdala (red), and right amygdala (purple) on T1-weighted volumetric MRI of a patient with MS in (a) coronal and (b) sagittal views.
Data from the National Institutes of Health (NIH)-funded MRI Study of Normal Brain Development (also known as the NIH Pediatric database, or NIHPD) were used to norm the MRI data obtained in the current study. The NIHPD consists of MRI data collected from the largest epidemiologically ascertained sample (n > 400) of children aged 4.5 to 18 years. The T1-weighted MRI protocol used for the present study was designed to mirror that used for the NIHPD. To normalize for age and sex, the volume of each brain structure was transformed into a Z-score by subtracting native volumes for each brain structure from the average obtained for age- and sex-matched controls from the NIHPD and then dividing by their standard deviation. Ages were matched to ± 6 m. Thus, the Z-score value represents the number of standard deviations by which each subject's brain volume (or regional brain volume) was above or below the average calculated for the age- and sex-matched normal population.
Statistical Analysis
Descriptive statistics and frequency of scores that were below normal expectations were used to summarize outcomes for the MS and HC group. The Komolgorov-Smirnov's test was used to assess normality across all outcomes. Multivariate analysis of variance was then carried out on the six memory outcomes (MFS Immediate and Delayed, WSR Total and Delayed, AVM, FM scores) to examine group differences, followed by post hoc analyses of variance (ANOVA). Likelihood Ratios were used to examine group differences with regard to proportion showing difficulties on the memory tests as well as reduced brain volumes. MRI measures were analyzed with ANOVA using both absolute and normalized volumes. Spearman correlations were used to assess the relationship between clinical variables (age at first attack, disease duration), normalized MRI variables (volume of whole brain, hippocampus, thalamus, amygdala, T2-weighted lesion volume) and memory outcomes. To reduce the number of correlational analyses, left and right volumes of each segmented brain structure, which were correlated with each other (all r values > .77), were averaged. Alpha levels were adjusted using Holm's sequentially rejective Bonferroni procedure (Holm, Reference Holm1977). Effect sizes were calculated using Cohen's d (for univariate comparisons) and Cramer's V (for non-parametric comparisons). To identify structure-function associations, Spearman correlations (one-tailed) were conducted using alpha levels adjusted using Holm's procedure. Analyses were conducted using SPSS Windows version 19.0 (SPSS Inc, Chicago, IL).
Results
The mean age of patients with MS at testing was 16.3 ± 2.3 years (range: 11–19 years) and 16.4 ± 2.3 years for HC participants (range: 11–20 years), F(1,57) = 0.05, p = .82. Mean years of parental education was 15.6 ± 2.2 years for the MS group and 15.7 ± 2.0 years for the HCs, F(1,57) = 0.08, p = .78. Average age at MS onset (defined from first symptom onset) was 11.9 ± 3.8 years (range: 4–16 years) and average disease duration was 4.3 ± 3.1 years (range: 0.4–13 years). Median EDSS score was 1.0 (range: 0–4). Seven patients (23%) reported that they experienced fatigue. Groups did not differ on a self-report questionnaire assessing symptoms of Depression (MS mean = 50.08 ± 11.13; HC mean = 46.4 ± 6.64; F(1,57) = 3.28, p = .08) or Anxiety (MS mean = 54.13 ± 10.82; HC mean = 51.84 ± 10.00; F(1,57) = 0.59, p = .45).
Overall intellectual ability (Full Scale IQ) was lower in the MS group (101.22 ± 12.36) compared with controls (110.12 ± 7.45), Mann-Whitney U = 234.50, p < .004, albeit both falling within the average range. This group difference in IQ reflected reduced performances on the verbal subtests (p < .01), but not on the performance-based subtests of the WASI in the MS group.
Memory Outcomes
As illustrated in Table 1, average-range performances were demonstrated across all of the neuropsychological measures in the MS group. Multivariate analysis substantiated overall lower memory scores in the MS group compared to the control group, Wilks’ Λ = .77, F(6,55) = 2.57, p = .03. However, when IQ was covaried in the multivariate analysis, this overall difference in memory function was no longer significant, Wilks’ Λ = .77, F(6,54) = 1.48, p = .21, indicating that the MS group did not demonstrate memory impairment independent of lower IQ. Using group data, patients with MS achieved lower scores than controls on two of the six memory tests (MFS – Immediate and Delayed recall) using an adjusted alpha level (α = .025). When IQ was entered as a covariate, the differences were no longer significant. For all memory outcomes, effect sizes (ES) were medium (>.50), except for WSR outcomes which had small ES. Disease duration and age of disease onset were not correlated with any of the memory measures.
Table 1 Comparison of the neuropsychological test results for the MS and control groups using standard scores
Bold indicates significant difference between groups. MFS = Memory for Stories; WSR = Word Selective Reminding; AVM = Abstract Visual Memory; FM = Facial Memory.
*Likelihood ratio (LR) computed based on proportion of scores falling ≤1 SD or more below norm.
**Effect size calculated for univariate group comparisons using Cohen's d and proportional differences using Cramer's V.
†Data missing for one participant in MS group (test not administered).
Considering base rates, 3 of 32 (9%) patients with MS performed in the impaired range (i.e., ≥ 1.5 SD below norm), and another 4 patients (12.5%) performed in the below average range (i.e. ≥ 1 SD below norm) on the following verbal memory measures: MFS-I, MFS-D, and WSR-L. All of the patients performed at least within the age expected level on a measure of delayed, unaided recall of a word-list (WSR-D), suggesting that verbal information was consolidated well for later retrieval when it was learned through rote repetition. On nonverbal memory measures, 7 of 31 (22.6%) patients demonstrated performance that fell 1 SD or more below norm on the Abstract Visual Memory test; 6 of 32 (18.9%) patients fell in this range on the FM test. None of the control participants had an impaired score, with the exception of one control on the MFS-I test. No significant differences were identified with regard to proportion of individuals showing deficits on the various memory tests; ES for these proportional comparisons were small (<.20).
Neuroimaging Results
The MS group showed lower total brain volume (1640.281 cm3 or z = −1.33) compared with controls (1709.417 cm3 or z = 0.002) using both brain volume adjusted for skull size (F(2,57) = 8.23; p = .006) and normalized brain volume (F(2,57) = 14.26; p < .001). As shown in Table 2, the MS group also showed lower normalized volumes in the amygdala (total and left-side) and thalamus (left, right, and total), using an adjusted alpha level (α = .007). Mean amygdala volume was reduced by 7% whereas mean thalamic volume was reduced by 14% in the MS group relative to the HC group, reflecting medium-to-large (>.60) and very large ES (>1.0), respectively. Between-group differences in thalamic volume (F(2,57) = 7.33; p = .009), but not amygdala volume (F(2,57) = 3.09; p = .08), persisted even after controlling for normalized brain volume in the analysis, suggesting a disproportionate reduction in thalamic volume related to other subcortical structures. Groups did not differ significantly with respect to left, right or total hippocampal volume with mean values falling well within the normal range for this structure compared with same age-peers of the same sex. Differences in hippocampal volume between patients and controls did not change even after controlling for head size (F(2,57) = 0.32; p = .58) or total brain volume (F(2,57) = 0.01; p = .92). Brain regions in the left and right amygdalae (r = 0.77), thalami (r = 0.93), and hippocampi (r = 0.85) were all highly correlated (collapsed across groups).
Table 2 Brain MRI volumetric data for the multiple sclerosis (MS) patients and healthy control (HC) group
Bold indicates significant difference between groups.
*Brain volumes normalized using NIHPD for all participants (reported as a z-score).
In comparison with the HC group, MS patients were more likely to show reduced volumes (defined as 1 SD below the normative values measured in the NIHPD cohort) in total brain (18/32 patients vs. 4/26 controls), thalami (18/32 patients vs. 3/26 controls), and amygdalae (9/32 patients vs. 0 controls), using an adjusted alpha level (α = .005). In contrast, reduced volume in hippocampi was only observed in 3 of 32 patients (9.4%) and did not differ from controls (1/26; 3.7%).
Within the MS group, longer disease duration correlated with lower whole brain (r = −0.42; p < .05), amygdala (r = −0.37; p < .05), and thalamic volumes (r = −0.37; p < .05). Younger age at disease onset correlated with lower whole brain (r = 0.36; p < .05), amygdala (r = 0.45; p < .05), and thalamic volume (r = 0.47; p < .05). Duration of disease and age at disease onset were not correlated with hippocampal volume. However, given that younger age at onset and longer disease duration were correlated with each other in our cohort, results should be interpreted with caution and used to understand the relative strength of the associations between these clinical variables with regard to the different volumetric measures.
Structure-Function Relationships
Correlations between regional brain MRI metrics and neuropsychological measures in the MS group are shown in Table 3. Full Scale IQ correlated with thalamic (r = 0.73; p < .005) and hippocampal volume (r = 0.47; p < .01). Regarding memory outcomes, total number of words learned (WSR-L) was positively related with total hippocampal volume (r = 0.43, p < .01) and total brain volume (r = 0.52; p < .01). Of note, the correlation between WSR–L score and total brain volume remained significant even after outliers from two patients with severe brain volume loss were removed from the data (r = 0.46; p < .05). No correlations were observed with the delayed recall portion of the WSR-D test, nor the MFS test. Performance on visual recognition measures was modestly correlated with thalamic volume, reaching significance for the AVM test (p < .01). T2 lesion volume was not associated with any memory measure. Figure 2 shows scatterplots contrasting regional volume against memory performance on select measures for the MS group. Memory outcomes did not correlate with any of the regional brain MRI measures for the control group.
Table 3 Correlations (Spearman's rank correlation coefficient, one-tailed) between total and regional brain volumes and cognitive test performances in the MS and control groups
MFS-I = Memory for Stories—Immediate; MFS-D = Memory for Stories––Delayed; WSR-L = Word Selective Reminding––Learning; WSR-D = Word Selective Reminding––Delayed; AVM = Abstract Visual Memory; FM = Facial Memory; FSIQ = Full Scale IQ.
αmeasures combine left and right volumes.
* p < .05; **p < .01.
Fig. 2 Scatterplots showing positive correlations between memory performance and brain MRI volumes (presented as z-scores) for MS patients. Top graphs A and B show hippocampal and total brain volume as correlates of Word Selective Reminding – total learning score. Bottom graphs C and D show amygdala and thalamic volume as correlates of Abstract Visual Memory performance.
Discussion
Our sample of children and adolescents with MS exhibited significantly reduced volume in the thalami and amygdalae, but not in the hippocampi, when compared with a normative sample and our own age-matched comparison group. Despite normal hippocampal volume, findings revealed a moderate association between hippocampal volume and performance on word-list learning, consistent with the well-established role of the hippocampus in learning and consolidation (Squire et al., Reference Squire, Stark and Clark2004). Because performance on this word-list learning test was overall quite high in the MS group—with approximately 80% of our patient sample performing at least within the average range—it can be argued that the inclusion of more patients with deficits in verbal learning would increase the strength of this correlation.
The relative sparing of hippocampal volume in light of significant reductions in whole brain volume and other deep gray matter structures contrasts with studies of adults with MS in which hippocampal atrophy or loss of hippocampal tissue integrity is reported (Anderson et al., Reference Anderson, Fisniku, Khaleeli, Summers, Penny, Altman and Miller2010; Papadopoulos et al., Reference Papadopoulos, Sumayya, Patel, Nicholas, Vora and Reynolds2009; Ramasamy et al., Reference Ramasamy, Benedict, Cox, Fritz, Abdelrahman, Hussein and Zivadinov2009; Roosendaal et al., Reference Roosendaal, Hulst, Vrenken, Feenstra, Castellins, Pouwels and Geurts2010; Sicotte et al., Reference Sicotte, Kern, Giesser, Arshanapalli, Schultz, Montag and Boohheimer2008). A possible explanation for this observed difference between adult and pediatric MS populations might be due to differences in the disease process itself. Previous studies examining hippocampal volume in adults with MS have included patients with secondary progressive and/or primary progressive MS types (Anderson et al., Reference Anderson, Fisniku, Khaleeli, Summers, Penny, Altman and Miller2010; Roosendaal et al., Reference Roosendaal, Hulst, Vrenken, Feenstra, Castellins, Pouwels and Geurts2010; Sicotte et al., Reference Sicotte, Kern, Giesser, Arshanapalli, Schultz, Montag and Boohheimer2008), whereas our sample consisted exclusively of patients with a relapsing-remitting disease course. Our differential findings might also reflect a potentially longer period of pre-clinical activity in adults with MS (even after matching for disease duration) that may possibly contribute to greater hippocampal pathology at time of evaluation; however, differences in underlying neuropathological processes remain speculative. Further understanding of the temporal progression for hippocampal atrophy and the emergence of memory dysfunction in pediatric-onset MS will require longitudinal investigation.
Normalized whole brain volume also emerged as a robust predictor of verbal learning. The relationship may reflect the involvement of more diffuse cerebral regions when listening to aurally presented information, organizing the unstructured word-list into meaningful units, and storing the information for later, unaided recall. These processes rely not only on medial temporal lobe structures, but also interactions with frontal-subcortical networks in the frontal-striatal-pallidal-thalamo-cortical network (Van Der Werf, Jolles, Witter, & Uylings, Reference Van Der Werf, Jolles, Witter and Uylings2003; Zola-Morgan & Squire, Reference Zola-Morgan and Squire1993). Findings also illustrated that volume of the thalami, and to a lesser extent the amygdalae, was more strongly associated with performance on a visual recognition test (AVM) than on measures of verbal learning and retrieval. The strong correlation between thalamic volume and visual recognition may be explained by the involvement of the lateral geniculate nucleus, a structure in the thalamus involved in visual processing. Consistent with this idea, a recent study in adults with MS (Benedict et al., Reference Benedict, Ramsamy, Munshauer, Weinstock-Guttman and Zivadinov2009) documented a stronger correlation between the learning component of a visual memory test (Brief Visuospatial Memory Test—Revised) and thalamic volume as compared with hippocampal and amygdala volume, lending further support for the involvement of the thalamus in aspects of visual memory. These findings highlight the importance of assessing brain regions additional to medial temporal lobe structures when examining the relationship between memory performance and MRI variables.
Memory impairment is commonly reported in pediatric-onset MS patients (Amato et al., Reference Amato, Goretti, Ghezzi, Lori, Zipoli, Portaccio and Trojano2008; MacAllister et al., Reference MacAllister, Belman, Milazzo, Weisbrot, Christodoulou, Scherl and Krupp2005). Given that over half of our patients with MS showed reduced brain volume, we expected memory dysfunction to be a prominent feature in our cohort. However, on our battery of memory tests, patients with MS demonstrated poorer performance only on the Memory for Stories test. Performance on all other memory measures did not differ between groups, and the difference in story recall was not maintained after covarying intellectual ability. Notably, our results did not reveal evidence of memory impairment in the majority (80%) of the patients. In fact, results in the MS group showed that verbal information is consolidated well for later retrieval when it is learned through rote repetition. These findings raise the question of why memory function appears relatively preserved in our MS sample. To answer this question, it is important to look at the disease and demographic characteristics of our MS cohort and to compare the sensitivity of our measures to those used in prior studies.
With regard to the characteristics of our MS cohort, the average age at disease onset was 11.9 years and average disease duration was 4.3 years. In comparison to an Italian (Amato et al., Reference Amato, Goretti, Ghezzi, Lori, Zipoli, Portaccio and Trojano2008) and American pediatric MS cohort (MacAllister et al., Reference MacAllister, Belman, Milazzo, Weisbrot, Christodoulou, Scherl and Krupp2005), our patient group had a slightly lower age of MS onset and a longer disease duration—both clinical factors that have been associated with worse outcome. However, our sample was recruited from a large Canadian pediatric health-care facility that is based in an ethnically diverse region of relatively high socioeconomic status. In light of these demographic characteristics, we applied strict matching criteria for our patients and controls, which included sex, parental education, and age at evaluation. We assessed parental education using both average years of maternal and paternal education attained as well as level of education completed (i.e., college, high school, etc.). Given that parents of youth with MS had, on average, completed post-secondary education (mean of 15.6 years of education), our MS group may represent a sample from a higher-than-average socioeconomic group relative to prior studies. Moreover, the lower level of education in the TOMAL-2 normative group (with 78% of the population having less than a Bachelor degree) relative to our study cohort may have inflated memory performance in our sample.
Our study suffers from several shortcomings. First, the small sample size precludes subgroup analyses to determine whether the structure-function relationships differ with regard to disease duration or age of disease onset. Second, cognitive reserve factors might impact an individual's ability to encode and retrieve new information. For example, processing speed and attention is positively correlated verbal memory span in typically-developing children (Cowan, Wood, Keller, Nugent, & Keller, Reference Cowan, Wood, Keller, Nugent and Keller1998; Greenstein, Blachstein, & Vakil, Reference Greenstein, Blachstein and Vakil2010). Our prior study (Till et al., Reference Till, Ghassemi, Aubert-Broche, Narayanan, Arnold, Desrocher and Banwell2011) examining the same cohort of children and adolescents with MS showed that deficits in attention and processing speed are the most common symptom; thus, the contribution of reduced attention and processing speed should also be considered when examining memory function. Intellectual ability may also impact the extent of organizational and retrieval strategies used by our patient cohort to aid with learning and memory function compared to other cohorts that have lower IQ. Evidence from a normative adult sample has shown that individuals with higher IQ demonstrate more efficient encoding and retrieval operations than individuals with lower IQ (Fritsch, Larsen, & Smyth, Reference Fritsch, Larsen and Smyth2007). Third, the memory measures used in the current study may underestimate the extent of memory impairment in our patient group due to low sensitivity of the test. In particular, the nonverbal memory tests used in the current study (AVM, FM) both assess visual recognition processes. Recognition tests are considered to be less sensitive to the detection of brain insult than retrieval tests that require unaided recall of recently presented information (Haist, Shimamura, & Squire, Reference Haist, Shimamura and Squire1992; Janowsky, Shimamura, Kritchevsky, & Squire, Reference Janowsky, Shimamura, Kritchevsky and Squire1989).
Conclusion
Our study uses quantitative brain volumetric measures to better understand the neural underpinnings of learning and memory functioning in children and adolescents with MS. Reduced volume was found in the total brain, as well as the amygdalae and thalami, but not the hippocampi, as is commonly reported in studies of adults with MS. From a clinical standpoint, our investigation shows that whole brain volume is a good overall predictor of memory outcome in youth with MS, and is more robust than a measure of inflammatory disease burden (i.e., T2-lesion volume), which did not correlate with any outcome. These findings can be used to direct further studies in pediatric MS examining whether memory impairment emerges with progressive brain atrophy, and why hippocampal tissue appears preserved relative to other brain regions.
Acknowledgments
This work was supported by grants from the Canadian Institutes of Health Research, Multiple Sclerosis Society of Canada, and Multiple Sclerosis Scientific Research Foundation. The authors thank Carolynn Darryl, Julie Coleman, and Stephanie Khan for assistance with recruitment and testing of research participants, Rezwan Ghassemi for conducting the lesion volume analysis, as well as the individuals who generously contributed their time to this research. No conflicts of interest are declared.
