The Evolution of Hand Proprioceptive and Motor Impairments in the Sub-Acute Phase After Stroke

Background Hand proprioception is essential for fine movements and therefore many activities of daily living. Although frequently impaired after stroke, it is unclear how hand proprioception evolves in the sub-acute phase and whether it follows a similar pattern of changes as motor impairments. Objective This work investigates whether there is a corresponding pattern of changes over time in hand proprioception and motor function as comprehensively quantified by a combination of robotic, clinical, and neurophysiological assessments. Methods Finger proprioception (position sense) and motor function (force, velocity, range of motion) were evaluated using robotic assessments at baseline (<3 months after stroke) and up to 4 weeks later (discharge). Clinical assessments (among others, Box & Block Test [BBT]) as well as Somatosensory/Motor Evoked Potentials (SSEP/MEP) were additionally performed. Results Complete datasets from 45 participants post-stroke were obtained. For 42% of all study participants proprioception and motor function had a dissociated pattern of changes (only 1 function considerably improved). This dissociation was either due to the absence of a measurable impairment in 1 modality at baseline, or due to a severe lesion of central somatosensory or motor tracts (absent SSEP/MEP). Better baseline BBT correlated with proprioceptive gains, while proprioceptive impairment at baseline did not correlate with change in BBT. Conclusions Proprioception and motor function frequently followed a dissociated pattern of changes in sub-acute stroke. This highlights the importance of monitoring both functions, which could help to further personalize therapies.

from the middle of the device's workspace), ensuring a comfortable resting position for the wrist.The index finger is attached to an adjustable finger module by Velcro straps.The centre of rotation of the end-effector is aligned with the MCP joint.(c) A tablet computer is placed above the hand, removing visual cues from the tested hand and providing an interactive graphical user interface displaying a simple gauge with a red indicator.There is a large variability in individual changes.For each metric there are 10, respectively 15, participants that considerably improved (black), but for the majority of the participants changes were too small to be classified as considerable (grey).Higher result indicates better performance.Table SM2: Comparison of clinical and robotic measures of proprioception in cases when the considerable improvement groups did not match.In 8 out of 9 cases when improvement in proprioception was detected by the robotic measure, it was in the ceiling of kUDT (kinaesthetic Up-Down).The reason cases when the change according to clinical measure was not detected by robotic were variable, including the change being just below the threshold of considerable improvement (e.g., participant numebr 10 and 12), or it could have also been linked to the subjectivity of the clinical scale.AE: Absolute Error.       Figure SM14: Classification agreement between clinical and robotic measures (60% agreement between proprioception measures, 58% agreement between motor measures).Change here means considerable improvement, defined for robotic measures as change larger than the smallest real difference (SRD) or change from impaired to non-impaired.For kUDT considerable change was selected as change by 1 point, and for FMA it was equal to the Minimal Detectable Change (MDC), 5.2 points or improved above 60 points.The characteristics of the group that changed in robotic task but not in clinical in proprioception are that these subjects were in the ceiling of the clinical scale.In motor tasks it was possible to detect subtle changes for individuals severely affected using robotics.
Changes not detected by robotics but captured in FMA can be explained by the fact that this scale considers the whole upper limb, not only the hand.

Figure SM1 :
Figure SM1: Schematic of the ETH MIKE robotic platform used for the assessment of index finger metacarpophelangeal joint.(a) Subjects are seated in front of the ETH MIKE robot, with their elbow supported on an arm rest.A wrist splint is worn to avoid any compensatory movements at the wrist.The device is inclined by 20 • to minimize parallax errors.(b) The hand is wrapped around a handle, which is set up at the wrist neutral position (0 • wrist flexion, 30 •

Figure SM2 :
Figure SM2: Changes over time of motor function, measured by two of the robotic task metrics (Active Range of Motion (AROM) and Extension Velocity (EV), two of 3 motor metrics).In black are marked individuals that improved considerably = change larger than the smallest real difference (SRD) or change from impaired to non-impaired.The dashed lines mark impairment thresholds (based on mean + 2SD of neurologically intact age-matched controls).

Figure SM3 :
Figure SM3: Results of the linear mixed effect model -change in proprioception as a dependent variable, change in the subcomponents of motor function, as measured by the robotic assessments, as fixed effects (N=45).

Figure SM4 :
Figure SM4: Results of the linear mixed effect model -delta proprioception (DProp) as the dependent variable, subgroup analysis (N=28).

Figure SM5 :
Figure SM5: Results of the linear mixed effect model -change in proprioception, measured by the robotic assessment, as the dependent variable, demographic and stroke-related factors as fixed effects.Abbreviations: kUDTkinesthetic Up-Down Test, FMA -Fugl-Meyer Assessments, TSS -Time Since Stroke (days), MOCA -Montreal Cognitive Assessment.

Figure SM6 :
Figure SM6: Results of the linear mixed effect model -change in the Maximum Fingertip Force as the dependent variable, demographic and stroke-related factors as fixed effects.Abbreviations: kUDT -kinesthetic Up-Down Test, FMA -Fugl-Meyer Assessments, TSS -Time Since Stroke (days), MOCA -Montreal Cognitive Assessment.

Figure SM7 :
Figure SM7: Results of the linear mixed effect model -change in Active Range of Motion (AROM) as the dependent variable, demographic and stroke-related factors as fixed effects.Abbreviations: kUDT -kinesthetic Up-Down Test, FMA -Fugl-Meyer Assessments, TSS -Time Since Stroke (days), MOCA -Montreal Cognitive Assessment.

Figure SM8 :
Figure SM8: Results of the linear mixed effect model -change in the Maximum Velocity Extension as the dependent variable, demographic and stroke-related factors as fixed effects.Abbreviations: kUDT -kinesthetic Up-Down Test, FMA -Fugl-Meyer Assessments, TSS -Time Since Stroke (days), MOCA -Montreal Cognitive Assessment.

Figure
Figure SM9: (A)The difference in baseline motor function in the group that did and did not considerably improve in motor function is significant for all motor metrics (FF, AROM, EV).The group "no change" means no change in neither proprioception nor motor function and only considering individuals impaired at baseline.(B) The difference in baseline proprioception (measured by AE) is not significant between the groups that did and did not improve in proprioception.

Figure SM11 :
Figure SM11: The difference between the group that did not change and the group that changed in at least one function (motor or proprioception) in terms of demographic and stroke-related factors -time since stroke, age at baseline, gender, type of stroke (haemorrhagic or ischemic), lateralization (right or left hemispheric stroke) and cognitive function (measured by Montreal Cognitive Assessment).Individuals that were not impaired in neither proprioception nor motor function at baseline (N=3) were not included in this analysis.

Figure SM12 :
Figure SM12: The relationship between hand impairments and functional hand use when all values in the floor effect of the Box & Block Test (BBT) were removed.The results remained the same.Both motor function (here Force Flexion, A) and proprioception (Absolute Error, B) correlated with the skilled hand use at discharge, measured by Box & Block Test (BBT).C) Impaired proprioception at inclusion did not correlate with functional recovery, as measured by delta BBT.D) At least partially preserved hand function was needed for improvement in proprioception, especially considering individuals with improvement capacity (i.e., impaired at T1).

Figure SM13 :
Figure SM13: The correlation between motor hand impairments and functional hand use at discharge was strong for both Active Range of Motion (AROM) and Extension Velocity (EV).

Figure SM15 :Figure SM16 :
Figure SM15: Grouping of participants based on MEP scores to compare the change in Box & Block Test (BBT).Participants with normal MEP response showed biggest improvement in BBT.

Table SM1 :
The threshold values of robotic and clinical measures used to determine if a participant improved considerably according to a given measure.The values are based on our previous study validating the robotic metrics and for clinical measures on literature (references provided in brackets, full citations are given in the main manuscript).Minimal Detectable Change (for clinical measures), Impairment Threshold: defined as age-matched control mean + 2x standard deviation for robotic and taken from literature for clinical measures.