The Cerebellum

, Volume 12, Issue 6, pp 882–891

Socioeconomic Status and the Cerebellar Grey Matter Volume. Data from a Well-Characterised Population Sample


    • Sackler Institute of Psychobiological Research, Institute of Health and WellbeingUniversity of Glasgow
    • Institute of Mental Health and Wellbeing, College of Medical, Veterinary and Life Sciences, Administration BuildingGartnavel Royal Hospital
  • Rajeev Krishnadas
    • Sackler Institute of Psychobiological Research, Institute of Health and WellbeingUniversity of Glasgow
  • G. David Batty
    • Social and Public Health Sciences UnitMedical Research Council
    • Clinical Epidemiology Group, Department of Epidemiology and Public HealthUniversity College London
  • Harry Burns
    • Scottish Government
  • Kevin A. Deans
    • Department of Clinical Biochemistry, NHS Greater Glasgow and ClydeGlasgow Royal Infirmary
    • Department of Clinical BiochemistryAberdeen Royal Infirmary
  • Ian Ford
    • Robertson Centre for BiostatisticsUniversity of Glasgow
  • Alex McConnachie
    • Robertson Centre for BiostatisticsUniversity of Glasgow
  • Agnes McGinty
    • Glasgow Clinical Research Facility
  • Jennifer S. McLean
    • Glasgow Centre for Population Health
  • Keith Millar
    • Sackler Institute of Psychobiological Research, Institute of Health and WellbeingUniversity of Glasgow
  • Naveed Sattar
    • Institute of Cardiovascular and Medical Sciences, College of Medical, Veterinary and Life SciencesUniversity of Glasgow
  • Paul G. Shiels
    • Institute of Cancer Sciences, College of Medical, Veterinary and Life SciencesUniversity of Glasgow
  • Carol Tannahill
    • Glasgow Centre for Population Health
  • Yoga N. Velupillai
    • Graduate Entry Medical SchoolUniversity of Limerick
  • Chris J. Packard
    • Glasgow Clinical Research Facility
  • John McLean
    • Sackler Institute of Psychobiological Research, Institute of Health and WellbeingUniversity of Glasgow

DOI: 10.1007/s12311-013-0497-4

Cite this article as:
Cavanagh, J., Krishnadas, R., Batty, G.D. et al. Cerebellum (2013) 12: 882. doi:10.1007/s12311-013-0497-4


The cerebellum is highly sensitive to adverse environmental factors throughout the life span. Socioeconomic deprivation has been associated with greater inflammatory and cardiometabolic risk, and poor neurocognitive function. Given the increasing awareness of the association between early-life adversities on cerebellar structure, we aimed to explore the relationship between early life (ESES) and current socioeconomic status (CSES) and cerebellar volume. T1-weighted MRI was used to create models of cerebellar grey matter volumes in 42 adult neurologically healthy males selected from the Psychological, Social and Biological Determinants of Ill Health study. The relationship between potential risk factors, including ESES, CSES and cerebellar grey matter volumes were examined using multiple regression techniques. We also examined if greater multisystem physiological risk index—derived from inflammatory and cardiometabolic risk markers—mediated the relationship between socioeconomic status (SES) and cerebellar grey matter volume. Both ESES and CSES explained the greatest variance in cerebellar grey matter volume, with age and alcohol use as a covariate in the model. Low CSES explained additional significant variance to low ESES on grey matter decrease. The multisystem physiological risk index mediated the relationship between both early life and current SES and grey matter volume in cerebellum. In a randomly selected sample of neurologically healthy males, poorer socioeconomic status was associated with a smaller cerebellar volume. Early and current socioeconomic status and the multisystem physiological risk index also apparently influence cerebellar volume. These findings provide data on the relationship between socioeconomic deprivation and a brain region highly sensitive to environmental factors.


CerebellumSocioeconomic statusDeprivationCognition


Recent data suggest that environmental factors may influence cerebellar grey matter volume more than other brain regions. Overall, heritable influences on cerebellar volume are less than for other brain regions particularly during development and early years of life [1, 2]. It also has a more protracted period of development, rendering it more vulnerable to the potential adverse effects of environmental factors [3]. It is well recognised that the cerebellum is also particularly vulnerable to toxins and other environmental poisons, with the cerebellar cortex and Purkinje neurons being especially at risk. For example, the rapidly developing cerebellum is susceptible to nutritional deficiencies like protein–energy malnutrition and zinc deficiency [4]. In humans, the most common toxin that is attributed to cerebellar damage is alcohol [5]. However, other risks include pharmaceutical agents like anticonvulsants and anti-neoplastic agents, other drugs of abuse like cocaine and heroin, and environmental toxins like mercury, lead, manganese and toluene/benzene derivatives [6]. As well as illustrating the effects of adversity on the cerebellum, these data also add to the growing awareness of the role of the cerebellum in cognition [79]. Taken together, these data suggest that cerebellar grey matter volume may be susceptible to a number of environmental factors throughout the life span.

Living in poverty is often associated with a reduction in environmental enrichment. Hackman and Farah emphasise the significance of the effect of socioeconomic status (SES) on neural systems [10]. Epidemiological studies have revealed an association between early life and adult individual level as well as neighbourhood level deprivation and raised inflammatory and cardiometabolic risk [1113]. Raised inflammatory and cardiometabolic risk markers in turn have been associated with poor cognitive functioning in adults [14, 15]. A number of reports have documented SES gradients in multi-system physiological indices like allostatic load (AL). Gruenewald et al. [16] have argued that multi-system AL indices may provide a more useful picture of the physiological toll that SES adversity takes on the body. It has already been demonstrated that AL levels are higher in lower SES groups.

The cerebellum is also particularly sensitive to high levels of glucocorticoids produced by either exogenous administration or in response to the stress induced, for example, by animal models of maternal separation [17]. In humans, Bauer et al. found that neglected children had smaller cerebellar volumes than those who were not neglected and that greater volume appeared to mediate better neuropsychological test performance [18].

Given that a specific aspect of deprivation—neglect—has been shown to have a detrimental effect on the cerebellar volume of children, it raises the question of whether more general environmental deprivation as may be seen in socioeconomic deprivation may also have such an effect on cerebellar volume.

If indeed there is a relationship between socioeconomic deprivation and cerebellar volumes, could greater inflammatory/cardiometabolic risk mediate the relationship between the two variables?

The aim of our study was to examine the relationship between early-life and current socioeconomic status and cerebellar volume. A second aim was to examine if a single multisystem physiological risk index (MPRI) derived from various inflammatory and cardiometabolic risk indices that was found to be highly correlated with SES mediated the relationship between SES and cerebellar volumes.

Materials and Methods


Participants were recruited as part of a larger study (Psychological, Social and Biological Determinants of Ill Health (PSoBiD); Details of the design of PSoBiD have been described elsewhere [1922]. Selection of participants was based on the Scottish Index of Multiple Deprivation 2004 (SIMD), which ranks small areas on the basis of multiple deprivation indicators across six domains, namely: income; employment; health; education, skills and training; geographic access and telecommunications; and housing. Sampling was stratified to achieve an approximately equal distribution of the 666 participants across males and females and age groups (35–44, 45–54 and 55–64 years) within the most (bottom 5 % of SIMD score) and least deprived areas (top 20 % of SIMD score). Participants could opt in for the neuroimaging component of the study. From a total of 327 male participants, 140 volunteered and 42 (21 from least deprived and 21 from most deprived areas) of these were randomly selected. Among the exclusion criteria for this neuroimaging component were neurodegenerative disorders (including family history of degenerative disorders), head injury, stroke, alcohol and drug misuse. The present paper presents the results from the 42 individuals. All participants gave informed consent and ethical approval for the study was obtained from Glasgow Royal Infirmary Research Ethics committee.

Mediators and Other Variables of Interest (Table 1)

Early Life and Current SES (Table 2)

Correspondence analysis was used to explore the factor structure of early and late SES [23]. This is similar to factor analysis for categorical data. These analyses confirmed that markers of early and late SES are well represented by single factors, and determined the corresponding weight associated with each level of each marker. By taking levels with positive and negative weights as representing relative deprivation or affluence, the following cut-offs were used to derive early and late SES scores: early life SES (ESES) consisted of the following items: number of siblings (>3 = 0), people per room (>1 = 0), paternal social class (IIIM or below = 0), parental housing tenure (not the owner = 0) and use of a car by the family (no car = 0). The current SES (CSES) score was derived from current income (<25,000 = 0), current social class (III or lower = 0) and current housing tenure (not the owner = 0). For each variable, those deemed to be least deprived scored 1 and those deemed to be most deprived scored 0. These scores were then summed for each, giving total score (0–5 for ESES, 0–3 for CSES), higher scores suggesting more affluence. We did not have data on current income from one of the participants. Excluding this person from the rest of the analysis did not change our results.
Table 1

Demographic and clinical characteristics of study participants






Age (years)





Alcohol—number of units per week





Glucose (mmol/L)





TRIG (mmol/L)





HDL (mmol/L)





Systolic BP (mmHg)





Diastolic BP (mmHg)





BMI (kg/m2)





Waist Hip ratio





Cortisol (nmol/L)





hsCRP (mg/L)





ICAM (ng/ml)





IL6 (pg/ml)





Insulin (uIU/ml)





Fibrogen (g/L)






























Stroop test





Choice reaction time





Trail making test A





Trail making test B





ICV (mm3)










BMI body mass index, CRP C-reactive protein, IL-6 interleukin-6, ICAM-1 intercellular adhesion molecule, Stroop test time difference, AVLT5 the fifth recall learning trial of the auditory verbal learning test

Table 2

Variables included in calculating socioeconomic status

Early life socioeconomic status

N = 42

Childhood overcrowding (no. people per room at age 11)





Fathers social class (n = 40)













Parents tenure status at age 11



Not the owner


Parents owned car at age 11





Number of siblings



3 or more


Current socioeconomic status

  Current social class (n = 41)













  Current income (n = 41)











  Current tenure status

Owner occupier




I professional, II managerial, IIIM skilled—manual, IIINM skilled—non-manual, IV partly skilled, V unskilled

Blood Sample Analysis

Ten to 12 h fasting morning blood samples were collected, separated and frozen at −80 °C within 1 h of venepuncture, except for samples for triglycerides, high density lipoprotein (HDL), high sensitivity C-reactive protein (CRP) and glucose, which were analysed on fresh plasma. High-sensitivity CRP was measured by an immunoturbidimetric assay (Roche Diagnostics Ltd., Burgess Hill, UK) and had a coefficient of variation (CV) of less than 3 %. Interleukin-6 (IL-6) and Intercellular Adhesion Molecule-1 (ICAM) were measured by sandwich enzyme-linked immunosorbent assay (ELISA; R&D Systems Europe Ltd., Abingdon, UK). The between batch CV for IL-6 was 8.3 % at a concentration of 2.84 pg/mL and 10.0 % at 5.38 pg/mL. The between batch CV for sICAM-1 was 5.5 % at an analyte concentration of 190 ng/mL and 8.1 % at 240 ng/mL. Fibrinogen was measured on an automated coagulometer (MDA-180, Organon Teknika, Cambridge, UK) with a between batch CV of 3.7 % at a fibrinogen concentration of 2.89 g/L. Cholesterol and triglyceride were determined by enzymatic colorimetric assays on a Roche 917 analyser (Roche Diagnostics Ltd., Burgess Hill, UK). All the lipid analysis had a between batch CV of less than 3 %. Glucose was measured by hexokinase/glucose-6-phosphate dehydrogenase assay on an Abbott c8000 analyser (Abbott Diagnostics, Maidenhead, UK). Between-batch CVs ranged (on the different Abbott c8000 analysers used) from 1.13 to 1.89 % at a glucose concentration of 3.23 mmol/L; from 1.10 to 1.45 % at 6.42 mmol/L and from 0.83 to 1.83 % at 20.4 mmol/L. Insulin was measured by a sandwich ELISA (Mercodia AB, Uppsala, Sweden). Between-batch analytical CV was 7.26 % at 6.04 mU/L and 7.85 % at 11.2 mU/L. Cortisol was measured by immunoassay on an Abbott c8000 analyser (Abbott Diagnostics, Maidenhead, UK). Between-batch analytical CV was 6 % at 99.7 nmol/L, 5.7 % at 522.8 nmol/L and 4.5 % at 897.7 nmol/L

Multisystem Physiological Risk Index

Risk indices were calculated from various biomarkers measured as described above. High sensitive C-reactive protein, cortisol, ICAM, fibrinogen, dDimer, IL6, blood glucose, BMI, WHR, systolic BP, diastolic BP, triglycerides and reversed Z score of HDL were used to calculate a risk index, based on their averaged Z scores. It was assumed that pro-inflammatory effects are in the same direction, i.e., a higher level of these markers would be pro-inflammatory. In this context, although cortisol is generally considered to be anti-inflammatory, taking into consideration the role of cortisol in mediating the effects of stress on brain structures, a greater level of cortisol was considered to be detrimental [24]. The Z scores were computed from the raw values of the above markers and were averaged over each individual to provide the MPRI. Greater score on the MPRI was considered to be suggestive of poorer health. Details of individual components of MPRI for the 42 subjects that underwent MRI are shown in Table 1.

Cognitive Assessment

The test battery was designed to assess the principal cognitive domains of memory, reaction and decision processes, and executive function. Executive function: the stroop test (the time taken to complete the colour–word task is reported) and trail making test (the time taken to complete trails A and B test are reported) [25, 26]. Choice reaction time: five-choice visual reaction time was measured in milliseconds by a computerised system. The total thinking and reaction time are reported here [27]. Memory: the Auditory Verbal Learning Test (AVLT). We report the score on trial 5, as an index of verbal memory and learning [28]. Participants also completed the National Adult Reading Test (NART-II) which provides an index of the individual’s peak achieved level of intellectual function, often termed an estimate of pre-morbid intelligence [29]. They also completed the 28-item GHQ consisting of four subscales: somatic symptoms, anxiety and insomnia, social dysfunction and severe depression, as an index of current mental health [30].

Image Acquisition

All MRI examinations were performed using GE Medical systems, 3T Signa Excite HD system (Milwaukee, USA) with an eight-channel phased array (receive only) head coil. An axial 3D T1-weighted IR-FSPGR was acquired with the following imaging parameters: TR = 6.8 ms, TE = 1.5 ms, inversion preparation time = 500 ms, flip angle = 12°, FOV = 26 cm, phase FOV = 70 %, matrix = 320 × 320, bandwidth = 31.25 kHz, number of slices = 160 and slab thickness = 1 mm. The acquisition time for this scan was 8 min 54 s.

Freesurfer Volume Extraction

FreeSurfer (FS) is a widely used MRI image analysis package ( The component of the FreeSurfer processing pipeline of most interest here was the subcortical structure segmentation. We used FreeSurfer version 5.0 on a Linux platform. The volume-based stream is designed to pre-process MRI volumes and label sub-cortical tissue classes. The stream consists of five stages (fully described in Fischl et al. [31, 32]). The first stage is an affine registration with Talairach space specifically designed to be insensitive to pathology and to maximise the accuracy of the final segmentation (a different procedure than the one employed by the surface-based stream). This is followed by an initial volumetric labelling. The variation in intensity due to the B1 bias field is corrected (again using a different algorithm than the surface-based stream). Finally, a high dimensional nonlinear volumetric alignment to the Talairach atlas is performed. After the pre-processing, the volume is labelled (see below). The volume-based stream has evolved somewhat independently from the surface-based stream. The volume-based stream only depends upon the skull stripping to create a mask of the brain in which the labelling is performed. The final segmentation is based on both a subject-independent probabilistic atlas and subject-specific measured values. The atlas is built from a training set, i.e., a set of subjects whose brains have been labelled by hand. These labels are then mapped into a common space (Talairach space for volumes) to achieve point-to-point correspondence for all subjects. Note that a “point” is a voxel in the volume or a vertex on the surface. At each point in space, there exists the label that was assigned to each subject and the measured value (or values) for each subject. Three types of probabilities are then computed at each point. First, the probability that the point belongs to each of the label classes is computed. The second type of probability is computed from the spatial configuration of labels that exist in the training set, which is termed the neighbourhood function. The neighbourhood function is the probability that a given point belongs to a label given the classification of its neighbouring points. The neighbourhood function is important because it helps to prevent islands of one structure in another at the structure edges. Third, the probability distribution function (PDF) of the measured value is estimated separately for each label at each point. For volume-based labelling, the measured value is the intensity at that voxel. For surfaced-based labelling, the measured value is the curvature in each of the principal directions at that vertex. The PDF is modelled as a normal distribution, so we only need to estimate the mean and variance for each label at each point in space. If there is more than one measured value, then the PDF is modelled as a multivariate normal for which we need to estimate the mean and variance–covariance matrix for each label [31, 32]. The FS image-processing pipeline ran automatically and was visually inspected at critical points in the processing pipeline (after the initial motion correction and skull stripping; autorecon1) and sub-cortical segmentation (autorecon2; in order to avoid errors permeating through the subsequent analyses). Cerebellar volumes were measured in cubic millimeter.

Data Analysis

Regression and Mediation Analysis

Multiple regression analysis was employed to examine the contribution of potential predictors on cerebellar volumes. We included four predictors of interest in our model. They were ESES, CSES, MPRI and the number of alcohol units consumed. Age and intracranial volume were included as nuisance covariates. Relative importance metrics were calculated using the relaimpo package in R [33]. This is an explanatory approach to determine the importance of each predictor variable in the full model (containing all predictors of interest). In theory, comparing the univariate contributions of each predictor could test this. However, in situations where there is significant correlation between predictors, the sum of univariate r2 will be much larger than the r2 in the full model. Therefore, calculating relative importance of each variable with respect to the other requires computational techniques that break down the full model r2. We used the Lindeman, Merenda and Gold method described by Lindeman et al. [34]. Briefly, this technique decomposes r2 by calculating the contribution of each predictor (the squared semi-partial) at all possible points of entry into the model and takes the average of those. In order to examine if ESES or CSES explained any variance over and above each other, we also conducted a hierarchical regression analysis.

We also conducted additional mediation analysis to see if the relationship between ESES and volumes were mediated by the risk index. The analysis tests the hypothesis that the SES accounts for variance in the mediator and in turn, this variance in the mediator accounts for a proportion of the variance of the cognitive deficits. In other words, the mediation analysis partitions the variance explained by the predictor into a part that is independent of the mediating variable (direct), and a part that is accounted for via the mediating variable (indirect; Fig. 1). We used the bootstrap method of Preacher and Hayes to estimate the indirect effect and bias-corrected 95 % confidence interval (CI) for each individual mediator based on 20,000 bootstrap samples using an SPSS macro [35]. This analysis requires no assumption regarding the underlying distributions since the statistical significance level is determined non-parametrically. Finally, we examined if there was correlation between cerebellar grey matter volume and performance on the above cognitive tests.
Fig. 1

The figure depicts the relationship between the predictor, mediator and the outcome variables (cerebellar grey matter volume). The mediation analysis partitions the total variance (total effect) explained by the predictor (socioeconomic status) into a part that is independent of the mediating variable (multi-system physiological risk index; direct effect), and a part that is accounted for via the mediating variable (indirect effect)


Demographics and details of other potential mediators are shown in Table 1. Relative importance of each predictor on the networks is shown in Fig. 2. After excluding ICV and age (the nuisance covariates), the model predicted 32.5 % of the variance. ESES (12.3 %) and CSES (11.4 %) together explained the greatest variance. The univariate relationship between SES and cerebellar grey matter volume is shown in Fig. 3. On hierarchical regression, CSES explained significant additional variance to ESES on grey matter volume (ΔR2 = 0.069, F change = 5.67, p = 0.02). However, ESES did not explain significant additional variance to CSES on grey matter volume (ΔR2 = 0.022, F change = 1.97, p = 0.17).
Fig. 2

Relative importance of predictors. The total variance explained (R2) is decomposed into contributions by individual predictors. Values shown are interpreted as percentage variance explained. MPRI multisystem physiological risk index, ESES early socioeconomic status, CSES current socioeconomic status. The numbers represent the relative s2 contribution (effect size) of each predictor
Fig. 3

Relationship between cerebellar volume and SES. ESES early socioeconomic status, CSES current socioeconomic status. Cerebellar volume are presented as standardised residuals after covarying for age, ICV and alcohol use

On univariate regression, both ESES (r2 = 0.17, beta = −0.11, t = −2.88, p = 0.006) and CSES (r2 = 0.14, beta = −0.12, t = −2.58, p = 0.01) predicted MPRI. In addition, MPRI predicted cerebellar grey matter volume (r2 = 0.16, beta = −0.4, t = −2.79, p = 0.008). MPRI mediated the relationship between ESES and grey matter volume (β = 1,055.10; SE, 615.63; 95 % CI = 41.56–2,477.78) as well as CSES and grey matter volume (β = 998.31; SE, 847.84; 95 % CI = 0.53–3,275.78).

Results of the correlation between cerebellar volumes and cognitive test performance are shown in Table 3. Performance on choice reaction time test and trail-making tests, correlated negatively with grey matter volume. Performance on the AVLT also correlated with the cerebellar grey matter volume (Table 3).
Table 3

Correlation of cognitive tests with cerebellar grey matter volumes


Cerebellar grey matter


Pearson correlation


Sig. (two-tailed)





Pearson correlation


Sig. (two-tailed)





Pearson correlation


Sig. (two-tailed)





Pearson correlation


Sig. (two-tailed)




Strp Stroop test (time difference), Tra trail making tests, CRT choice reaction time, AVLT5 the fifth recall learning trial of the auditory verbal learning test

*p < 0.05


The findings in this study were as follows: individual level socioeconomic status predicted cerebellar volumes. Lower SES was associated with smaller cerebellar grey matter volumes. Both early life and current socioeconomic status explained the greatest variance. On regression analysis, low current SES explained significant additional variance to low early SES on grey matter volume. Conversely, low early SES had no additional effect to current SES on grey matter volume. MPRI mediated the relationship between early/current life SES and grey matter volume in cerebellum.

There are a number of approaches to understanding the relationship between SES across the life course and health outcomes in adulthood: the ‘critical period model’, where exposure to deprivation during particular periods of development has long-term irreversible effects on adult health independent of adult circumstances; the ‘pathway model’, where the early life environment by itself does not directly have an effect on health outcomes, but sets people on a life trajectory that affects health status in adulthood; the ‘cumulative model’, where exposure to unfavourable social environments throughout life affects health status in a dose-dependent fashion [36]. While the results of our relative importance statistics found that both ESES and CSES explain almost equal variance on cerebellar grey matter volume, the results of our hierarchical regression suggest that CSES explained additional variance independent of ESES on CGMV, while ESES did not. While this is inconsistent with the critical period hypothesis, the cross-sectional nature of our cerebellar volume data makes it difficult to differentiate between a pathway and a cumulative model. In addition, in our sample, ESES and CSES showed significant correlation with each other. That is, those who scored low on ESES also tended to score low on CSES scores. This increases the likelihood that those who were poor during childhood remained poor through to adulthood indicating lack of social mobility.

Nevertheless, some reasonable conclusions can be reached based on previous studies that the effect of SES on CGMV may indeed be cumulative. As a brain region, cerebellum is highly sensitive to environmental influence and indeed heritable influences on volume are less than for other brain regions [1]. The cerebellum also has a very protracted period of development, rendering it more vulnerable to environmental “toxicity” [18]. Neurogenesis continues postnatally through first 2 years—with some data suggesting further development into late childhood and early adulthood [37]. Children exposed to early deprivation show smaller L and R superior–posterior cerebellar lobe volumes than controls [18]. There is good evidence that early stress may have a negative effect on the cerebellar volumes. However, there is also good evidence that the cerebellum is particularly vulnerable to the effects of toxins throughout the life span [37]. Is there a biological basis for the association between SES and CGMV? In our study, we attempted to capture physiological risk factors that may have a role in the relationship between SES and CGMV. We initially calculated an MPRI and measure that was representative of a cumulative physiological risk across different physiological systems. Indeed, poorer SES was associated with greater MPRI. Gruenewald et al. [16] suggest that SES adversity across the life course may accumulate to negatively affect the functioning of biological regulatory systems important to functioning and health in later adulthood. In a recent study, they examined whether greater life course SES adversity experience was associated with higher scores on a multi-system AL index of physiological function in adulthood. They found that an increased AL score calculated from 24 biomarkers from 7 different physiological systems was associated with greater SES adversity at both childhood and adulthood, and indeed cumulatively across the life course. They conclude that SES adversity experience may accumulate across the life course to have a negative effect on multiple biological systems in adulthood. The MPRI used in our study can be interpreted as a biological marker of deprivation and an index of cumulative allostatic load similar to that of Gruenewald et al. [16]. The MPRI was found to mediate the relationship between SES and cerebellar volume. In other words, there was significant shared variance between SES and MPRI in explaining cerebellar grey matter volume.

However, the relative contribution of early versus current/adult SES to the physiological risk is less easy to unpick. Gruenewald et al. [16] suggest that biological functioning in middle and later adulthood may be affected by recent SES adversity but may also have residual effects of SES adversity earlier in the life course and a number of studies have found attenuation of the association between childhood SES and adult health markers on controlling for adult SES. We however found that both the relationship between ESES and CSES and CGMV were indeed mediated by the MPRI. The MPRI includes both markers of inflammation and metabolic disturbance. Both have an effect on brain structure and function. For example, inflammatory mediators have been linked to neurodegeneration and structural remodelling as well as modulation of neurogenesis in the hippocampus [38, 39]. Microglia in the vicinity can respond to these mediators with augmented TNF release and neuronal damage can ensue via direct toxicity of TNF and due to impaired astrocyte uptake of glutamate leading to excitotoxic neurotoxicity [40].

In relation to the other physiological mediators, other studies have shown that glucose metabolism is different in the cerebellum than in cerebral structures [41]. Hoogendam et al. [42] also report that a lower total cholesterol level is related to a smaller cerebellar volume.

As alcohol is widely seen as a key confound in imaging studies that include the cerebellum, we included this as a covariate in the model. Alcohol use did not predict grey matter volume, and did not have any effect on the association between SES and grey matter volume.

There is increasing awareness of the role the cerebellum plays in cognitive function. This includes possible involvement in reward-based associative learning and predictive control in emotional and cognitive learning. To date, the specific influence of social deprivation on cerebellum and its cognitive role has been difficult to unpick. However, our data resonate with those of Bauer et al. who used structural MRI to compare cerebellar volumes in healthy and previously neglected children [18]. This study also noted larger volumes yielded better performance in tests of memory and planning. Our data are consistent with these in terms of the cognitive domains tapped and, as such, raises the issue of the connectivity of the cerebellum to other brain regions implicated in the cognitive functioning. For example, it has been demonstrated that stimulation of the dentate nucleus of the cerebellum evokes dopamine release in medial prefrontal cortex and Purkinje cells in the superior–posterior lobe project to prefrontal areas via ventral dentate and thalamus [43, 44]. Rogers et al. investigated neuronal circuitry whereby cerebellum modulates this DA release in the mPFC and found two distinct neuronal circuits contribute to mPFC DA release, providing more evidence for the functional connectivity of cerebellum and PFC [45]. In a recent meta-analysis of studies reporting cerebellar activation in selected task categories by Stoodley and Schahmann, peaks were identified in lobule VI and Crus I for language and working memory; lobule VI for spatial tasks; lobules VI, Crus I and VIIB for executive functions and lobules VI, Crus I and medial VII for emotional processing [46]. Further, data from this meta-analysis also give support to an anterior sensorimotor and posterior cognitive/emotional dichotomy in human cerebellum. While MacLullich et al. observed a relationship between cerebellum area (not volume) and cognitive performance, Hogan et al. using voxel-based morphometry and structural equation modelling, found that grey matter volume in the cerebellum predicted general intelligence factor controlling for total intracranial volume as well as grey and white matter volume in frontal lobes [47, 48]. In this study, they also found sex differences in that a significant relationship was found between cerebellum and GM volume and G in men but not women. Research indicates that cerebellum shrinks proportionally more with age in men than women [49].

While the positive features of this study include a community-based sample and a very well-characterised cohort, there are also some limitations to be acknowledged: the cross-sectional design limits our ability to attribute causation. From these data, it is not possible to determine whether inflammatory risk factors are causing structural brain change or vice versa.

We argue that deprivation across the life course is negatively associated with cerebellar volume but there may be other explanations for differences in structure. We have an apparent limitation in this being only a male sample. However, we deliberately recruited only men to avoid the potential confounds around the effects of sex hormones mediating adaptive structural plasticity in the brain, which has been extensively investigated in, e.g. the hippocampus. Stress and sex hormones exert their effects on brain structural remodelling through both classical genomic and non-genomic mechanisms, and do so in collaboration with neurotransmitters and other mediators [50]. Thus, we took the view that a mixed sample would be more difficult to interpret. Linking SES to morphological changes in the brain is challenging but a sufficient number of studies have now been done to allow reviews of the data to be carried out. The parent study—PSoBiD—was designed to explore the differences in a number of biomarkers between the least and most deprived groups with the greatest power. Therefore, the nature of the sampling technique (based on neighbourhood deprivation) used for recruiting subjects led to a bimodal distribution of a number of independent variables (including ESES and CSES) used in this study. However the dependent variable (grey matter volume) was normally distributed. Including neighbourhood level deprivation as a covariate in the model would have been inappropriate [51], as there was significant correlation between ESES/CSES and neighbourhood deprivation. This however meant that we could not tease out neighbourhood and individual level deprivation on the cerebellum. Another limitation of this study is that the white matter in the folia of the cerebellum is tightly packed and therefore usually of sub-voxel resolution. Therefore, our cerebellar grey volume measurements reflect the “bulk” anatomy and are not sensitive to subtle changes in the folia structure. To our knowledge, no publicly available brain structure segmentation method exists that can model such partial volume voxels in the cerebellum. In summary, our data demonstrate a strong relationship between cerebellar structure, physiological risk index/allostatic load and SES (early and current).


We thank Dr Mortimer and Theresa Sackler Foundation for their support. This work was funded by the Glasgow Centre for Population Health, a partnership between NHS Greater Glasgow and Clyde, Glasgow City Council and the University of Glasgow, supported by the Scottish Government. The Glasgow Centre for Population Health had a role in study design, data collection and analysis, decision to publish and the preparation of the manuscript.

Copyright information

© Springer Science+Business Media New York 2013