Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 1 of 32

1

Locomotor recovery in spinal cord injury - insights beyond walking speed and distance

Authors Lea Awai* PhD, Armin Curt1 MD 1

1

Spinal Cord Injury Center, Balgrist University Hospital, University of Zurich, CH-8008 Zurich, Switzerland.

Running head: Locomotor recovery in spinal cord injury *Corresponding author: Lea Awai Spinal Cord Injury Center Balgrist University Hospital Forchstrasse 340 CH-8008 Zürich Switzerland Tel: +41 44 386 37 34 Fax: +41 44 386 37 31 [email protected] Title characters: 85 Running head characters: 40 Word counts: - Abstract: 253 - Body: 4032 Figures: 4 (color) Tables: 1 References: 26

Armin Curt Spinal Cord Injury Center Balgrist University Hospital Forchstrasse 340 CH-8008 Zürich Switzerland Tel: +41 44 386 39 01 Fax: +41 44 386 39 09 [email protected]

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 2 of 32

2 Abstract Recovery of locomotor function following incomplete spinal cord injury (iSCI) is clinically assessed through walking speed and distance while improvements in these measures might not be in line with a normalization of gait quality and are on their own insensitive at revealing potential mechanisms underlying recovery. The objective of this study was to relate changes of gait parameters to the recovery of walking speed while distinguishing between parameters that rather reflect speed improvements from factors contributing to overall recovery. Kinematic data of 16 iSCI subjects were repeatedly recorded during inpatient rehabilitation. The responsiveness of gait parameters to walking speed was assessed by linear regression. Principal component analysis (PCA) was applied on the multivariate data across time in order to identify factors that contribute to recovery after iSCI. Parameters of gait-cycle and movement dynamics were both responsive and closely related to the recovery of walking speed, which increased by 96%. Multivariate analysis revealed specific gait parameters (intralimb shape normality and consistency) that, although less related to speed increments, loaded highly on PC1 (58.6%) explaining the highest proportion of variance (i.e., recovery of outcome over time). Interestingly, measures of hip, knee and ankle ROM showed varying degrees of responsiveness (from very high to very low) while not contributing to gait recovery as revealed by PCA. The conjunct application of two analysis methods distinguishes gait parameters that simply reflect increased walking speed from parameters that actually contribute to gait recovery in iSCI. This distinction may be of value for the evaluation of interventions for locomotor recovery.

Key words: spinal cord injury, recovery, responsiveness, locomotion

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 3 of 32

3 Introduction The improvement of functional outcome measures following treatment or rehabilitation programs is generally termed recovery while the evaluation of underlying factors contributing to recovery remains challenging. Clinically, functional recovery is assessed by standardized tests that evaluate the ability of a person to accomplish specific tasks representing activities of daily living (e.g., functional independence measure (FIM),1, Fugl-Meyer assessment (FMA),2 spinal cord independence measure (SCIM-III)3) while these functional assessments are specifically designed for a defined group of patients (e.g., stroke, multiple sclerosis, spinal cord injury). Locomotor recovery is primarily captured by measures of walking speed and distance4, 5 that can be easily compared between different groups and disorders. Being able to walk from A to B within a certain time is functionally relevant to most people and can be assessed without the need for elaborate equipment. However, monitoring walking speed alone is insufficient to reveal possible underlying mechanisms of recovery. The question is by what means do patients attain a faster walking speed? Possible explanations may include increased muscle strength6, 7 or learning of new movement strategies,8 some of which may rely on processes such as reorganization of neural circuits and sprouting of specific pathways at the level of the brain/brainstem and spinal cord.9, 10 In addition to clinical tests, electrophysiological recordings provide objective insights into the integrity of specific spinal pathways. The latencies of somatosensory- (SSEP) and motor evoked potentials (MEP) in incomplete spinal cord injury (iSCI) patients are indicative of the state of myelination of axons and were shown to remain relatively unchanged following injury despite functional improvements, suggesting that mechanisms such as compensation and adaptation of movement strategies may contribute to functional recovery.11 However,

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 4 of 32

4 neurophysiological measures cannot be used to extrapolate to motor recovery on a behavioral level. Thus, a possible approach to reveal mechanisms underlying motor recovery is the combination of a multivariate analysis considering the intricate relation between multimodal measures (no selection bias of single parameters), and the evaluation of responsiveness of single outcome metrics to a gold standard. Partial disruption of spinal pathways and diminished supraspinal input in iSCI provide the opportunity to assess the capacity to compensate for an impairment (improving effectiveness within the limits of impairment) and changes in the control of walking (i.e., plasticity allowing for adaptations in neural control) of the human locomotor system to recover walking function in the likely absence of true restitution via repair mechanisms.12-14 In the present study detailed kinematic and statistical analyses of gait parameters in ambulating iSCI subjects recorded at several time points during rehabilitation served to disentangle contributing factors (influencing recovery) from gait parameters that rather follow (i.e., are a consequence of) improved walking speed. We hypothesized that quantitative measures of speed and distance evolve differently from qualitative measures of complex gait coordination, and the distinct evaluation allows to discern increases in the effectiveness of walking from improvements in motor control.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 5 of 32

5 Materials and Methods Subjects Patients undergoing in-patient rehabilitation due to iSCI were recruited for this study as soon as they were able to ambulate without the assistance of an additional person. Patients with comorbidities influencing or inhibiting gait were excluded. Reference gait data were obtained from 11 healthy control subjects with no neurological disorder or any other impairment affecting their gait. All subjects had to give written informed consent prior to participation. The study was approved by the ethics committee of the Canton of Zurich, Switzerland.

Protocol Subjects were recorded at least at two time points during recovery. Healthy controls were recorded once. Subjects were asked to walk barefooted overground along a straight 8 m walkway at their preferred walking speed. The first and last few steps were excluded from analysis because of acceleration/deceleration effects. Several runs were recorded in order to obtain at least 10 gait cycles. Overground walking at preferred speed was chosen because subjects need to self-initiate and maintain walking (as opposed to treadmill walking) and it reveals a person's natural and inherent walking pattern. Patients were allowed to use an assistive device if needed while using the same device throughout all assessment time points. Lower-limb kinematics were recorded using 8 infrared cameras (T10, Vicon motion systems Ltd., Oxford, UK) at 200 Hz and two synchronized high-speed digital video cameras (pilot series, Basler AG, Ahrensburg, D) at 100 Hz. 16 reflective markers (16

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 6 of 32

6 mm diameter) were placed on bony landmarks of the lower body according to the Vicon Plug-in Gait model. Patients were allowed to rest whenever needed.

Outcome measures One gait cycle ranged from initial contact of one foot to the subsequent initial contact of the same foot. Data were time-normalized to one gait cycle by linear interpolation and data of at least 10 gait cycles were analyzed. Parameters of the more affected limb (lower motor score), or right limb in symmetric patients or control subjects, were used for analysis. Evaluated parameters included preferred walking speed [km/h] and gait-cycle parameters: step length [cm], cadence [strides/min], stance phase [%], and singlesupport phase [%]. Kinematic data provided information on lower-body intralimb coordination (gait quality) by evaluating hip-knee cyclograms, where the knee angle (abscissa) was plotted against the hip angle (ordinate). Further parameters extracted from kinematic data were hip-knee angular velocity at toe-off (resultant velocity vector of hip and knee angular change per unit time during stance-to-swing transition), ankle angular dorsiflexion velocity at toe-off, endpoint velocity at toe-off (velocity of the 2nd metacarpal marker in the sagittal plane), maximal hip-, knee-, and ankle range of motion (ROM) during a gait cycle, angular component of coefficient of correspondence (ACC)15 as a measure for hip-knee cyclogram consistency. The hip-knee cyclogram shape difference was quantified by the square root of the sum of squared distances (SSD)13 between the cyclogram of a patient (mean of all cycles) and a mean standard cyclogram (derived from control subjects). Foot clearance was defined as the maximal height above ground of the 2nd metacarpal marker during the first two thirds of swing phase [mm] (excluding the vertical peak just prior to heel strike).

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 7 of 32

7 A greater SSD reflects greater shape difference from normal and therefore a higher degree of impairment. The same is true for the stance phase: a shorter stance phase corresponds to improved gait. We therefore multiplied these two values by -1 in order to align all measures of progression. The improvement of a parameter was evaluated by its change from an early to a late time point (Δ value) expressed as absolute values and as a percentage change in order to illustrate the magnitude of improvement or decrement of gait-related variables over time. For further regression analysis, the recovery of a parameter was defined as its difference between the last time point and the first time point divided by the amount of days between the two time points. This rate of recovery (change per unit time) accounted for the different lengths of time intervals between the assessment dates.

Statistical analysis The improvement of each parameter quantified by the percentage increase was evaluated by comparing the values at late time point to early time point using a paired ttest for normally distributed data and Wilcoxon signed-rank test for non-normally distributed parameters. In order to evaluate the respective influence of different gaitrelated parameters on the outcome of walking speed, all parameters were normalized to have a mean value of 0 and standard deviation 1. This prevents an overestimation of big numbers (e.g., ROM, angular velocity) and, conversely, an underestimation of small values (e.g., ACC). The responsiveness to a change in walking speed of 14 potentially predictive gait-related parameters was assessed using a linear regression model:17

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 8 of 32

8 where

;

describes the change of walking speed per unit time and

represents the change of a gait-related parameter per unit time. change in

if there is no change in

unit change of

and

is the amount of

indicates the average increase in

per

. Therefore, the -values inform on how strongly changes in particular

gait parameters are related to changes in walking speed (Δ speed). Pairwise relations to preferred walking speed at each time point were performed using Spearman’s nonparametric correlation coefficient rho and data points were approximated for visualization by a linear or a 2nd degree polynomial fit, as appropriate, in a least squares sense.

Principal component analysis Principal component analysis (PCA) can be used to reduce the complexity of a multivariate set of data by generating new variables (principal components, PCs) that are constructed from the original variables under the constraint of explaining the highest variance in data while incremental PC numbers explain decreasing amounts of data variance. In other words, PC1 accounts for the highest amount of variance in data, therefore maximally separating distinct behaviors while the orthogonal PC2 explains the second largest proportion of variance, and so forth. Using this method, specific data clusters may emerge that are typically determined by a few PCs representing distinct groups of variables that mutually interact to explain a common behavior of outcome. In the present study PCA was applied on the correlation matrix of the above-mentioned 15 gait-related variables derived from the first assessment time point (early) and the last assessment time point (late) as well as healthy control data. A scree plot and overdetermination of PCs (at least two variables with factor loadings ≥ 0.75) were used to

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 9 of 32

9 estimate the number of PCs retained to explain the data. A cutoff of 0.75 was chosen because loadings express correlations between PC scores (new variables in PC domain) and original variables. Correlations ≥ 0.75 are generally considered to be strong correlations.16 The extraction of variables determining the PCs may change over time. This should be taken into account when evaluating data of different time points. To reveal the consistency of the obtained loading pattern across time we calculated Pearson’s r between loadings for each variable at early and late time points. In order to compare whether different groups of patients are distinguishable within the emerging PC domain using various grouping variables (i.e., initial walking speed, neurological condition, time point, cause of SCI), a one-way ANOVA was applied and, if appropriate, a post-hoc Tukey's test was performed for group-by-group comparisons.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 10 of 32

10 Results Subjects In total 16 patients (7 females, 9 males; age 51.2 ± 15.7 years; height 169.9 ± 8.6 cm; weight 67.1 ± 12.6 kg) met the inclusion and exclusion criteria and were evaluated for this study (Table 1). The kinematic data of 11 healthy control subjects (7 females, 4 males; age 39.2 ± 15.3 years; height: 174.5 ± 8.4 cm; weight: 70.1 ± 15.6 kg) with no neurological disorders and no walking impairments were used to compute the standard hip-knee cyclogram for the SSD calculation and to provide kinematic reference data. 15 out of 16 patients required an assistive device for safe walking at the early time point. Because changing the assistive device might confound the results, we let patients walk with the same assistive device at the late time point. However, 6 of the 16 patients could actually walk without any assistive device at the late time point.

Walking speed Only one patient could not improve walking speed (-0.34 km/h) while all other patients showed increased walking speeds over time. The mean increase was 96% (1.12 ± 1.08 km/h, mean ± SD; paired t-test: p < 0.001). The initial walking speed of patients was not related to their change in speed (delta speed; r = 0.174, p = 0.520; Figure 1A).

Responsiveness of gait parameters More than half of the parameters significantly improved from an early (74 ± 59 days post-incident; mean ± SD) to a late time point (216 ± 116 days, Figure 1B). However, the knee and ankle ROMs as well as the foot clearance and the cyclogram shape difference (SSD) did not change. On average, the cyclogram shape difference did not change over

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 11 of 32

11 time (no improvement of intralimb coordination) from the first to the last time point of assessment (paired t-test: p = 0.087) while the cyclogram reproducibility (ACC) increased during recovery (Wilcoxon signed rank test: p = 0.006). Examining the influence of changes in gait variables on changes in preferred walking speed using linear regression analysis showed that changes in walking speed were most closely accompanied by changes in hip ROM (β = 1.12), followed by cadence (β = 1.10), hip-knee angular velocity at toe-off (β = 1.07), and step length (β = 1.00, Figure 1C). Changes in the quality of intralimb coordination, quantified by the shape difference of the hip-knee cyclogram compared to normal (β = 0.35), the foot clearance (β = 0.22) and the cycle-to-cycle cyclogram consistency (β = 0.09) were least related to an increase in preferred walking speed.

Multivariate analysis Principal component analysis applied on all variables at both time points revealed that PC1 alone explained 58.6% of the total variance in data while the first two components collectively explained 73.7% of variance. The scree plot and PC over-determination suggested to retain the first two PCs. PC1 was highly loaded by gait-cycle parameters, preferred walking speed, angular velocities at toe-off, cyclogram shape difference, and cyclogram consistency (Figure 2A). PC2 only received high loadings from the knee- and ankle ROM. After segregating data into early and late time point during recovery in comparison to healthy control data, ANOVA and post-hoc Tukey's revealed that the data only showed group differences along PC1, suggesting that the variables with high loadings on PC1 explained a large part of gait recovery (approaching normal) across time (Figure 2B). When initial speed was selected as the grouping variable, the ANOVA

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 12 of 32

12 and post-hoc Tukey’s analyses yielded differences between initially slow walkers (≤ 1km/h) and both medium (≤ 2 km/h) and fast walkers (≤ 3km/h) but not between medium and fast walkers (Figure 2C). Choosing the cause of the SCI as the grouping variable did not yield any differences among groups. A matching of the loading factors on PC1 and PC2 between the early and late time point using Pearson’s correlation revealed a robust outcome over time expressed by the correlation coefficient r = 0.930 for PC1 variable loadings and r = 0.920 for PC2 variable loadings.

Speed dependence The speed relation of gait parameters remained relatively consistent across time except for the knee- and ankle ROM: The SSD and ACC showed a non-linear behavior with respect to preferred speed at early and late time point with far-from-normal values in slow walkers while approaching normal and plateauing in fast walkers (Figures 3A, C). The speed relation of step length (early: rho = 0.838, late: rho = 0.943, Figure 3E) and hip ROM (early: rho = 0.406, late: rho = 0.580, Figure 3B) both showed a similar and adequate relation to walking speed at an early compared to a late time point, even though the relation was weaker in the hip ROM. Step length is shown as one representative of gait-cycle parameters, which all showed similar behavior. The speed relation of the more distal joint ROMs was very weak at an early time point (knee: rho = 0.015, ankle: rho = -0.212) but improved at a late time point (knee: rho = 0.596, ankle: rho = 0.309) while absolute values still remained rather different from normal (Figures 3D, F).

Recovery of intralimb coordination

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 13 of 32

13 In contrast to speed and time-distance parameters, which almost consistently improved over time, the intralimb coordination as a readout for gait quality showed a rather discrete behavior. With regard to cyclogram impairment patients behaved quite distinctively and could be allocated to three patterns of recovery, irrespective of speed or cadence improvements: 1) bilateral recovery of intralimb coordination with improved shape and consistency of cyclogram in both limbs over time, 2) bilaterally no improvement in intralimb coordination despite higher walking speed and cadence, and 3) dissociated unilateral cyclogram recovery, with an improved cyclogram in one leg and an invariant or deteriorating cyclogram in the other leg (Figure 4).

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 14 of 32

14 Discussion Recovery of locomotor function after iSCI underlies a complex process where many factors contribute to a gain in walking function. Measures of responsiveness and regression analyses are limited at revealing the intricate contribution of variables to the multidimensional behavior of gait recovery. The combined application of regression and multivariate analysis (PCA) disentangled parameters that rather react to improved speed in contrast to measures that, although not responsive over time, have a significant impact on recovery. The conjunct use of the two analysis methods revealed that, in the absence of improvements in gait quality despite significant increases in walking speed over time, functional recovery is mainly attributable to the acquisition of compensatory strategies and increased efficiency of a pathological gait rather than restitution (i.e., normalization) of motor control of walking.11 The distinction between these underlying mechanisms of gait recovery might offer new insights for the refined evaluation of treatment interventions.

Responsiveness and speed dependence Except for one subject all patients increased their preferred walking speed. Interestingly, the extent of improvement in walking speed was not related to initial speed while the final speed outcome was dependent on baseline values as the gain in speed was relatively constant irrespective of initial speed. In other words, initially slow walkers also had a relatively slow walking speed at final assessment. This suggests that recovery of walking speed following iSCI to some extent relies on an inherent capacity of functional improvement irrespective of initial impairment. Measuring relative changes of outcome variables over time indeed reflects their responsiveness as a measurement tool but does

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 15 of 32

15 not scale their functional relevance. For that purpose, variables are compared to a gold standard (as a proxy for recovery), in this case walking speed. However, major concerns here are that the selection of a proxy produces a bias. Moreover, the responsiveness to increased walking speed may partially be attributable to recovery, but is also largely influenced by biomechanical modulations that accompany increased walking speed. 19

18,

The outcome of the regression analysis can be interpreted such that greater speed

was mainly achieved by larger maximal deflections of the hip joint rather than the more distal joint angles. The speed relation of the hip ROM remained relatively consistent across time (i.e., similar correlation at the early and late time point) but significantly changed in the knee- and ankle ROM. These findings confirm that neural control across lower-limb joints is not uniform20 while the hip seems to have a specific role in gait control.21 The distinct behavior of distal lower-limb joints was also reflected by the outcome of the PCA where knee- and ankle ROM loaded onto PC2 (not closely related to walking recovery) and may reflect their distinct roles in motor control and recovery. In addition, distal joint angles may be more strongly impeded by confounders such as increases in muscle tone (eventually presenting as spasticity) and changes in mechanical properties of the muscle-tendon compartments (i.e., joint contractures).22 The use of compensatory strategies with the aim of a better performance (i.e., a stiff leg may allow for faster and more reliable walking) may further contribute to abnormal walking patterns. Gait-cycle parameters remained in accordance with walking speed when compared to healthy subjects both at the early and late time point and therefore reflect increased walking speed rather than underlying changes in motor control. The intralimb pattern (consistency and shape), however, was less responsive to regained walking speeds. Obviously, a normalization of the intralimb coordination relies on more

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 16 of 32

16 complex control mechanisms and on the unimpaired function of involved neural structures that may not have recovered to pre-injury levels to provide normal behavior. It therefore appears that spinal networks can be trained to some extent to enable increases in walking speed while the persistently diminished supraspinal drive prevents an improvement of lower limb coordination, suggesting a critical dependence on supraspinal control. iSCI gait patterns are probably immediately triggered by the loss of descending fibers and subsequently promoted by the reorganization of spared neuronal circuits following deprivation of supraspinal inputs.23, 24, 26, 28

Determinants of recovery The present findings highlight the crucial importance of distinguishing between parameters showing changes by simply responding to increased walking speed (i.e., responsiveness) and those measures that actually contribute to recovery of walking function while not necessarily being responsive. PC1, by distinguishing gait outcome across time, consists of variables that strongly contribute to overall recovery (i.e., the different gait parameters followed over time form an unbiased construct of recovery) in a mutually interrelated way, which does not necessarily require paralleled improvements over time. Given that gait-cycle parameters increased over time and also loaded highly onto PC1, one might conjecture a strong contribution to impairment and recovery. Upon closer examination, however, these variables revealed intact modulation with respect to speed already at an early time point, suggesting that they may be little affected by the injury and simply increase over time due to speed increments. In contrast, variables related to the intralimb coordination (SSD, ACC) were impaired in the iSCI population

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 17 of 32

17 and showed little responsiveness to improvements in speed. A previous study showed that the cyclogram shape difference (SSD) of iSCI patients does not normalize as a consequence of increasing speed from slow to preferred walking while it converged to a very uniform shape in control subjects.13 Taken together, these findings suggest that the complex multi-segmental coordination is less susceptible to both immediate changes in speed as well as increases in speed in the course of recovery. Despite diminished responsiveness, the degree of alteration of intralimb coordination determines gait recovery, i.e., deficits in multi-segmental lower limb coordination influence overall gait recovery. Hip ROM, on the other hand, showed the highest responsiveness to speed improvements but did not considerably contribute to recovery as revealed by PCA. Thus, whether parameters are well modulated in patients or impaired due to injury while being unrelated to speed and invariant across time may intricately affect recovery and reveal specific information on dynamic changes in distinct underlying locomotor control systems.

Patterns of recovery As discussed, the initial speed was of no particular predictive value to estimate speed improvements. However, initially slow walkers (≤ 1km/h) had a different recovery profile compared to moderate or fast walkers (≤ 2 or ≤ 3km/h, respectively) as revealed by PCA. The lower limb patterns of iSCI subjects observed over time revealed 3 distinct recovery profiles. These profiles differed in their left/right symmetry of recovery while the intralimb pattern may or may not show any convergence towards normal. The recovery profiles of intralimb patterns were not consistently paralleled by either progression of speed or cadence that both increased on average in iSCI subjects despite absent shape

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 18 of 32

18 normalization. The distinct behavior of within limb coordination is complementary to the relatively uniform changes taking place in commonly assessed time-distance measures. This makes it a valuable readout to reflect underlying gait control mechanisms and recovery processes not reflected by any of the other outcome metrics and highlights the importance of evaluating complementary variables in order to disentangle differential contributions to impairment and recovery.

Limitations Major drawbacks of the present study were the irregular visit intervals and the varying number of visits. We tried to overcome this problem by only looking at the first and the last time point of assessment and evaluating the rate of change of the various parameters (Δ value/Δ time). Also, we could only include patients who were able to walk independently and already regained some locomotor capacity. We therefore might have missed a very sensitive phase during which patients transitioned into being able to start walking and major effects of plasticity within the sensory-motor system may have occurred. Using assistive devices may affect the biomechanics of walking. Therefore, the absolute values may be affected while we tried to minimize the effect attributable to the walking aid by using the same device throughout all recording sessions.

Conclusions The combination of unbiased multivariate analysis and regression analysis followed by detailed post-hoc analyses of complementary outcome measures including gait-cycle parameters and intralimb coordination revealed that some measures remain well controlled in iSCI even at an early stage of recovery. In addition, improvements in

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 19 of 32

19 measures as a consequence of speed increases can be distinguished from contributing factors that may determine walking outcome. The distinction of these parameters may help to identify underlying processes of compensatory mechanisms and recovery of motor control and is crucial for the conduct and evaluation of future interventional studies.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 20 of 32

20 Acknowledgements Funding This study was supported by the European Commission’s Seventh Framework Program [CP-IP 258654; NEUWalk]; the Clinical Research Priority Program CRPP Neurorehab UZH; the Wolf Foundation, Switzerland. We would like to thank B. Huber and M. Stüssi for their assistance with subject recruitment and recordings.

Author disclosure statement No competing financial interests exist.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 21 of 32

21 References 1.

Kidd D, Stewart G, Baldry J, Johnson J, Rossiter D, Petruckevitch A and Thompson AJ. (1995). The Functional Independence Measure: a comparative validity and reliability study. Disabil. Rehabil. 17, 10-14.

2.

Gladstone DJ, Danells CJ and Black SE. (2002). The fugl-meyer assessment of motor recovery after stroke: a critical review of its measurement properties. Neurorehabil. Neural Repair 16, 232-240.

3.

van Hedel HJ and Dietz V. (2009). Walking during daily life can be validly and responsively assessed in subjects with a spinal cord injury. Neurorehabil. Neural Repair 23, 117-124.

4.

Severinsen K, Jakobsen JK, Pedersen AR, Overgaard K and Andersen H. (2014). Effects of resistance training and aerobic training on ambulation in chronic stroke. Am. J. Phys. Med. Rehabil. 93, 29-42.

5.

Thompson AK, Pomerantz FR and Wolpaw JR. (2013). Operant conditioning of a spinal reflex can improve locomotion after spinal cord injury in humans. J. Neurosci. 33, 2365-2375.

6.

Kim CM, Eng JJ and Whittaker MW. (2004). Level walking and ambulatory capacity in persons with incomplete spinal cord injury: relationship with muscle strength. Spinal Cord 42, 156-162.

7.

Petersen JA, Spiess M, Curt A, Dietz V and Schubert M. (2012). Spinal cord injury: one-year evolution of motor-evoked potentials and recovery of leg motor function in 255 patients. Neurorehabil. Neural Repair 26, 939-948.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 22 of 32

22 8.

Buurke JH, Nene AV, Kwakkel G, Erren-Wolters V, Ijzerman MJ and Hermens HJ. (2008). Recovery of gait after stroke: what changes? Neurorehabil. Neural Repair 22, 676-683.

9.

Raineteau O and Schwab ME. (2001). Plasticity of motor systems after incomplete spinal cord injury. Nat. Rev. Neurosci. 2, 263-273.

10.

van den Brand R, Heutschi J, Barraud Q, DiGiovanna J, Bartholdi K, Huerlimann M, Friedli L, Vollenweider I, Moraud EM, Duis S, Dominici N, Micera S, Musienko P and Courtine G. (2012). Restoring voluntary control of locomotion after paralyzing spinal cord injury. Science 336, 1182-1185.

11.

Curt A, Van Hedel HJ, Klaus D and Dietz V. (2008). Recovery from a spinal cord injury: significance of compensation, neural plasticity, and repair. J. Neurotrauma 25, 677-685.

12.

Knikou M and Mummidisetty CK. (2014). Locomotor training improves premotoneuronal control after chronic spinal cord injury. J. Neurophysiol. 111, 2264-2275.

13.

Awai L and Curt A. (2014). Intralimb coordination as a sensitive indicator of motor-control impairment after spinal cord injury. Front. Hum. Neurosci. 8, 148.

14.

Dietz V and Curt A. (2006). Neurological aspects of spinal-cord repair: promises and challenges. Lancet Neurol. 5, 688-694.

15.

Field-Fote EC and Tepavac D. (2002). Improved intralimb coordination in people with incomplete spinal cord injury following training with body weight support and electrical stimulation. Phys. Ther. 82, 707-715.

16.

Dawson-Saunders B and Trapp RG. (2004). Basic and Clinical Biostatistics. 4th ed. New York: Lange Medical Books/McGraw-Hill.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 23 of 32

23 17.

Husted JA, Cook RJ, Farewell VT and Gladman DD. (2000). Methods for assessing responsiveness: a critical review and recommendations. J. Clin. Epidemiol. 53, 459-468.

18.

Hanlon M and Anderson R. (2006). Prediction methods to account for the effect of gait speed on lower limb angular kinematics. Gait Posture 24, 280-287.

19.

Pepin A, Norman KE and Barbeau H. (2003). Treadmill walking in incomplete spinal-cord-injured subjects: 1. Adaptation to changes in speed. Spinal Cord 41, 257-270.

20.

Hicheur H, Terekhov AV and Berthoz A. (2006). Intersegmental coordination during human locomotion: does planar covariation of elevation angles reflect central constraints? J. Neurophysiol. 96, 1406-1419.

21.

Dietz V, Muller R and Colombo G. (2002). Locomotor activity in spinal man: significance of afferent input from joint and load receptors. Brain 125, 2626-2634.

22.

Krawetz P and Nance P. (1996). Gait analysis of spinal cord injured subjects: effects of injury level and spasticity. Arch. Phys. Med. Rehabil. 77, 635-638.

23.

Hubli M, Bolliger M and Dietz V. (2011). Neuronal dysfunction in chronic spinal cord injury. Spinal Cord 49, 582-587.

24.

Beauparlant J, van den Brand R, Barraud Q, Friedli L, Musienko P, Dietz V and Courtine G. (2013). Undirected compensatory plasticity contributes to neuronal dysfunction after severe spinal cord injury. Brain 136, 3347-3361.

26.

Dietz V. (2010). Behavior of spinal neurons deprived of supraspinal input. Nat. Rev. Neurol. 6, 167-174.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 24 of 32

24

28. Calancie B, Lutton S and Broton JG. (1996). Central nervous system plasticity

after spinal cord injury in man: interlimb reflexes and the influence of cutaneous

stimulation. Electroencephalogr. Clin. Neurophysiol. 101, 304-315.

Figure legend

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 25 of 32

25

Figure 1. Responsiveness and speed dependence. (A) The initial speed of patients does not predict their gain in walking speed as revealed by the low Pearson correlation coefficient r. (B) Most parameters improved from an early to a late time point. Asterisks depict those variables that had a p-value < 0.01 (bold), the hashtag marks those

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 26 of 32

26 parameters with p-values < 0.05 (italic) when comparing the variables between the two time points. (C) The stem plot illustrates rates of change of gait related variables (Δ values divided by Δ time; Δ = late - early). Green circles represent positive changes and red circles negative changes. Black dot = mean. ACC = angular component of coefficient of correspondence, AngVel = angular velocity at toe-off, EndpointVel = velocity of the 2nd metatarsal joint marker at toe-off in the sagittal plane, LEMS = lower extremity motor score, ROM = range of motion, SSD = square root of sum of squared distances.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 27 of 32

27

Figure 2. Principal component analysis applied on 15 gait-related parameters revealed that the first two PC’s alone explained almost 75% of the variance in data (A). Differences between early and late gait outcome emerged along the PC1 axis but not the PC2 axis and patient data was significantly distinguishable from control data at any time point along PC1 (B). A segregation of the data according to the initial walking speed revealed that slow walkers (≤ 1km/h) differed from both medium (≤ 2km/h) and fast walkers (≤ 3km/h) with respect to PC1 but not PC2 (C). Asterisks depict p-values < 0.05.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 28 of 32

28

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof. Page 29 of 32

29

Figure 3. Speed dependence. The relation of various gait-related parameters to

preferred speed was quantified by Spearman’s correlation coefficient rho at an early

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 30 of 32

30 (red) and a late (green) time point. The shaded gray area indicates the ± 2SD interval of healthy control subjects. SSD and ACC data (A and C) were fitted by a 2nd degree polynomial while the other data contain a linear fit (B, D, E, F). ACC = angular component of coefficient of correspondence, ROM = range of motion, SSD = square root of sum of squared distances.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 31 of 32

31

Figure 4. Patterns of intralimb recovery. Gait analysis at 3 time points of rehabilitation (n = 13) revealed 3 distinct patterns of intralimb recovery. Recovery and patterns of no improvement were revealed bilaterally while some patients showed a dissociated, unilateral pattern of recovery. The walking speed improved in all 3 groups and was paralleled by the cadence. ACC = angular component of coefficient of correspondence, SSD = square root of sum of squared distances.

Journal of Neurotrauma Locomotor recovery in spinal cord injury - insights beyond walking speed and distance (doi: 10.1089/neu.2015.4154) This article has been peer-reviewed and accepted for publication, but has yet to undergo copyediting and proof correction. The final published version may differ from this proof.

Page 32 of 32

32 Table 1. Descriptive data of patients Initial ID

Age [years] Sex

Cause of SCI

Level of SCI

Initial LEMS

speed km/h

01

73

m

Traumatic

C5

48

0.19

03

36

f

Disc prolapse

T7/8

32

1.12

04

47

f

Ischemic

T9/10

45

2.54

05

30

m

Traumatic

L2

48

2.88

08

60

f

Ischemic

T5

27

1.17

09

48

f

Traumatic

T7

30

0.86

10

39

f

Spinal myelitis

C, T

38

1.15

13

23

m

Traumatic

C7

19

0.13

30

36

m

Traumatic

T4

40

1.97

31

60

m

Traumatic

C3

48

1.50

32

47

m

Traumatic

C5

46

0.79

34

63

m

Tumor

T4

46

0.41

37

71

m

Ischemic

T11

43

0.11

39

76

f

Tumor

T1

37

0.20

40

58

m

Ischemic

T4

45

2.37

44

53

f

Traumatic

C4

48

1.27

SCI = spinal cord injury; C = cervical, T = thoracic, L = lumbar; LEMS = lower extremity motor score (max = 50).

Locomotor Recovery in Spinal Cord Injury: Insights Beyond Walking Speed and Distance.

Recovery of locomotor function after incomplete spinal cord injury (iSCI) is clinically assessed through walking speed and distance, while improvement...
1MB Sizes 0 Downloads 11 Views