Head size, age and gender adjustment in MRI studies: a necessary nuisance?
Research highlights
► Brain volume correlates with head size (TIV) and differs between the sexes. ► Increasing age has a negative association with brain volumes and cortical thickness. ► This can confound observational studies or reduce statistical power. ► For cortical thickness, we recommend adjustment for age, and gender. ► For some volumes and for voxel-based morphometry, we recommend age, gender and TIV.
Introduction
Cerebral atrophy is a characteristic and often defining feature of a number of degenerative diseases including Alzheimer's disease (AD) (Chan et al., 2003, Jack et al., 2004), frontotemporal lobar degeneration (FTLD) (Chan et al., 2001a, Whitwell et al., 2008) and Huntington's disease (HD) (Henley et al., 2009, Tabrizi et al., 2009). Magnetic resonance imaging (MRI) is increasingly used to investigate these diseases (Barnes et al., 2007, Bohanna et al., 2008, Rovira & Leon, 2008) and imaging forms part of recommended investigations in dementia (Knopman et al., 2001, Waldemar et al., 2007) not only to exclude causes of symptoms, but also to determine characteristic patterns of atrophy associated with particular diagnoses (Dubois et al., 2007, McKeith et al., 1996, Neary et al., 1998).
Progressive improvements in MRI contrast and resolution mean that increasingly detailed measurements of regional or cortical volumes, thicknesses, surface areas and curvatures of structures or classes of tissue are now possible. There is increasing interest in using such measurements in randomized clinical trials to determine efficacy of putative disease-modifying treatments, as well as in observational case–control studies. Comparability of treatment groups in the former is assured (at least in large trials) by randomization. In contrast, in observational case–control studies, upon which we focus in this paper, subject groups (defined by disease status) often differ according to confounders that affect the outcome of interest, rendering simple group comparisons biased. One approach for dealing with such confounding is to match cases and controls for the confounding variables. A second approach is to collect information on confounding factors, and adjust for these, usually by including them as covariates in a statistical regression model. In addition to considerations of bias, adjusting for a predictive covariate increases the power of between-subject comparisons.
In observational MRI studies, age, gender and head size are the most commonly included so-called ‘nuisance’ variables. However, studies vary as to which of these variables are allowed for. Such differences in covariate selection may be reasonable, however, since whether a variable confounds comparisons between groups depends both on whether the variable affects the particular outcome under study and also whether the variable is balanced between subject groups. A balanced variable is not necessarily one which shows no significant differences across groups since non-statistically significant differences in smaller studies may still lead to confounded results.
A number of studies have shown that age is associated with lower whole-brain (Courchesne et al., 2000, Gur et al., 1991, Scahill et al., 2003), temporal lobe and hippocampal volumes (Scahill et al., 2003), grey matter volume (Courchesne et al., 2000, Ge et al., 2002, Good et al., 2001, Guttmann et al., 1998, Pell et al., 2008, Raz et al., 1997, Smith et al., 2007, Taki et al., 2004), and cortical thickness (Sowell et al., 2007). Effects of age on white matter are more varied: some studies have shown no significant effect (Good et al., 2001, Smith et al., 2007, Taki et al., 2004), whilst others show an increase until middle age followed by a decline (Courchesne et al., 2000, Ge et al., 2002) whilst others again show decline with age (Guttmann et al., 1998, Lemaitre et al., 2005). Two studies which show no overall significant effect of age on white matter volume do show some areas of white matter volume decline with age using voxel-based techniques (Good et al., 2001, Taki et al., 2004).
Male gender has been shown to be associated with larger cerebral volumes (Gur et al., 1991, Sowell et al., 2007) which disappears with head size correction (Scahill et al., 2003). Greater decline of grey matter volume with age in males has also been reported in some (Ge et al., 2002, Raz et al., 1997, Taki et al., 2004) but not other (Lemaitre et al., 2005) studies. Females have also been shown to have thicker cortex across many regions of the brain (Luders et al., 2006, Sowell et al., 2007).
Although measurement of age and gender when used are obvious, head size measures vary widely (Pengas et al., 2009) and some studies have used body height as a proxy (Raz et al., 1997, Sowell et al., 2007). Most commonly, an estimate of intracranial volume is used but there is a wide range of methods including tissue compartment addition (grey matter plus white matter plus cerebrospinal fluid (CSF)) (Courchesne et al., 2000, Lemaitre et al., 2005, Rudick et al., 1999, Smith et al., 2007); registration-based estimation (Smith et al., 2002) estimates generated from points making an ellipsoid (Pfefferbaum et al., 2000), area of mid-sagittal slice section (Raz et al., 1997), and semi-automatic segmentation of every tenth slice (Scahill et al., 2003). In addition there are many ways in which head size correction is applied including: simple division of volumes (Chan et al., 2001b), regression association correction (Scahill et al., 2003) and statistical adjustment with the nuisance variable as a covariate (Good et al., 2002).
Using the same images from a set of normal control subjects, we aimed to assess which nuisance variables are associated with specific metrics often used in imaging studies: region of interest (ROI) volumes, cortical thickness and voxel-based morphometry (VBM). Our hypothesis was that these variables would have large associations with cerebral structures and tissue compartments but we did not want to assume that the nature of these associations would be identical across structures. We aimed to quantify the independent contribution of each of the nuisance variables in explaining differences in volumes and cortical thicknesses between subjects, from which we could make recommendations as to which require consideration in observational studies.
Section snippets
Subjects
All individuals had been recruited as normal control subjects for research projects at the National Hospital for Neurology and Neurosurgery, London, UK. These projects included longitudinal studies on sporadic AD (Schott et al., 2005), familial AD (Ridha et al., 2006), FTLD (Rohrer et al., 2008), HD (Henley et al., 2006, Henley et al., 2009), and progressive supranuclear palsy (Paviour et al., 2006) together with a cross-sectional magnetization transfer ratio imaging study in AD (Ridha et al.,
Results
In total 78 subjects were included in this study and the demographic data for these subjects are reported in Table 1.
Discussion
We have investigated the associations of age, gender and TIV with volumetric measures and cortical thickness using MRI. Unsurprisingly, each covariate alone had an effect on nearly every metric we assessed: age had a negative association with volumes and thicknesses, men had larger volumes and smaller cortical thicknesses than women and TIV had positive association with volumes and a non-significant negative association with thicknesses. However, within the same model, some variables showed
Acknowledgments
This work was undertaken at UCLH/UCL who received a proportion of funding from the Department of Health's NIHR Biomedical Research Centres funding scheme. The Dementia Research Centre is an Alzheimer's Research Trust Co-ordinating Centre. Dr. Barnes is supported by the Alzheimer's Research Trust (UK). Professor Fox is supported by the Medical Research Council and NIHR (UK). Ms. Lehmann is supported by the Alzheimer's Society (UK). Dr. Clarkson is supported by the Technology Strategy Board
References (52)
A fast diffeomorphic image registration algorithm
Neuroimage
(2007)- et al.
Unified segmentation
Neuroimage
(2005) - et al.
Magnetic resonance imaging as an approach towards identifying neuropathological biomarkers for Huntington's disease
Brain Res. Rev.
(2008) - et al.
Change in rates of cerebral atrophy over time in early-onset Alzheimer's disease: longitudinal MRI study
Lancet
(2003) - et al.
Cortical surface-based analysis. I. Segmentation and surface reconstruction
Neuroimage
(1999) - et al.
Differential effects of aging and Alzheimer's disease on medial temporal lobe cortical thickness and surface area
Neurobiol. Aging
(2009) - et al.
Research criteria for the diagnosis of Alzheimer's disease: revising the NINCDS-ADRDA criteria
Lancet Neurol.
(2007) - et al.
Cortical surface-based analysis. II. Inflation, flattening, and a surface-based coordinate system
Neuroimage
(1999) - et al.
Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain
Neuron
(2002) - et al.
Interactive algorithms for the segmentation and quantitation of 3-D MRI brain scans
Comput. Meth. Programs Biomed.
(1997)
Thresholding of statistical maps in functional neuroimaging using the false discovery rate
Neuroimage
A voxel-based morphometric study of ageing in 465 normal adult human brains
Neuroimage
Automatic differentiation of anatomical patterns in the human brain: validation with studies of degenerative dementias
Neuroimage
Age- and sex-related effects on the neuroanatomy of healthy elderly
Neuroimage
Selection of the control group for VBM analysis: influence of covariates, matching and sample size
Neuroimage
Brain structure in men remains highly heritable in the seventh and eighth decades of life
Neurobiol. Aging
Issues with threshold masking in voxel-based morphometry of atrophied brains
Neuroimage
Tracking atrophy progression in familial Alzheimer's disease: a serial MRI study
Lancet Neurol.
MR in the diagnosis and monitoring of multiple sclerosis: an overview
Eur. J. Radiol.
Accurate, robust, and automated longitudinal and cross-sectional brain change analysis
Neuroimage
Age and gender effects on human brain anatomy: a voxel-based morphometric study in healthy elderly
Neurobiol. Aging
Biological and clinical manifestations of Huntington's disease in the longitudinal TRACK-HD study: cross-sectional analysis of baseline data
Lancet Neurol.
Voxel-based morphometry of human brain with age and cerebrovascular risk factors
Neurobiol. Aging
Rates of brain atrophy over time in autopsy-proven frontotemporal dementia and Alzheimer disease
Neuroimage
Structural magnetic resonance imaging-derived biomarkers for Alzheimer's disease
Biomark. Med.
Rates of global and regional cerebral atrophy in AD and frontotemporal dementia
Neurology
Cited by (390)
Genetic and brain similarity independently predict childhood anthropometrics and neighborhood socioeconomic conditions
2024, Developmental Cognitive NeuroscienceAnhedonia and Delay Discounting: Differing Patterns of Brain-Behavior Relationships in Healthy Control Participants Versus Individuals With Posttraumatic Stress Disorder
2024, Biological Psychiatry: Cognitive Neuroscience and Neuroimaging