OUP user menu

Motor cortex and gait in mild cognitive impairment: a magnetic resonance spectroscopy and volumetric imaging study

Cédric Annweiler, Olivier Beauchet, Robert Bartha, Jennie L. Wells, Michael J. Borrie, Vladimir Hachinski, Manuel Montero-Odasso
DOI: http://dx.doi.org/10.1093/brain/aws373 859-871 First published online: 25 February 2013

Summary

Gait disorders are common in the course of dementia, even at the stage of mild cognitive impairment, owing to probable changes in higher levels of motor control. Since motor control message is ultimately supported in the brain by the primary motor cortex and since cortical lesions are frequent in the dementia process, we hypothesized that impairments of the primary motor cortex may explain the early gait disorders observed in mild cognitive impairment. Our purpose was to determine whether the neurochemistry of the primary motor cortex measured with proton magnetic resonance spectroscopy, and its volume, were associated with gait performance while single and dual-tasking in mild cognitive impairment. Twenty community dwellers with mild cognitive impairment, aged 76 years (11) [median (interquartile range)] (30% female) from the ‘Gait and Brain Study’ were included in this analysis. Gait velocity and stride time variability were measured while single (i.e. walking alone) and dual tasking (i.e. walking while counting backwards by seven) using an electronic walkway (GAITRite System). Ratios of N-acetyl aspartate to creatine and choline to creatine and cortical volume were calculated in the primary motor cortex. Participants were categorized according to median N-acetyl aspartate to creatine and choline to creatine ratios. Age, gender, body mass index, cognition, education level and subcortical vascular burden were used as potential confounders. Participants with low N-acetyl aspartate to creatine (n = 10) had higher (worse) stride time variability while dual tasking than those with high N-acetyl aspartate to creatine (P = 0.007). Those with high choline to creatine had slower (worse) gait velocity while single (P = 0.015) and dual tasking (P = 0.002). Low N-acetyl aspartate to creatine was associated with increased stride time variability while dual tasking (adjusted β = 5.51, P = 0.031). High choline to creatine was associated with slower gait velocity while single (adjusted β = −26.56, P = 0.009) and dual tasking (adjusted β = −41.92, P = 0.022). Cortical volume correlated with faster gait velocity while single (P = 0.029) and dual tasking (P = 0.037), and with decreased stride time variability while single tasking (P = 0.034). Finally, the probability of exhibiting abnormal metabolite ratios in the primary motor cortex was 63% higher among participants with major gait disturbances in dual task. Those with compromised gait velocity in dual task had a 2.05-fold greater risk of having a smaller cortical volume. In conclusion, the neurochemistry and volume of the primary motor cortex were associated with gait performance while single and dual tasking. Stride time variability was mainly sensitive to neuronal function (N-acetyl aspartate to creatine), whereas gait velocity was more affected by inflammatory damage (choline to creatine) and volumetric changes. These findings may contribute to a better understanding of the higher risks of mobility decline and falls in subjects with mild cognitive impairment.

  • gait
  • primary motor cortex
  • mild cognitive impairment
  • proton magnetic resonance spectroscopy
  • magnetic resonance imaging
  • volumetry

Introduction

Gait disorders are a public health concern among older adults due to their adverse consequences including falls, loss of independence, institutionalization and death (Alexander, 1996; Montero-Odasso et al., 2005; Beauchet and Berrut, 2006). They are particularly common throughout the course of dementia and even at the stage of pre-dementia (i.e. mild cognitive impairment) (Beauchet et al., 2008; Verghese et al., 2008; Montero-Odasso et al., 2012). Specifically, it has been reported that gait disorders in mild cognitive impairment involve both the propulsion component of gait, as illustrated by decreased gait velocity (Verghese et al., 2008; Maquet et al., 2010), but also its dimension of stability, as illustrated by increased stride time variability (Verghese et al., 2008; Beauchet et al., 2013). These findings are particularly obvious while using the dual-task paradigm (‘walking while talking’), indicating that performing an attention task while walking results in the deterioration of gait among participants with mild cognitive impairment, the so-called ‘dual-task cost’ (Verghese et al., 2008; Maquet et al., 2010; Montero-Odasso et al., 2012). This observation suggests the possibility that cognitive decline can lead to gait disorders among older adults with mild cognitive impairment independent of declines in muscle strength, tone or osteoarticular functions that may accompany ageing (Beauchet and Berrut, 2006; Beauchet et al., 2008). Specifically, high levels of motor control involve cognitive function to produce complex motor responses that are adapted to multiple sensory inputs and environmental constraints (Beauchet et al., 2008). The integration of afferent information in the brain generates a global motor control message that is ultimately supported by the primary motor cortex (Graziano et al., 2002; Meier et al., 2008). Located in the posterior portion of the frontal lobe, the primary motor cortex receives considerable converging inputs from many cortical and subcortical regions involved in the high levels of gait control such as the frontal lobe and the thalamus, and it provides the largest contribution to the descending corticospinal tract, with some of these neurons synapsing directly onto α-motor neurons to appropriately execute movements (Graziano et al., 2002). Supporting the importance of the primary motor cortex in gait control, recent functional imaging studies of healthy older adults have reported that the primary motor cortex has a critical role in the executive network of locomotion (la Fougère et al., 2010; Jahn and Zwergal, 2010).

As cortical lesions are frequent in the course of dementias (Modrego and Fayed, 2011; Niskanen et al., 2011), we hypothesized that impairments of the primary motor cortex, as the final integrator of all brain processes involved in gait control, may explain, at least in part, the early gait disorders observed in mild cognitive impairment. The purpose of this study was to determine whether the neurochemistry in primary motor cortex measured by proton magnetic resonance spectroscopy (1H-MRS) was associated with gait performance in older participants diagnosed with mild cognitive impairment while single and dual tasking. Our secondary objective was to examine the relationship between the volume of the primary motor cortex and gait performance in mild cognitive impairment.

Materials and methods

Participants

We studied 20 participants with mild cognitive impairment aged 76 years (11) [median (interquartile range)] (30% female) recruited in the second wave of the ‘Gait and Brain Study’ from September 2011 to March 2012. The ‘Gait and Brain Study’ is a longitudinal prospective cohort study designed to determine whether early gait disorders can predict cognitive and mobility decline, progression to dementia and frailty status among older adults with mild cognitive impairment (Montero-Odasso et al., 2009). The subset of participants included in the present analysis corresponded to the 20 participants who received 1H-MRS of the primary motor cortex allowing metabolite measurements. Power analysis determined that 20 participants were sufficient to find significant correlations between primary motor cortex metabolites and gait performance in mild cognitive impairment (Machin et al., 1997). The sampling and data collection procedures have been described elsewhere in detail (Montero-Odasso et al., 2009). Mild cognitive impairment identification was based on the consensus criteria by Dubois and Albert (2004), which included the presence of subjective memory complaints from the patient and family, objective memory impairment (assessed using a complete battery of neuropsychological tests designed to examine a full range of cognitive functions), preserved general intellectual function (assessed clinically), absence of significant functional impairment and absence of clinical dementia. Exclusion criteria were diagnosis of a terminal illness, life expectancy <12 months, pending nursing home placement, hip or knee joint arthroplasty within the preceding 6 months, inability to walk independently, use of walking aids and diagnosis of dementia. Study participants received a complete medical examination, consisting of structured questionnaires and a standardized clinical and neurocognitive examination. Additionally, gait and balance assessments and 3.0 T MRI of the brain were performed.

Gait assessment

Gait velocity (cm/s) and stride time variability [coefficient of variation of stride time = (mean/standard deviation) × 100] were measured using a computerized walkway with embedded pressure sensors (GAITRite®, 600 × 64 × 1 cm, active electronic surface area 792 × 610 cm, with a total of 29 952 pressure sensors, scanning frequency 60 Hz, software version 3.8, CIR Systems). The GAITRite® system is a reliable tool for gait analysis that has been validated for several gait protocols by our centre and others, including for assessments of gait velocity and variability (Montero-Odasso et al., 2009; Beauchet et al., 2011). Two walking conditions were successively measured in non-randomized order: walking at self-selected pace (i.e. single-task condition) and then walking while counting aloud backwards by 7 starting from 100 (i.e. dual-task condition). Before each condition, a trained evaluator gave standardized verbal instructions to the participants. Participants walked in a quiet, well-lit room wearing their own footwear according to validated protocols published elsewhere (Kressig et al., 2006; Montero-Odasso et al., 2009). To avoid acceleration and deceleration effects, participants started walking 1 m before reaching the electronic walkway and completed their walk 1 m beyond it. The change in gait velocity (Δgait velocity) and stride time variability (Δstride time variability) related to the dual-task condition compared with the single-task condition (i.e. the dual-task cost) was calculated with the following formula: Δgait velocity,stride time variability = {[gait velocity or stride time variability while dual tasking − gait velocity or stride time variability while single tasking]/[(gait velocity or stride time variability while dual tasking + gait velocity or stride time variability while single tasking)/2]} × 100 (Beauchet et al., 2005). The lowest quintile of Δgait velocity was defined to represent a major decrease in gait velocity while dual tasking, while the highest quintile of Δstride time variability was defined to represent a major increase in stride time variability while dual tasking. These dual-task costs reflected the ability to maintain gait performance while attention demands were increased. They are classically interpreted as the interference arising from sharing attention resources when simultaneously completing two attention-requiring tasks compared with one at a time (Arnell and Duncan, 2002).

Magnetic resonance imaging data acquisition

All magnetic resonance data were acquired on a 3.0 T Siemens Tim Trio MRI, using a 32-channel head coil. Each examination included the acquisition of sagittal 3D T1-weighted inversion-prepared rapid acquisition with gradient echo (MP-RAGE) anatomical images (repetition time/echo time = 2300/2.9 ms, inversion time = 900 ms, flip angle = 9°, averages = 1, field of view = 256 × 240 × 192 mm, matrix = 256 ×240 × 160) covering the entire brain to produce high grey matter/white matter contrast. Fluid-attenuated inversion recovery (FLAIR) coronal images were also acquired (acquisition matrix = 256 × 232, reconstructed to 512 × 464 matrix, field of view = 220 × 200 mm, thickness = 4 mm, gap = 0.5 mm, 41 slices, repetition time = 8 s, echo time = 120 ms, inversion time = 2400 ms, flip angle = 130°, averages = 1) for the assessment of subcortical white matter hyperintensities (SWMH).

Magnetic resonance spectroscopy

The anatomical images were used to guide the placement of a 20-mm isotropic voxel on the leg and foot region of the right motor cortex (Fig. 1) (Youstry et al., 1997). Spectra were localized by point-resolved spectroscopy (repetition time/echo time = 2000/135 ms, voxel size = 8 cm3). Both water-suppressed (averages = 192) and water-unsuppressed spectra (averages = 8) were acquired. Data were line-shape corrected by combined QUALITY deconvolution and eddy current correction (Bartha et al., 2000) using the unsuppressed water as a reference. Any unsuppressed water signal remaining in the water-suppressed spectrum at 4.7 parts per million was removed by subtracting peaks between 4.1 and 5.1 parts per million identified by a Hankel singular value decomposition algorithm (Kassem and Bartha, 2003). Resultant metabolite spectra were fitted in the time domain using a Levenberg–Marquardt minimization routine incorporating a template of prior knowledge of metabolite line shapes. The analysis software (fitMAN) is incorporated into a graphical user interface written in our laboratory in the IDL (version 5.4 Research Systems Inc.) programming language (Bartha et al., 1999). The acquisition of metabolite prior knowledge data has been previously described in detail (Bartha et al., 1999; Kassem and Bartha, 2003; Kowalczyk et al., 2012). Briefly, high resolution in vitro spectra were acquired for all visible metabolites including solutions (pH adjusted to 7.04) of N-acetyl aspartate, creatine and choline using the same sequence that was used to acquire all in vivo data. Each metabolite solution contained sodium 3-trimethylsilyl propionic acid as a reference for chemical shift, zero-order phase and Lorentzian damping (line width). The high-resolution metabolite spectra were fitted to produce metabolite templates that were subsequently used to fit in vivo spectra.

Figure 1

The spectroscopy voxel placed in the right motor cortex is outlined in yellow on sagittal (left), axial (middle) and coronal (right) images.

In the current study, magnetic resonance spectra were successfully acquired in all participants. Figure 2 shows the spectrum acquired in one participant along with the fitted result and the residual (the difference between the fit and the spectrum). The individual metabolite components for N-acetyl aspartate, choline and creatine are also provided. N-acetyl aspartate was examined because it has been linked to the functional status of neuronal mitochondria and is therefore considered as a marker of neuronal health and function that decreases in diseases that adversely affect neuronal integrity (Moffett et al., 2007); creatine provides a measure of oxidative energy stores (Wozniak and Lim, 2006), and choline is a measure of cellular turnover that is elevated in inflammatory processes and ischaemic lesions of the brain due to gliosis or ischaemic damage to myelin (Ferraz-Filho et al., 2009; Karaszewski et al., 2010). More specifically, metabolite ratios (N-acetyl aspartate/creatine and choline/creatine) were calculated. Metabolite ratios relative to creatine provide reproducible and sensitive measurements that remove quantification errors associated with tissue partial volume effects that may arise from having different fractions of white/grey matter and CSF in selected voxels (Jensen et al., 2006). Ratios can also be more sensitive in detecting metabolite changes when one metabolite in the ratio (e.g. the numerator) increases, while the other metabolite in the ratio (e.g. the denominator) decreases. In our sample, participants were separated into two groups based on their N-acetyl aspartate/creatine and choline/creatine ratios, using a threshold N-acetyl aspartate/creatine ratio of the median value (i.e. N-acetyl aspartate/creatine = 1.17), and similarly a threshold choline/creatine ratio of the median value (i.e. choline/creatine = 0.58).

Figure 2

Representative 1H-MRS data from the motor cortex in one subject. Data (grey line) are shown with the fit spectrum (black line) superimposed. The residual is shown above with the individual metabolite components included in the analysis: N-acetylaspartate (NAA), creatine (Cr), choline containing compounds (Cho). ppm = parts per million.

Volumetric magnetic resonance imaging analysis

The volumetric 3D T1-weighted images were segmented using the FreeSurfer software package (version 5.1.0; Fischl et al., 2002) to calculate the grey matter volume of the primary motor cortex (i.e. Brodmann area 4) and of a number of control regions of interest thought to be involved in motor control such as the premotor cortex and supplementary motor area (i.e. Brodmann area 6), the primary somatosensory cortex (i.e. Brodmann areas 1–3), the hippocampus, the cortex of the frontal lobe and of the superior parietal lobule, the posterior cingulate, the thalamus, the basal ganglia and the cerebellar cortex volume. FreeSurfer is a set of tools that automatically segments and labels these structures based on established processing steps; the technical specifications of these procedures have been described previously (Dale and Sereno, 1993; Dale et al., 1999; Fischl et al., 1999a, b, 2001, 2002, 2004a, b; Fischl and Dale, 2000; Segonne et al., 2004; Han et al., 2006; Jovicich et al., 2006). Briefly, image processing included the following steps as described on the Freesurfer website (http://surfer.nmr.mgh.harvard.edu): removal of non-brain tissue using a hybrid watershed/surface deformation procedure (Segonne et al., 2004); automated Talairach transformation; segmentation of the subcortical white matter and deep grey matter structures (Fischl et al., 2002, 2004a); tessellation of the grey matter/white matter boundary; automated topology correction (Fischl et al., 2001; Segonne et al., 2004); registration to a spherical atlas (Fischl et al., 1999b); parcellation of the cerebral cortex into units, including Brodmann area 4, based on gyral and sulcal structure (Fischl et al., 2004b; Desikan et al., 2006); surface inflation and creation of surface-based data (Dale et al., 1999) (Fig. 3). The procedures for the measurement of cortical volume have been validated against histological analysis (Rosas et al., 2002) and manual measurements (Kuperberg et al., 2003; Salat et al., 2004). Freesurfer morphometric procedures have demonstrated good test–retest reliability across scanner manufacturers and across field strengths (Han et al., 2006).

Figure 3

Automated parcellation of the primary motor cortex in one representative subject (Brodmann area 4, in white) with FreeSurfer on the 3D rendering of the brain from the T1-weighted magnetic resonance images. Top view, normal (A) and inflated (D) surface; lateral view, normal (B) and inflated (E) surface; midsagittal view, normal (C) and inflated (F) surface.

Covariables

Age, gender, body mass index, cognitive performance, education level and subcortical vascular burden were included as potential confounders in the data analysis. Body mass index (kg/m2) was calculated as weight (kg)/height2 (m2). Weight was measured with a beam balance scale and height with a height gauge. Cognitive performance as a covariable was assessed using the Montreal Cognitive Assessment, a 30-point test assessing multiple aspects of cognition: short-term memory, visuospatial abilities, executive functioning, attention, concentration and working memory, language and orientation to time and place (Nasreddine et al., 2005). The Montreal Cognitive Assessment was specifically designed for the detection of mild cognitive impairment and has proven to be sensitive to deficits in cognition within this group, with very high internal consistency and good retest reliability (Nasreddine et al., 2005). In addition, education level was reported with a structured questionnaire. Participants who passed at least the Elementary School Recognition Certificate were considered to have higher education level compared with those who did not. Finally, the subcortical vascular burden was measured using the semiquantitative visual rating scale for SWMH devised by Fazekas et al. (1987) applied to the T2-weighted FLAIR images. SWMH were graded on a four-point scale of increasing severity: 0, normal; 1, punctuate foci; 2, beginning confluence of foci; and 3, large confluent areas. The reliability of the Fazekas scale is high with intra-operator correlation coefficient of 0.85 (Leaper et al., 2001), and magnetic resonance images were scored under the supervision of a neuroradiologist (I.B.G.) by a single observer (C.A.) who was blinded to participants’ clinical information, including age, gender, prior imaging findings or cardiovascular disease risk factors.

Statistics

Participants’ characteristics were summarized using medians (interquartile range) or frequencies and percentages, as appropriate. First, comparisons between participants categorized according to median values of metabolite ratios in the primary motor cortex were performed using the Chi-square test or the Mann–Whitney U test, as appropriate. Comparison of gait measures between single task and dual task was performed using the Wilcoxon test. Second, a Pearson correlation matrix was used to determine which gait measures in single and dual task were specifically linked to metabolite ratios in the primary motor cortex. Bivariate and SWMH-adjusted partial Pearson correlations were also used to examine the relationships between gait measures in single and dual task and the volumes of the primary motor cortex and other regions of interest. Third, multiple linear regressions were used to examine the association of low N-acetyl aspartate/creatine and high choline/creatine (independent variables) with gait velocity and stride time variability while single and dual tasking (dependent variables) after adjustment for age, gender, Montreal Cognitive Assessment score and SWMH grade. Separated analyses were performed for each model. Fourth, the risk difference of abnormal metabolite ratios in the primary motor cortex and the effect size of decreased primary motor cortex volume were calculated for participants who exhibited major gait disturbances while dual tasking compared with those who did not. All statistics were performed using SPSS (version 19.0; SPSS). Risk difference and effect size were obtained with RevMan v5.1 (The Nordic Cochrane Centre, Copenhagen, Denmark).

Standard protocol approvals, registrations and patient consents

The study was conducted in accordance with the ethical standards set forth in the Helsinki Declaration (1983). Ethics approval was obtained from the University of Western Ontario Research Ethics Board for Health Sciences Research Involving Human Subjects. Written informed consent was obtained at enrolment according to protocols approved by the local institutional review board.

Results

As illustrated in Table 1, median gait velocity was 119.9 (20.0) cm/s while walking alone and 95.5 (49.3) cm/s while dual tasking (P = 0.006). The stride time variability was 1.8% (0.8) while walking alone and 3.7% (3.1) while dual tasking (P = 0.001). Four participants (20%) experienced a major decrease in gait velocity and a major increase in stride time variability while dual tasking compared with single tasking. Participants with N-acetyl aspartate/creatine below the median (n = 10) exhibited higher (worse) stride time variability while dual tasking than those with N-acetyl aspartate/creatine above the median (P = 0.009). Participants with choline/creatine above the median exhibited slower (worse) gait velocity while single and dual tasking than those with choline/creatine below the median (P = 0.015 and P = 0.002, respectively). These participants also had a lower primary motor cortex volume compared with those with low choline/creatine (P = 0.047).

View this table:
Table 1

Characteristics and comparison of participants with mild cognitive impairment categorized according to median values of N-acetyl aspartate/creatine and choline/creatine ratios in the primary motor cortex

Total cohort (n = 20)N-acetyl aspartate/creatine ratioCholine/creatine ratio
<1.17 (n = 10)≥1.17 (n = 10)P-valuea≤0.58 (n = 10)>0.58 (n = 10)P-valuea
Clinical measures
    Age (years)76.0 (11.0)75.5 (12.0)76.0 (11.0)0.79670.5 (8.0)77.5 (7.0)0.247
    Female gender, n (%)6 (30.0)4 (40.0)2 (20.0)0.3293 (30.0)3 (30.0)0.999
    Body mass index (kg/m2)26.1 (6.9)27.7 (6.4)23.7 (5.5)0.14326.8 (8.7)25.9 (6.9)0.393
    Montreal Cognitive Assessment score (/30)25.0 (5.0)25.5 (5.0)25.0 (5.0)0.31525.5 (5.0)25.0 (5)0.853
Neuroimaging measures
    N-acetyl aspartate/creatine ratio in primary motor cortex1.17 (0.26)1.02 (0.24)1.25 (0.18)<0.0011.19 (0.20)1.16 (0.43)0.739
    Choline/creatine ratio in primary motor cortex0.58 (0.09)0.60 (0.13)0.55 (0.06)0.3930.54 (0.06)0.63 (0.11)<0.001
    Volume of primary motor cortex (mm3)9431.00 (1237.00)8862.00 (360.49)9512.44 (178.01)0.1379570.80 (164.28)8724.89 (375.70)0.047
    Subcortical WMH grade (/3)1.0 (1.0)1.0 (1.0)1.5 (1.0)0.7391.5 (1.0)1.0 (1.0)0.739
Gait measures
    Gait velocity (cm/s)
        Single-task119.9 (20.0)118.7 (47.4)121.1 (12.9)0.999127.1 (31.8)111.8 (22.9)0.015
        Dual-taskb95.5 (49.3)87.5 (70.5)108.2 (32.4)0.165122.3 (58.8)83.8 (24.6)0.002
    Stride time variability (%)
        Single-task1.8 (0.8)2.0 (1.1)1.6 (0.9)0.0751.6 (0.5)2.1 (0.6)0.105
        Dual-taskb3.7 (3.1)5.0 (11.5)3.4 (1.9)0.0073.4 (2.3)4.5 (6.0)0.063
  • Data are presented as median (interquartile range) where applicable.

  • a Comparisons of participants with low and high ratios (i.e. low N-acetyl aspartate/creatine versus high N-acetyl aspartate/creatine, and low choline/creatine versus high choline/creatine) based on Mann–Whitney U test or Chi-square test, as appropriate.

  • b Walking while counting backwards by seven.

  • WMH = white matter hyperintensities.

  • P < 0.05 indicated in bold.

Tables 2 and 3 report the correlations between the magnetic resonance findings in the primary motor cortex and gait measures while single and dual tasking. As illustrated in Table 2, low N-acetyl aspartate/creatine correlated positively with increased stride time variability while dual tasking (correlation coefficient r = 0.52 with P = 0.018). High choline/creatine correlated negatively with gait velocity while single and dual tasking (r = −0.58 with P = 0.008, and r = −0.65 with P = 0.002, respectively). Table 3 shows that the primary motor cortex volume correlated with faster gait velocity while single (r = 0.50 with P = 0.029) and dual tasking (r = 0.48 with P = 0.037), and with decreased stride time variability while single tasking (r = −0.49 with P = 0.034). Controlling for the SWMH grade in partial models did not alter the results (r = 0.52 with P = 0.027 for gait velocity in single task; r = 0.49 with P = 0.039 for gait velocity in dual task; r = −0.49 with P = 0.040 for stride time variability in single task; r = 0.05 with P = 0.839 for stride time variability in dual task). Conversely, gait velocity and stride time variability did not correlate with the volumes of the other regions of interest either in single or dual task (Table 3).

View this table:
Table 2

Correlation matrix of magnetic resonance findings and gait performance among participants with mild cognitive impairment while single and dual tasking (n = 20)

Characteristic123456
1. Low N-acetyl aspartate/creatine ratioa0.20−0.10−0.160.410.52*
2. High choline/creatine ratiob−0.58**−0.65**0.210.26
3. Gait velocity while single tasking0.84**−0.12−0.17
4. Gait velocity while dual tasking0.51−0.43
5. Stride time variability while single tasking−0.18
6. Stride time variability while dual tasking
  • Correlation matrix of abnormal metabolite ratios in the primary motor cortex and gait performance.

  • a N-acetyl aspartate/creatine ratio < 1.17.

  • b Choline/creatine ratio > 0.58.

  • *P < 0.05 (two-tailed); **P < 0.01 (two-tailed).

View this table:
Table 3

Correlation matrix of brain volumes and gait performance among participants with mild cognitive impairment while single and dual tasking (n = 20)

Characteristic1234567891011121314
1. Volume of primary motor cortex0.450.350.48*0.03−0.26−0.010.030.230.300.50*0.48*−0.49*0.08
2. Volume of premotor cortex plus supplementary motor area0.56*0.48*−0.060.300.130.200.76**0.340.140.03−0.230.31
3. Volume of primary somatosensory cortex0.260.170.340.270.280.420.140.310.23−0.420.22
4. Volume of hippocampus−0.200.06−0.040.440.390.400.360.34−0.15−0.02
5. Volume of frontal lobe cortex0.09−0.080.07−0.210.030.310.27−0.22−0.24
6. Volume of superior parietal lobule−0.190.050.40−0.32−0.19−0.19−0.060.22
7. Volume of posterior cingulate0.350.320.36−0.070.070.07−0.11
8. Volume of thalamus0.430.77**0.220.15−0.32−0.02
9. Volume of basal ganglia nuclei0.410.200.09−0.170.30
10. Volume of cerebellar cortex0.300.28−0.26−0.10
11. Gait velocity while single tasking0.84**−0.12−0.17
12. Gait velocity while dual tasking0.51−0.43
13. Stride time variability while single tasking−0.18
14. Stride time variability while dual tasking
  • *P < 0.05 (two-tailed); **P < 0.01 (two-tailed).

Multiple linear regressions showed that low N-acetyl aspartate/creatine in the primary motor cortex was directly associated with increased stride time variability while dual tasking (adjusted β = 5.51 with P = 0.031) (Table 4). The covariables were not associated with stride time variability, except the SWMH grade that was positively associated with increased stride time variability in dual task (adjusted β = 3.00 with P = 0.045). Additionally, N-acetyl aspartate/creatine as a continuous variable was associated with stride time variability while dual tasking {adjusted β = −10.37 [95% confidence interval (CI) −19.91 to −0.84], P = 0.035}. High choline/creatine in the primary motor cortex was inversely associated with gait velocity while single and dual tasking (β = −26.56 with P = 0.009, and β = −41.92 with P = 0.022, respectively) (Table 5).

View this table:
Table 4

Association between low N-acetyl aspartate/creatine ratioa and stride-to-stride variability of stride time while single and dual tasking

Stride-to-stride variability of stride time
Single taskDual taskd
β (95% CI)P-valueβ (95% CI)P-value
Low N-acetyl aspartate/creatine ratioa0.68 (−0.11 to 1.47)0.0875.51 (0.60 to 10.42)0.031
Age0.02 (−0.03 to 0.07)0.347−0.22 (−0.54 to 0.09)0.148
Female gender0.08 (−0.71 to 0.87)0.830−1.77 (−6.67 to 3.13)0.447
Body mass index0.05 (−0.04 to 0.14)0.278−0.16 (−0.71 to 0.40)0.553
Montreal Cognitive Assessment score−0.12 (−0.26 to 0.02)0.0780.13 (−0.73 to 1.00)0.741
High education levelb−0.41 (−1.51 to 0.70)0.438−0.64 (−7.46 to 6.19)0.842
SWMH grade−0.02 (−0.49 to 0.45)0.9423.00 (0.08 to 5.91)0.045
  • Multiple linear regressions examining the cross-sectional associations between abnormal metabolite ratios in the primary motor cortex (independent variables) and gait performance while single and dual tasking (dependent variables)c,d among participants with mild cognitive impairment, adjusted for potential confounders (n = 20).

  • β significant (i.e. P < 0.05) indicated in bold.

  • β = coefficient of regression corresponding to a change in stride time variability.

  • a N-acetyl aspartate/creatine ratio < 1.17.

  • b Defined when Elementary School Recognition Certificate was passed.

  • c Separated analyses were used for each model.

  • d Walking while counting backwards by seven.

View this table:
Table 5

Association between high choline/creatine ratioa and gait velocity while single and dual tasking

Gait velocity
Single taskDual taskb
β (95% CI)P-valueβ (95% CI)P-value
High choline/creatine ratioa−26.56 (−45.27 to −7.86)0.009−41.92 (−76.60 to −7.24)0.022
Age1.46 (2.97 to 0.06)0.0581.13 (3.93 to 1.68)0.399
Female gender2.05 (23.34 to 19.24)0.8371.91 (41.38 to 37.57)0.918
Body mass index1.81 (4.37 to 0.75)0.1500.30 (4.45 to 5.05)0.891
Montreal Cognitive Assessment score2.21 (1.68 to 6.10)0.2400.69 (7.91 to 5.53)0.838
High education levelc20.07 (7.08 to 47.22)0.1334.24 (46.10 to 54.57)0.858
SWMH grade3.59 (17.09 to 9.90)0.5722.08 (27.10 to 22.94)0.859
  • Multiple linear regressions examining the cross-sectional associations between abnormal metabolite ratios in the primary motor cortex (independent variables) and gait performance while single and dual tasking (dependent variables)b,d among participants with mild cognitive impairment, adjusted for potential confounders (n = 20).

  • β significant (i.e. P < 0.05) indicated in bold.

  • β = coefficient of regression corresponding to a change in gait velocity.

  • a Choline/creatine ratio > 0.58.

  • b Walking while counting backwards by seven.

  • c Defined when Elementary School Recognition Certificate was passed.

  • d Separated analyses were used for each model.

Finally, we found that the risk of low N-acetyl aspartate/creatine in the primary motor cortex was 63% higher among participants with a major increase in stride time variability in dual task compared with those without major increase in stride time variability (Fig. 4A). The effect size of the difference in primary motor cortex volume of −0.16 (95% CI −1.04 to 0.72) indicated that there were no between-group differences with regards to the primary motor cortex volume. Lastly, the risk of high choline/creatine in the primary motor cortex was 63% higher in the case of a major decrease in gait velocity while dual tasking (Fig. 4B). The effect size of the difference in primary motor cortex volume between the participants with a major decrease in gait velocity while dual tasking compared with those without decreased gait velocity was 2.05 (95% CI 0.99 to 3.11), indicating that the primary motor cortex volume was 2.05 standard deviations lower in participants with a decreased gait velocity in dual-task (P = 0.016). Using the ‘Common Language Effect Size’ approach of McGraw and Wong, the participants with a decreased gait velocity under dual task had a 2.05-fold greater risk of having a smaller primary motor cortex volume than those without decreased gait velocity under dual task (Dunlap, 1999).

Figure 4

Forest plots for the risk difference of low N-acetyl aspartate/creatine* or high choline/creatine ratios and for the effect size of a smaller volume of the primary motor cortex among participants with mild cognitive impairment who experienced major gait disturbances while dual tasking compared with those who did not. Horizontal lines correspond to the 95% confidence interval. The vertical line corresponds to a risk difference or an effect size of 0.0, equivalent to no between-group difference. (A) Participants with major increase in stride time variability while dual tasking. (B) Participants with major decrease in gait velocity while dual tasking. Cho = choline; CI = confidence interval; Cr = creatine; GV = gait velocity; MCI = mild cognitive impairment; MRS = magnetic resonance spectroscopy; NAA = N-acetyl aspartate; STV = stride-to-stride variability of stride time. *N-acetyl aspartate/creatine ratio < 1.17. †Choline/creatine ratio > 0.58. ‡Major decrease in gait velocity while dual tasking (i.e. walking while counting backwards by seven) defined as being in the lowest quintile of Δgait velocity; major increase in stride time variability while dual tasking defined as being in the highest quintile of Δstride time variability.

Discussion

This study of older adults with mild cognitive impairment showed that the metabolite ratios and volume of the primary motor cortex were associated with gait performance in both single and dual task conditions. Specifically, slower gait velocity was associated with higher choline/creatine and a smaller primary motor cortex volume, while greater stride time variability was associated with lower N-acetyl aspartate/creatine and to a lesser extent with a smaller primary motor cortex volume. Since slower gait velocity and greater stride time variability are robust markers of mobility decline and falls (Verghese et al., 2008; Beauchet et al., 2009; Montero-Odasso et al., 2012), our findings may contribute to a better understanding of the higher risk of mobility decline and falls in people with mild cognitive impairment. Moreover, participants whose gait was more compromised while dual tasking had a 63% higher risk of presenting abnormal metabolite ratios in the primary motor cortex. As well, those who decreased their gait velocity in dual task had a >2-fold greater risk of having a smaller primary motor cortex volume. This result indicates that the primary motor cortex is not only associated with gait per se but also with the maintenance of a constant walking pattern while attention demands are increased.

Our study provides evidence of an association between gait performance and the neurochemistry and volume of the primary motor cortex in older adults with mild cognitive impairment. While the role of the primary motor cortex in the executive network of normal locomotion is increasingly recognized (la Fougère et al., 2010; Jahn and Zwergal, 2010), the involvement of the primary motor cortex dysfunction in gait disorders in mild cognitive impairment and dementia has not been questioned until now. For instance, only two studies have examined gait in relation to 1H-MRS during the early stages of dementia (Ben Salem et al., 2008; Zimmerman et al., 2009), and neither of them focused on the primary motor cortex. One study reported that stride length variability correlated negatively with hippocampal N-acetyl aspartate/creatine (r = −0.56, P = 0.04) in 14 participants with mild cognitive impairment (Zimmerman et al., 2009), which could be explained by the role of the hippocampus in the retrieval of complex foot movement sequences necessary for regular gait patterns (Lafleur et al., 2002). In the other study, the authors selected 67 participants with probable pre-dementia (Mini-Mental State Examination score ranging between 25 and 30; diagnosis of mild cognitive impairment not specified) and showed that a modified Tinetti scale correlated positively with N-acetyl aspartate/creatine in the basal ganglia (r = 0.016, P = 0.05) (Ben Salem et al., 2008); a result consistent with previous studies highlighting the major role of the basal ganglia in single-task gait (Whelan, 1996). These two 1H-MRS studies supported the idea that the supratentorial grey matter is involved in gait control, but unfortunately they did not determine whether the primary motor cortex dysfunction is involved in gait disorders in mild cognitive impairment. The potential involvement of the primary motor cortex appears essential, as it is part of the final common pathway of the cerebral motor program to the spinal cord. Any impairment of the primary motor cortex may result in an altered motor control signal even in the absence of other brain abnormalities, including those in the hippocampus and basal ganglia. Briefly, the cerebral motor programme is organized into five consecutive steps: sensory feedback, intention, planning, programming and execution (Beauchet et al., 2008). The latter step of execution (i.e. the conversion of motor programs into movements) depends on the motor cortex, particularly the primary motor cortex (Christensen et al., 2000; Graziano et al., 2002; Meier et al., 2008). The successful completion of any movement, including walking, therefore depends on the proper functioning of the primary motor cortex (Christensen et al., 2000), so does the adaptation of the locomotion to environmental influences (Chan, 1983; Alain et al., 2007). Since the course of dementia is marked by thinning of the primary motor cortex and changes of cortical excitability (Im et al., 2008; Niskanen et al., 2011), it is therefore not surprising that we found the neurochemistry and volume of the primary motor cortex were associated with gait performance in mild cognitive impairment.

The literature on brain volumetry and gait disorders in the prodromal stage of dementia remains extremely scarce (Annweiler et al., 2012). A previous attempt to find a correlation between stride length variability and the volume of the hippocampus in individuals with mild cognitive impairment was not successful (Zimmerman et al., 2009). Our study provides new insight into regional brain atrophy in mild cognitive impairment associated with gait disorders. We found that gait performance was associated with the volume of the primary motor cortex specifically. Conversely, neither gait velocity nor stride time variability correlated with the volumes of the frontal cortex (executive control function), the parietal cortex (visuospatial attention), the somatosensory cortex (sensory homunculus), the hippocampus (memory), the posterior cingulate (motor imagery), the basal ganglia nuclei (automatic motor skills) and the thalamus and cerebellum (balance) (Table 3). These results underscore the likely involvement of the primary motor cortex changes in the onset of high-level gait disorders during the prodromal stage of dementia and reinforce our initial hypothesis of altered neurochemistry in this tissue region.

Interestingly, we found that both the propulsion and stability components of gait were explained by the primary motor cortex neurochemistry, but each by a different metabolite ratio. Stride time variability was associated with N-acetyl aspartate/creatine, a neuronal marker of health and biochemical function (Moffett et al., 2007), whereas gait velocity was associated with choline/creatine, a marker of inflammatory processes (Ferraz-Filho et al., 2009). Inflammation is long known to be associated with poorer physical performance in older adults (Cesari et al., 2004). The most commonly accepted explanation is that chronic inflammation influences physical performance by its effect on body composition, namely by accelerating hypermetabolism and relative anorexia (Roubenoff et al., 1994), with consequent loss of muscle mass and strength (Visser et al., 2002). Interestingly, our finding of an association between slower gait velocity and higher choline/creatine locally in the primary motor cortex suggests that inflammation may influence physical performance not only by its catabolic effect on muscles but also by a local inflammatory response in the brain, particularly in the primary motor cortex. Although unspecific, inflammatory changes illustrated by high choline/creatine are likely due, in the absence of brain tumour or infection, to chronic ischaemic demyelination (Karaszewski et al., 2010). Thus, the finding of high choline/creatine in the primary motor cortex of our participants with mild cognitive impairment may also support the notion that decreased gait velocity in mild cognitive impairment is related to cerebrovascular disease, consistent with previous literature showing slower gait velocity in the presence of SWMH on the corticospinal tract containing efferent motor fibres originating from the primary motor cortex (Annweiler and Montero-Odasso, 2012). For instance, it has been reported in 61 older adults with mild cognitive impairment that the grade of SWMH was associated with a poor performance on the Timed Up and Go test (Onen et al., 2008). This result is also in line with previous diffusion tensor imaging studies arguing that the disconnection by SWMH of motor networks served by the corticospinal tract is responsible for gait disorders (Bhadelia et al., 2009; Srikanth et al., 2010; de Laat et al., 2011a, b), specifically slower gait velocity (de Laat et al., 2011a) and falls (Koo et al., 2012).

Beyond inflammatory changes in the primary motor cortex, which appear mainly related to gait velocity, we also found that impaired function in the primary motor cortex, illustrated by lower N-acetyl aspartate/creatine, was associated with unstable gait and increased stride time variability. This result reinforces the current thinking that stride time variability may serve as a surrogate marker of the function and efficiency of higher-level motor control (Hausdorff, 2005). For instance, an indication in this direction was the finding of an inverse association between stride time variability and motor control function in older adults (Hausdorff, 2005; Beauchet et al., 2008, 2012) or the observation that patients with Parkinson’s disease and Alzheimer’s disease have higher stride time variability than cognitively normal older adults (Beauchet et al., 2008). Our results support the idea that the regularity of gait is associated with neural function, even at the molecular level, and from the early stages of dementia.

Besides gait performance per se, we also found that the primary motor cortex neurochemistry and volume were associated with the ability to maintain a stable walking pattern while attention demands were increased. The dual-task paradigm, which requires participants to perform an attention-demanding task while walking to assess changes compared with the reference single-task paradigm, explores the involvement of cortical level in gait control (Beauchet et al., 2005). The underlying hypothesis is that the changes in gait observed when dual tasking, such as decreased gait velocity or increased stride time variability, are caused by competition for cortical resources between the two simultaneously performed tasks (Beauchet and Berrut, 2006; Montero-Odasso et al., 2012). The ability to divide attention and properly perform two tasks simultaneously relies mainly on the frontal-executive functions (Miyake et al., 2000), which can become impaired with cerebrovascular disease owing to the localization of SWMH on the frontal–subcortical circuits (Oosterman et al., 2004), as evidenced here by the association between greater SWMH grade and worse gait performance in dual task (Table 4). However and importantly, in the current study, participants experiencing major gait disturbances in dual task compared with single task had also greater risks of having abnormal primary motor cortex metabolite levels and a smaller primary motor cortex volume (Fig. 4). These results indicate that the ability to maintain a constant walking pattern while attention demands are increased may also be supported in part by the primary motor cortex. Such findings are consistent with prior transcranial magnetic stimulation observations providing evidence that, although rhythmical gait movements seem automatic and reflexive in nature, they actually rely on a neuronal network that involves the motor cortex (Barthélemy et al., 2011).

The strengths of this study included the collection of data from a single scanner and research centre, the rigorous 1H-MRS acquisition and post-processing methods leading to reduced data variance (Kowalczyk et al., 2012) and the reliability of the Freesurfer morphometric procedures (Han et al., 2006). Magnetic resonance spectroscopy was performed using a 3.0 T scanner, which provides greater signal-to-noise ratio and greater spectral resolution compared with lower field strengths (Bartha et al., 2002). Additionally, gait performance was measured using a standardized and validated automated computerized gait analysis technique, and regression models were applied to measure adjusted associations.

Regardless, a number of limitations also existed. First, we analysed metabolite levels by normalizing to creatine, rather than calculating absolute concentrations. However, most studies also use ratios, and creatine level is considered relatively stable. Second, 1H-MRS data were acquired from a single region of interest, based on our a priori hypothesis and supported by the results of the current volumetric analysis. Although acquiring spectroscopy data from multiple brain locations is difficult due to the length of time required for data acquisition, future studies could be designed to determine whether the associations between gait measures and metabolite levels are specific to the motor cortex. Third, at present, our study is cross-sectional, which limits conclusions regarding causality. A future prospective analysis of the ‘Gait and Brain Study’ cohort should examine metabolite changes in the primary motor cortex and other regions of interest over time and elucidate whether such changes have prognostic significance on gait decline. Finally, although highlighting an association between the primary motor cortex and gait in amnestic mild cognitive impairment specifically would provide greater insight into gait disorders across the Alzheimer’s disease spectrum, subtyping mild cognitive impairment was not pursued in our analysis due to the limited sample size.

Our results showed for the first time among older adults diagnosed with mild cognitive impairment that abnormal metabolite ratios in the primary motor cortex and a lower primary motor cortex volume were associated with poor gait performance while single and dual tasking. Taken together, these results underscore the possible involvement of decreased neuronal function and viability in the primary motor cortex causing gait disorders observed in mild cognitive impairment. Additional work is needed to better understand the determinants of gait disorders in the course of cognitive decline and dementia. Understanding the neuroanatomical correlates of higher-level gait disorders may offer a powerful mechanism to act on mobility decline and falls in older adults with mild cognitive impairment and to maintain function late in life.

Funding

The ‘Gait and Brain Study’ is supported by an operating grant from the Canadian Institutes of Health Research (Principal investigator: Manuel Montero-Odasso; operating grant number MOP 211220). C.A. holds a research fellowship at the Gait and Brain Lab and the University of Western Ontario, funded by a grant from the Canadian Institutes of Health Research - Institute of Aging (CIHR-IA), and holds a research grant from Servier Institute in France. M.M-O.’s program in ‘Gait and Brain’ function is supported, in part, by grants from the CIHR-IA, the Drummond Foundation, the Physician Services Incorporated Foundation of Canada (PSI), the Ontario Ministry of Research and Innovation, and by Department of Medicine Program of Experimental Medicine (POEM) Research Award from the University of Western Ontario. He is the first recipient of the Schulich Clinician-Scientist Award and recipient of the CIHR New Investigator Award.

Acknowledgements

The authors thank Susan W. Muir, PhD, Karen Gopaul, MSc and Anam Islam, MSc, from the ‘Gait and Brain Lab’, Lawson Health Research Institute, the University of Western Ontario, London, Ontario, Canada, for their help in the participants’ assessment and data gathering. Irene B. Gulka, MD, from the Department of Radiology, London Health Sciences Centre, London, Ontario, Canada, for her kind advice. Izabella Kowalczyk, MSc, and Sandy Goncalves, MSc, from the Robarts Research Institute, Department of Medical Biophysics, Schulich School of Medicine and Dentistry, the University of Western Ontario, London, Ontario, Canada, for their kind advice. There were no compensations for these contributions.

Abbreviations
1H-MRS
proton magnetic resonance spectroscopy
SWMH
subcortical white matter hyperintensities

References

View Abstract