We assessed day-to-day limb-use strategies in individuals with acquired (n = 18, 10 below elbow) or congenital (n = 11, 10 below elbow) hand absence (see Table 1 for demographic details and prosthetic limb usage). Specifically, we quantified the extent to which individuals utilise their residual arms, relative to their intact hands (Figure 1A), in daily activities. We collected data from acceleration monitors, worn by a sub-set of 20 participants (8 congenital) while engaged in their normal routines. A laterality index, reflecting the relative number of movements performed by the intact hand vs the residual arm (Figure 1B), showed that both groups moved the intact hand more than the residual arm (t(7) = 3.48, p=0.01; t(11) = 10.60, p=0.001, one-sample t-test compared to zero for the congenital and acquired groups, respectively). This could be due both to greater use of the intact hand and also to the different position of the accelerometer on the two sides of the body (wrist vs arm), resulting in different acceleration profiles. Our main aim was to compare the degree of this laterality across the two 1-handed groups. We found that the acquired group exhibited stronger relative reliance on their intact hand, as demonstrated in a significantly greater lateralisation scores towards the intact hand, compared to the congenital group (t(18) = −2.67, p=0.016, Figure 1B).

Figure 1

Download asset Open asset

These findings were further confirmed using questionnaire ratings from all 1-handed participants, quantifying the extent to which individuals incorporate their residual arm in daily tasks (‘Materials and methods’). Ratings were significantly greater for the congenital group (t(1,27) = 3.65, p=0.001), suggesting they use their residual arm more frequently in their daily routine, compared with the acquired group (Figure 1C). This group difference remained significant after accounting for level of amputation and degree of functional and cosmetic prosthetic limb usage as covariates (F_(1,24) = 8.217, p=0.009). The questionnaire ratings for residual arm usage were negatively correlated with the accereometry-based laterality indices (r₍₁₈₎ = −0.53, p=0.012, Figure 1D), validating the questionnaires as a measurement of habitual usage strategies between the residual and intact limbs. A similar correlation was also obtained after accounting for level of amputation and degree of functional and cosmetic prostheses usage (r₍₁₅₎ = −0.451, p=0.034).

Based on these dissociated adaptive motor strategies, we predicted distinct patterns of limb representation between the two 1-handed groups. All 1-handed participants and 22 intact (2-handed) controls underwent fMRI, involving simple unilateral hand and arm movements (fingers/elbow flexion and extension). Voxel-wise activation (β) patterns were compared between the three groups during hand and arm movements across the entire brain. Each of these two separate whole-brain contrasts resulted in a single cluster, centred on the hand knob of the central sulcus of the deprived hemisphere, spanning both the pre- and the post-central gyri. In line with their increased usage of the residual arm, congenital 1-handed participants showed increased activation during residual arm movements, compared with the other two groups (Figure 2B). The resulting cluster showed two distinct local peaks of activation—in the anterior and posterior banks of the central sulcus (t = 3.73; x = 38, y = −20, z = 54 for the anterior peak and t = 3.93; x = 48, y = −20, z = 54 for the posterior peak, coordinates are based on the MNI 152 brain template). Conversely, acquired 1-handed participants, who displayed increased use of their intact hand, correspondingly showed significantly increased activation during intact hand movements compared to the other groups (Figure 2C). The resulting cluster had a single peak, centred on the posterior bank of the central sulcus (t = 5.24; x = 42, y = −18, z = 48 in MNI 152 space).

Figure 2

Download asset Open asset

Even though activations used to generate these two contrasts were measured during movements of different limbs (hand vs arm), and on different sides of the body (limbless side vs intact side), the differential activation was restricted in both cases to the hand area, as demonstrated by activation patterns in the control group (Figure 2A and inserts in Figure 2B,C). No other significant clusters were identified in these comparisons, or using the homologous contrasts between the other groups.

As demonstrated in Figure 2, the two independent clusters, showing increased representation of the residual arm or the intact hand in the congenital and acquired groups, overlapped with the hand area of the control participants. This suggests that the area that would typically represent the missing hand has been recruited to support increased representation of the residual arm or intact hand. To investigate use-dependent plasticity specifically in the ‘deprived cortex’, we constructed an independent region of interest (ROI), based on the conjunction between phantom/nondominant hand movements in acquired amputees/2-handed controls. Amputees often experience vivid sensations of a phantom hand. Recently, this phenomenology has been supported by empirical evidence demonstrating that movements of a phantom limb elicit both central and peripheral motor signals that are different from those found during imagined movement (Reilly et al., 2006; Raffin et al., 2012). This phenomenon of maintained representation in the sensorimotor cortex, allowed us to reliably localise the representation of the missing hand (Makin et al., 2013). Here, this pre-defined ROI was used to interrogate activity relating to other body parts, thus allowing us to test the usage-driven hypotheses independently. The borders of this ‘deprived cortex’ ROI, spanning primary somatosensory and motor cortices, are shown in Figure 3A. As the spatial resolution and co-registration methods used here are insufficient to reliably dissociate the somatosensory and motor primary cortices, cortical areas spanning the pre- and post-central gyri will be described here as ‘sensorimotor’.

Figure 3

Download asset Open asset

We next extracted fMRI activation values (β) during movement conditions within the independent ROI. A mixed effect analysis of variance showed a significant interaction between groups and limbs (F_(2,47) = 3.67, p=0.033), reflecting dissociated recruitment of the deprived cortex by movements of the residual arm and intact hand in the two 1-handed groups (Figure 3B–C). Post-hoc tests confirmed increased activation when congenital participants moved their (relatively over-used) residual limb, compared to acquired participants (Mann Whitney U = 6, n = 20, p=0.001) and compared to controls (t(30) = 2.37, p=0.025). Similarly, greater activation within the deprived cortex was observed when acquired participants moved their (relatively over-used) intact hand, compared to congenital participants (t(26) = −3.68, p=0.001) and controls (t(37) = 2.96, p=0.005). These results were independent of participant A07, who lost her hand at a young age (t>2.81, p<0.008) (note that this participant was not included in the arm analysis above, due to her amputation level). The difference between the two 1-handed groups was also significant when accounting for level of amputation, and degree of functional and cosmetic prosthetic limb usage (F_(1,15) = 11.357, p=0.004 and F_(1,23) = 8.172, p=0.009 for residual arm and intact hand movements, respectively).

To complement the between-group findings reported above, we also performed within-group analyses to test for assymetrical patterns of limb representation. In particular, we tested whether the activity in the deprived cortex during movement of the over-used (compensatory) limb was greater than activity in the homologous cortex (in the intact hemipshere) during movements of the corresponding (opposite) limb (Figure 4). For the congenital group, activation in the deprived cortex during movements of the (contralateral) residual arm was greater than activation in the homolgous intact cortex during movements of the (contralateral) intact arm (t(1,9) = 2.45, p=0.037). Similarly, for the acquired group, activation in the deprived cortex during movements of the (ipsilateral) intact hand was greater than activation in the homologous intact cortex during movements of the (ipsilateral) phantom hand (Makin et al., 2013) (t(1,16) = 3.08, p=0.007). These results confirm that, within both groups, activity in the deprived cortex relating to the compensatory limb is greater than activity in the homologous cortex, relating to the corresponding opposite limb.

Figure 4

Download asset Open asset

In addition to the hypothesised increase in activation for the limb favoured for adaptive usage, the ROI analysis in Figure 3B–C also revealed unexpected reduced activation in the 1-handed groups during movements of the limb which is not favoured for over-use, compared with controls. Specifically, the congenital group, who over-use the residual arm, showed reduced activation in the deprived cortex during movements of the intact hand (t(31) = −2.61, p=0.014), whereas the acquired group, who over-use the intact hand, showed reduced activation in the deprived cortex during movements of the residual arm (Mann Whitney U = 97, n = 39, p=0.025). This result potentially reflects a competitive relationship between the representations of the two limbs in the deprived cortex in 1-handed individuals.

To test whether functional reorganisation associated with adaptive movement strategies is present outside the primary sensorimotor cortex, we also explored patterns of limb representation in the cerebellar hand area (lobule V of the anterior lobe) of the acquired and congenital 1-handed groups. We defined an independent ROI for the deprived cerebellum, using the same criteria as used to construct the deprived cortex ROI. Note that the deprived cerebellum ROI is ipsilateral to the missing hand (and opposite to the deprived cortex ROI). Similarly to the deprived cortex, the deprived cerebellum showed increased activation in the congenital group during residual arm movements (t(26) = −3.33, p=0.003), and increased activation in the acquired group during intact hand movement (t(18) = 2.75, p=0.013). This suggests that the patterns of altered representation identified in this study are not limited to reorganisation of body-part representations in the primary sensorimotor cortex.

For the group results described above, we have compared the intact hand of 1-handed participants to the dominant hand of controls, and the residual arm of 1-handed participants to the non-dominant arm of controls. However, a third of the individuals with acquired deprivation lost their dominant hand (Table 1). To test whether group differences in hand dominance could influence the results, activation levels for non-dominant hand and dominant arm movements were extracted from a homologous ROI contralateral to the dominant hand, in an age-matched sub-group of the control participants. ROI-based comparisons between the acquired and control groups were repeated as above, while accounting for handedness using this sub-group, with similar results (t(30) = 3.1, p=0.004 for increased representation during intact hand movements and t(37) = −2.25, p=0.032 for decreased residual arm representation during residual arm movements in acquired amputees vs controls).

Phantom limb pain, prevalent in acquired amputees, may discourage these participants from using their residual arm. Indeed, within the acquired group a trend for a correlation between chronic phantom pain scores and usage was found (r₍₁₆₎ = −0.40, p=0.099), such that individuals with worst chronic pain tend to use their residual arm less. Therefore, to test whether phantom pain could explain our observed imaging results, we repeated the comparisons between the two 1-handed groups described above (using a univariate GLM) with chronic phantom pain ratings as a covariate. Increased activation was maintained both in the congenital group (relative to the acquired group) during residual arm movements (F_(1,19) = 12.57, p=0.002) and in the acquired group (relative to the congenital group) during intact hand movements (F_(1,27) = 6.10, p=0.021). These results suggest that the activation patterns presented above are not merely epiphenomenal to phantom pain.

Although the congenital and acquired 1-handed groups differed in terms of usage strategies and activation patterns in the deprived cortex, individuals across these groups showed considerable overlap (Figure 3D): acquired 1-handed individuals who nevertheless had high residual arm usage showed less of the intact hand activation that was typical of their group. Such individuals demonstrate that the differential reorganisational patterns observed in the two 1-handed groups may not be solely determined by deprivation age (or previous afferent/efferent experience), but rather could relate to usage strategies. Indeed, when considering usage across both 1-handed groups we found that greater usage of the residual arm was associated with less activation during intact hand movements in the deprived cortex (r₍₂₆₎ = −0.43, p=0.021), even when accounting for age of sensory deprivation (r₍₂₅₎ = −0.36, p=0.032). This effect was strengthened when repeating the same partial correlation while only considering the prosthetic limb users (r₍₁₉₎ = −0.61, p=0.003 and r₍₁₈₎ = −0.56, p=0.005, respectively), suggesting that rehabilitation may strengthen the relationship between adaptive plasticity and usage, however further research will be required to identify the relationship between prosthesis usage and adaptive brain plasticity.

Finally, the white matter connections supporting the functional changes associated with limb-usage were studied in voxel-wise comparisons (Smith et al., 2006) of fractional anisotropy (FA, a measure of white matter microstructure), along the white matter skeleton. No significant differences in FA were found between the three groups, suggesting that deprivation in itself may be insufficient to induce significant white matter change (although see Langer et al. (2012) for a study showing reduced FA in the corticospinal tract following several weeks of arm immobilisation). However, across both 1-handed groups, individuals showing greater activation in the deprived cortex during intact hand movements had higher FA values in the corticospinal tract and the inferior fronto-occipital fasciculus (IFO) in the deprived hemisphere (Figure 5). This voxel-wise non-parametric correlation was performed while accounting for participants’ age and deprivation age (as nuisance regressors). No significant correlations were identified between FA and degree of activation in the deprived cortex during residual arm movements.

Figure 5

Download asset Open asset

18 individuals with sustained unilateral upper limb amputation (mean age ± s.e.m = 46 ± 3, 6 with absent right hand), 11 individuals with a congenital unilateral upper limb deficit (age = 35 ± 4, 3 with absent right hand) and 22 intact controls (age = 41 ± 3, 7 left hand dominant) were recruited for the study (see Table 1 for demographic details). Recruitment was carried through the Oxford Centre for Enablement and Opcare in accordance with NHS national research ethics service approval. Informed consent and consent to publish was obtained in accordance with ethical standards set out by the Declaration of Helsinki (1964) and with procedures approved by the NHS (REC ref: 10/H0707/29). Data from one individual with acquired deprivation and one control were discarded, due to excessive head movements (>3 mm) during fMRI and problems during DTI data acquisition (respectively).

In order to assess potential differences in rehabilitation strategies between the two amputee groups, use of residual arm was initially assessed using a revised version of the Motor Activity Log (Uswatte et al., 2006). 1-handed participants were requested to rate how frequently (never; sometimes; very often) they incorporate their residual arm (stump; either directly, or using a prosthesis) in an inventory of daily activities, requiring varying degrees of motor control. The following items were used: taking money out of wallet; opening envelope; picking up/lifting glasses; picking up/holding up a phone; wiping off a kitchen counter or other surface; getting out of a car; stabilizing paper while writing; stabilizing dishes while carrying; carrying a cup or a can; carrying bags; getting up from a chair with arm rests; pulling chair away from table before sitting down; holding a book or a magazine/turning pages; typing on a keyboard/pressing mouse buttons; controlling a computer mouse; putting on your socks; putting on your shoes; tying shoe laces; inserting your (intact) arm through a sleeve; putting on makeup base, lotion, or shaving cream on face; washing hand or face; drying your hand or face; combing your hair; buttoning a shirt; zipping up a coat; peeling fruit skin; using a fork or spoon for eating.

As previous motor activity logs were designed to assess mobility of a paralyzed hand, the questionnaire was modified while considering the unique confounds of unilateral amputees (Lindner et al., 2010), and was aimed to accommodate participants with various levels of deprivation. In accordance with the original questionnaire, the items represented commonly encountered actions and covered a comprehensive range of activities, to accommodate various levels of skill of the residual arm. We focused on frequency, rather than quality of movement rating, as we were interested in both prosthetic limb and stump usage, which could not have been easily assessed. We used a scale of three frequency ratings, rather than five, since pilot testing showed that individuals tended to ignore the second and fourth options. Each item was scored (between 0 and 2), and the sum was divided by 54, such that individuals were rated on a scale between 0 to 1.

To validate the usage questionnaires, and to study the relationship between increased reported residual arm usage and intact hand usage, limb acceleration data were collected from a subset of 21 limb-absent individuals (8 with congenital absence) using the GeneActiv accelerometers (sample rate: 100 Hz). Participants were given two strapped sensors, which they were asked to place on their wrist (intact hand) or the proximal aspect of the upper arm (residual arm) during 2 days with a typical daily routine (mean number of hours per day ± s.e.m. = 15:08 ± 37 min). Data from one acquired participant was discarded, due to hardware malfunction. Data for each of the three movement axes was initially smoothed using a 500 ms kernel, to discard high frequency noise. To quantify the number of movements executed with each limb, the difference between maximal and minimal acceleration values were initially calculated within a sliding window of 400 ms for each movement axis separately. In each time window, movements were identified as difference in acceleration that was above a threshold of 0.2 m/s² in at least one axis, provided that this increase was preceded and followed by periods of no movement (a difference between maximal and minimal acceleration below 0.2 m/s²) in all axes. To account for whole body movements, as well as differences in number of hours of recordings, a ratio between the two limb movements was used, rather than an absolute number of movements. The movement laterality ratio was chosen [(intact−residual)/(intact+residual)], as it portrays the extent of intact hand usage, given the contribution of residual arm movements.

Task-based fMRI: participants were visually instructed to move their left/right hand (finger movements), left/right arm (elbow movements), feet (bilateral toe movements) or lips. The protocol comprised of alternating 12 s periods of movement and ‘rest’. Each of the six conditions was repeated four times, in a counterbalanced order. Here we focus on results from intact/dominant hand and residual/nondominant elbow movements (see Makin et al., 2013) for information about phantom movements). Participants received extensive training on the degree and form of movements expected. Note that the movements were easy to execute and did not require any expertise. To confirm that appropriate movements were made at the instructed times, task performance was monitored visually both on- and off-line, using video recordings.

The MRI measurements were obtained using a 3 Tesla Verio scanner (Siemens, Erlangen, Germany) with a 32-channel head coil. Anatomical data were acquired using a T1-weighted magnetization prepared rapid acquisition gradient echo sequence (MPRAGE) with the parameters: TR = 2040 ms; TE = 4.7 ms; flip angle = 8°, voxel size = 1 mm isotropic resolution. Functional data based on the blood oxygenation level-dependent (BOLD) signal were acquired using a multiple gradient echo-planar T2*-weighted pulse sequence, with the parameters: TR = 2000 ms; TE = 30 ms; flip angle = 90°; imaging matrix = 64 × 64; FOV = 192 mm axial slices. 46 slices with slice thickness of 3 mm and no gap were oriented in the oblique axial plane, covering the whole cortex, with partial coverage of the cerebellum. Two sets of whole brain diffusion weighted volumes were acquired using a generalized autocalibrating partially parallel acquisitions (GRAPPA) sequence, using the following parameters: 60 directions (plus 8 vol without diffusion weighting); b = 1000 s/mm; 65 axial slices; voxel size 2 × 2 × 2 mm; TR = 9600 ms; TE = 87 ms.

All imaging data were processed using FSL 5.1 (www.fmrib.ox.ac.uk/fsl). Data collected for individuals with absent right hands (6 acquired and 3 congenital participants), were mirror reversed across the mid-sagittal plane prior to all analyses so that the hemisphere corresponding to the missing hand was consistently aligned. Data collected for left-hand dominant controls (n = 7) were also flipped, in order to account for potential biases stemming from this procedure.

Functional data were analysed using FMRIB’s expert analysis tool (FEAT, version 5.98). The following pre-statistics processing was applied to each individual run: motion correction using FMRIB’s Linear Image Registration Tool (MCFLIRT [Jenkinson et al., 2002]); brain-extraction using BET (Smith, 2002); mean-based intensity normalization; high pass temporal filtering of 300 s; and spatial smoothing using a Gaussian kernel of FWHM (full width at half maximum) 5 mm. Time-series statistical analysis was carried out using FILM (FMRIB’s Improved Linear Model) with local autocorrelation correction. Functional data were aligned to structural images (within-subject) initially using linear registration (FMRIB’s Linear Image Registration Tool, FLIRT), then optimized using Boundary-Based Registration (Greve and Fischl, 2009). Structural images were transformed to standard MNI space using a non-linear registration tool (FNIRT), and the resulting warp fields applied to the functional statistical summary images.

To compute task-based statistical parametric maps, we applied a voxel-based general linear model (GLM), as implemented in FEAT. The block design paradigm was convolved with a gamma function (Friston et al., 1998), and its temporal derivative was used to model the activation time-course at the individual level. For our main comparisons, contrasts between the intact/dominant hand and the residual/nondominant arm conditions were defined against the feet condition. Further contrasts were defined between intact/dominant arm (in the congenital group) and nondominant/missing hand (in the acquired group) vs feet movements. Group level analysis of spatial maps was carried out using FMRIB’s Local Analysis of Mixed Effects (FLAME). Intact hand movements in the 1-handed groups was compared with dominant hand movements in controls, and residual arm movements were compared with non-dominant arm movements. Data from participants with above elbow deprivation (1 congenital and 7 acquired participants) were excluded from the analysis of activity during arm movements. The cross-subject GLM included planned comparisons between each of the groups against the other two, while accounting for the unbalanced comparison. Z (Gaussianised T/F) statistic images were thresholded using clusters determined by Z>2 and a family-wise-error corrected cluster significance threshold of p<0.05 was applied to the suprathreshold clusters.

To visualise the human homunculus, activation maps during feet, hands, arms and lip movements in the control group were thresholded at Z>3, each map was mirror flipped and maps representing each of the four body parts were averaged, such that a symmetrical representation of each body part was achieved. For presentation purposes, statistical parametric activation maps were projected on the inflated surface of a representative participant’s cortex, using FreeSurfer.

We followed the same ROI selection and thresholding procedures as previously reported in Makin et al. (2013). Briefly, it has been demonstrated that movements of a phantom limb elicit both central and peripheral motor signals that are different from those found during imagined movement (Reilly et al., 2006; Raffin et al., 2012). Since the acquired, but not the congenital participants, displayed group activation in the sensorimotor hand knob during phantom hand movements (Makin et al., 2013), the congenital group was excluded from the ROI definition. The ‘deprived cortex’ ROI was therefore defined using the conjunction of missing/non-dominant hand movements (compared to feet movements) in the acquired and control groups only. A pre-determined threshold of Z>7 was chosen, yielding a single cluster, centred on the hand knob of the central sulcus, contralateral to the missing hand. A second cluster in lobule V of the cerebellum, ipsilateral to the nondominant/phantom hand was defined separately. For within-participant comparisons, a homologous ROI for the intact hand was defined using the conjunction between controls and all 1-handed participants (using a pre-determined threshold of Z>8), yielding a comparable cluster contralateral to the intact hand. GLM parameter estimate values (β) of the low level statistical parametric maps of the contrasts described above were extracted from all voxels underlying the pre-determined ROIs and then averaged for each participant.

Voxelwise statistical analysis of the FA data were carried out using Tract-Based Spatial Statistics (TBSS [Smith et al., 2006]). TBSS projects all subjects’ FA data onto a mean FA tract skeleton, before applying voxelwise cross-subject statistics. Diffusion data were initially corrected for eddy-currents and head motion using affine registration to the average of the non-diffusion-weighted volumes. A diffusion tensor model was fitted at every voxel to derive FA maps, which were non-linearly registered to group-specific templates created from 11 participants of each group (FA thresholds > 0.2). The resulting FA images were temporally concatenated into a single 4D file and averaged to create a mean ‘skeleton’, representing the centre of all white matter tracts onto which participant-specific FA values were projected. Tract-based spatial statistics was performed voxel-wise, using a GLM and permutation-based non-parametric testing. The GLM included either planned comparisons between the three groups, or the beta values extracted from the deprived cortex ROI during intact hand movements for each of the individuals with hand absence. An age regressor of no interest was also included for the correlation analysis. Clusters were formed at t > 2 and tested for significance at p<0.05, corrected for multiple comparisons across space. For presentation purposes, the FA clusters were dilated to fill the corresponding tracts.

For analysis of limb-use data, independent-samples two-tailed t-tests were used to compare congenital and acquired groups.

For ROI-based analysis of fMRI activation, intact hand movements in the 1-handed groups was compared with dominant hand movements in controls, and residual arm movements were compared with non-dominant arm movements. Data from participants with above elbow deprivation were excluded from the analysis of activity during arm movements. Between-group effects were initially statistically compared using a mixed-level ANOVA. For each limb, two comparisons were planned based on our behavioural results. Therefore, for follow-up unpaired t-tests of these planned comparisons, a p-value of p≤0.025 was used, to account for multiple comparisons. Other within/between-group effects were statistically compared using paired/independent-samples two-tailed t-tests (α < 0.05). Where significant departure from normality was found (based on the Shapiro-Wilk test), nonparametric tests were used (Man-Whitney or Wilcoxon, as appropriate).

To test for correlations between questionnaire, acceleration data, and fMRI activation, two-tailed Pearson tests were used. To exclude the involvement of various confounds from the correlation analysis (e.g., level of amputation, age at deprivation), post-hoc one-tailed partial correlation analysis was performed. To exclude the involvement of various confounds (e.g., phantom pain) from the group comparisons, univariate GLM comparisons were carried, with the confounding measurement as a covariate. To account for the potential contribution of phantom pain on the BOLD differences, chronic levels of phantom pain were assessed in each 1-handed individual. Chronic pain magnitude was calculated by dividing pain intensity (0: ‘no pain’—10: ‘worst pain imaginable’) by frequency (1—‘all the time’, 2 —‘daily’, 3—‘weekly’, 4—‘several times per month’ and 5—‘once or less per month’). This measure therefore reflects the chronic aspect of the pain as it combines both frequency and intensity, as used previously (Draganski et al., 2006; Makin et al., 2013). Comparison between correlations was assessed using a two-tailed Fisher r-to-z transformation. Statistical analysis was carried with SPSS version 18.

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Supporting bodies: Royal Society; Marie Curie Actions; European Community’s Seventh Framework Programme FP7/2007-2013 under grant agreement number PITN-GA-2008-290011; Wellcome Trust; National Institute for Health Research (NIHR) Oxford Biomedical Research Centre based at Oxford University Hospitals NHS Trust and University of Oxford; MRC. We thank Nancy Rawlins for help with accelerometry data collection. We thank Scott Frey and Karen Reilly for helpful comments on a preliminary version of the manuscript. We thank Opcare for help with volunteer recruitment, and our participants for their contribution to the study.

Human subjects: Informed consent and consent to publish was obtained in accordance with ethical standards set out by the Declaration of Helsinki (1964) and with procedures approved by the NHS (REC ref: 10/H0707/29).

© 2013, Makin et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.