Professional Documents
Culture Documents
research-article2019
CRE0010.1177/0269215519829795Clinical RehabilitationTurville et al.
CLINICAL
Original article REHABILITATION
Clinical Rehabilitation
Abstract
Objective: The aim of this study was to evaluate if somatosensory retraining programmes assist people
to improve somatosensory discrimination skills and arm functioning after stroke.
Data sources: Nine databases were systematically searched: Medline, Cumulative Index to Nursing and
Allied Health Literature, PsychInfo, Embase, Amed, Web of Science, Physiotherapy Evidence Database,
OT seeker, and Cochrane Library.
Review methods: Studies were included for review if they involved (1) adult participants who had
somatosensory impairment in the arm after stroke, (2) a programme targeted at retraining somatosensation,
(3) a primary measure of somatosensory discrimination skills in the arm, and (4) an intervention study
design (e.g. randomized or non-randomized control designs).
Results: A total of 6779 articles were screened. Five group trials and five single case experimental designs
were included (N = 199 stroke survivors). Six studies focused exclusively on retraining somatosensation
and four studies focused on somatosensation and motor retraining. Standardized somatosensory
measures were typically used for tactile, proprioception, and haptic object recognition modalities. Sensory
intervention effect sizes ranged from 0.3 to 2.2, with an average effect size of 0.85 across somatosensory
modalities. A majority of effect sizes for proprioception and tactile somatosensory domains were greater
than 0.5, and all but one of the intervention effect sizes were larger than the control effect sizes, at least
as point estimates. Six studies measured motor and/or functional arm outcomes (n = 89 participants), with
narrative analysis suggesting a trend towards improvement in arm use after somatosensory retraining.
Conclusion: Somatosensory retraining may assist people to regain somatosensory discrimination skills
in the arm after stroke.
1Occupational Therapy, School of Allied Health, Human 5Department of Physiotherapy, Austin Health, Melbourne,
Services and Sport, College of Science, Health and VIC, Australia
Engineering, La Trobe University, Melbourne, VIC, Australia
2Neurorehabilitation and Recovery, Stroke Division, Florey Corresponding author:
Institute of Neuroscience and Mental Health, Melbourne, Leeanne M Carey, Occupational Therapy, School of Allied
VIC, Australia Health, Human Services and Sport, College of Science, Health
3Department of Occupational Therapy, School of Allied Health, and Engineering, La Trobe University, Bundoora, Melbourne,
Australian Catholic University, Melbourne, VIC, Australia VIC 3086, Australia.
4School of Psychology and Public Health, College of Science, Email: L.Carey@latrobe.edu.au
Health, and Engineering, La Trobe University, Melbourne,
VIC, Australia
2 Clinical Rehabilitation 00(0)
Keywords
Stroke, somatosensation, treatment, intervention, effect size
Database of Research into Stroke (DORIS). The fol- Data extraction and quality assessment
lowing websites were also searched: http://www
.opengrey.eu; http://www.controlled-trials.com; and Data were extracted according to established crite-
http://www.ClinicalTrials.gov. ria of key evaluation (e.g. blinding, independent
The search strategy was broad to address the assessment) and content (e.g. dose, theory) param-
diverse terminology in this field and involved col- eters. The primary author (MLT) extracted data and
laborative work of two authors (MLT and LMC). a second author (LSC) verified accuracy and com-
MeSH terms and free text terms were used, along pleteness of extracted data. In the case of missing
with truncation and wild card symbols to retrieve data, authors were contacted in an attempt to obtain
all relevant keywords. Search terms are listed in and include such data. The quality of studies was
Supplemental Appendix 1. To complete word cita- evaluated using the following tools: (1)
tion tracking, key authors and references were Physiotherapy Evidence Database (PEDro-P) scale
entered in Science Citation Index to identify fur- for randomized and non-randomized group
ther studies. The reference lists of relevant papers designs,19 or (2) Risk of Bias in N-of-1 Trials
and reviews were hand searched to ensure all appli- (RoBiNT) scale for single case experimental
cable research was obtained. designs.20 Two authors (MLT and LSC) completed
quality assessment independently, and a third
reviewer (LMC) was consulted if discrepancies
Study selection arose.
Studies were included if they involved (1) a sample
of adults with diagnosis of stroke and measured
Data synthesis and analysis
somatosensory impairment in the arm; (2) an inter-
vention for retraining somatosensory discrimina- Effect sizes were computed from single groups of
tion skills in the arm; (3) a primary outcome related participants (i.e. individual studies) undertaking
to a somatosensory discrimination modality; and either intervention and/or control conditions. To
(4) a study design such as randomized control trial, understand somatosensory changes, effect sizes
controlled clinical trial, or single case experimental were calculated for each somatosensory modality
design. When present, outcomes of arm function outcome (i.e. postphase mean result minus pre-
were also extracted from included studies. phase mean result, divided by the pretest stand-
A study was excluded if the (1) intervention ard deviation). We used the pretest standard
focused on somatosensory stimulation only; (2) deviation to standardize the raw metric effect size
study did not measure somatosensation discrimina- because it provided a representative measure of
tion skills as a primary outcome; (3) intervention variability between individuals in the preinter-
used sensation as a modality to improve movement vention state. This index most closely resembles
in the arm, not sensation; (4) article was written in Glass’ Delta (Glass δ) and enabled computation
a language other than English or not full text; (5) of a standardized effect size for each intervention
publication was a descriptive case study/series, group.21 This was necessary given that not all
cohort study, review, book chapter, or editorial; or studies possessed control groups, and our aim
(6) sample did not have somatosensory impair- was to describe effect sizes from all groups pos-
ment. Two review authors (MLT and LSC) sessing sufficient data, not just those obtained
reviewed article titles and abstracts independently from parallel group trials.
to identify potentially relevant studies. Full text of We also reviewed the literature to obtain rele-
identified articles was then retrieved and two vant data that would provide a frame of reference
review authors (MLT and LSC) independently for interpreting intervention group results. For esti-
screened studies according to inclusion criteria. mates of effects produced by control conditions,
When differences occurred, a third reviewer (LMC) we used results from (1) the control groups of
was consulted for confirmation of criteria. somatosensory discrimination intervention studies
4 Clinical Rehabilitation 00(0)
with controlled clinical design; (2) other soma- The majority of studies did not provide suffi-
tosensory intervention studies that did not apply cient information to compute standard errors for
discrimination retraining, but did provide relevant mean changes, thus confidence intervals for stand-
control data; and (3) observational studies of arm ardized effect sizes, formal effect size pooling, and
somatosensory recovery after stroke. We adopted forest plots could not be constructed. Therefore, we
this approach given the lack of control data availa- conducted a descriptive meta-analysis based on
ble in many (N = 7) of the somatosensory discrimi- point estimates of mean changes.
nation studies. Finally, narrative analysis was used for addi-
We identified studies that (1) measured recov- tional outcomes related to arm functioning
ery of somatosensation post-stroke in a control because there was insufficient data from which
group or observational study; (2) assessed soma- effect sizes could be calculated. Descriptive sta-
tosensory discrimination in the arm and/or hand in tistics and narrative analysis were employed to
at least one sensory modality; (3) published mean characterize the population sample and interven-
and standard deviation data from which effect sizes tion characteristics.
could be calculated; (4) included a sample of at
least 10 stroke survivors; and (5) measured change
Results
over a period of weeks to months.
We found three studies to act as control proxies: A flowchart of search results is presented in Figure 1.
(1) the Connell et al.22 study that assessed soma- Ten studies were included in this review25–34 and are
tosensory recovery at 0–6 months after stroke in presented in Table 1.
two rehabilitation facilities (N = 80); (2) the Wolny
et al.23 intervention study on Butler’s neuromobili-
Study designs and quality assessment
sations, which included a control group (N = 32)
using ‘traditional’ post-stroke therapy (e.g. multi- The study designs and quality assessment are detailed
disciplinary treatment); and (3) Cambier et al.24 in Table 1. In general, the quality of study designs
randomized control trial study on intermittent was relatively low for group trials (i.e. less than
pneumatic compression, with a control group that 5/10), with the exception of one randomized control
received standard therapy plus sham short wave trial33 achieving 10/10 on the PEDro-P scale19 (Table
therapy to their affected arm (N = 12). 1). Single case experimental design studies (n = 5)
Where possible, we pooled data from single resulted in more consistent quality ratings that ranged
case experimental design intervention and con- from 13 to 18 (out of 30) on the RoBiNT scale20
trol phases to calculate respective effect sizes. To (Table 1). For detailed information on quality ratings,
achieve consistency with published group stud- see Supplemental Appendix 2.
ies, we calculated for each case the difference
between start and end of baselines to obtain con-
Participant characteristics
trol group mean changes. The standard deviation
used to obtain the quasi-Glass δ was estimated This review involved 199 stroke survivors (M = 20,
from a pooled sample of baseline phase pretests SD = 17.44) and, on average, participants were
(i.e. the first tests in the time series, which repre- 59.5 years old (SD = 10.23) and 1.8 years post
sented individual differences in preintervention stroke (range: 3 weeks–6.2 years). Male partici-
states of impairment). For intervention effects, pants were more common than females (65%
we calculated the change from end of baseline to males; 35% females), and the sample contained
end of intervention for each case. These differ- nearly equal numbers of right-sided (48%) and left-
ences were averaged and the mean change was sided strokes (52%). Typically, participants did not
then standardized using the standard deviation of have significant communication difficulties, previ-
the baseline phase pretest, as for the baseline ous neurological conditions, or perceptual difficul-
change effect. ties, such as unilateral spatial neglect.
Turville et al. 5
Authors Study Quality N Focus of Somatosensory measures Motor and/or arm functioning measures Other outcome
(Year) design rating retraining measures
Yekutiel and CT 3/10 39 Sensation Location of touch: 20 repetitions of a blunt pencil on No formal measures. Reported in discussion that two Not measured
Guttman hand with vision occluded. Client points to picture of clients reported they could now use affected hand
(1993)25 hand to identify location of touch more than before treatment.
Elbow proprioception: affected elbow moved to 10
flexion angles, and without vision, client matches
position with unaffected arm.
Two-point discrimination: client reported if felt 1 or 2
pins on fingers, palm, or forearm; 32 trials
Tactile object recognition: identify 20 common objects
using affected hand with vision occluded.
Byl et al. CT 4/10 18 Sensation Sensory discrimination: Fine motor control: Lower limb
(2003)28 and motor Graphesthesia: modified subtest of Sensory Integration Digital reaction time musculoskeletal
Praxis Test (SIPT) Perdue Pegboard test performance
Kinaesthesia: subtest of SIPT Musculoskeletal performance Gait: velocity
Stereognosis: Byl-Cheney-Boczai test of stereognosis Manual muscle test: strength California
Range of motion Functional
Wolf Motor Functional Test Evaluation
Byl et al. CT 4/10 45 Sensation Sensory discrimination: Fine motor control: Café 40: functional
(2008)30 and Motor Graphesthesia: modified subtest of Sensory Integration Digital reaction time performance
Praxis Test (SIPT) Finger tapping speed Gait: velocity
Kinesthesia: subtest of SIPT Strength
Stereognosis: Byl-Cheney-Boczai test of stereognosis Grip and pinch
Wolf Motor Functional Test
Carey et al. RCT 10/10 50 Sensation Texture discrimination: Fabric Matching Test (FMT) Sequential Occupational Dexterity Assessment Not measured
(2011)33 Proprioception: Wrist Position Sense Test (WPST) (SODA)
Tactile object recognition: functional Tactile Object Motor Activity Log (MAL)
Recognition Test (fTORT) Results from the SODA and MAL were not reported
in published study.
De Diego RCT 5/10 21 Sensation Tactile sensibility: Semmes-Weinstein monofilament. Arm motor function: Fugl Meyer Assessment (FMA): Stroke Impact
et al. and motor Proprioception: Three tests of sensory discrimination upper limb section Scale-16 (SIS-16):
(2013)34 with the person blindfolded: Arm use and motion quality: Motor Activity Log Autonomy and
Direction of passive motion (flexion extension – (MAL) Independence
neutral) in the joints of the upper limb (elbow, wrist,
metacarpophalangeal and thumb).
Consistency discrimination: ordering 4 sponges of
different consistency from less to more consistency.
Weight discrimination of objects (Scalha et al., 2011)
Clinical Rehabilitation 00(0)
Table 1. (Continued)
Single case experimental design studies
Turville et al.
Authors Study Quality N Focus of Somatosensory measures Motor and/or arm functioning measures Other outcome
(Year) design rating retraining measures
Carey et al. SCED 18/30 8 Sensation Tactile Discrimination Test No formal measures. Qualitative reporting of Not measured
(1993)26 Proprioceptive Discrimination Test subjective comments from participants.
Smania et al. SCED 17/30 4 Sensation Tactile discrimination (modified version of Carey 1993 Paper manipulation: six trials where clients had to Not measured
(2003)27 and motor measure) crumple paper into a ball.
Joint position sense (modified version of Carey 1993 Motor sequences: perform four sequences of alternative
measure) movement involving first digit with other digits.
Pressure sensation (modified version of Dannenbaum Reaching and grasping: two or three cylinders
and Dykes 1990) positioned on a wood board.
Weight discrimination: 30 trials where client had two Thumb-index grip force control: using a piston each
weights in hands and had to report if weights were person had to adjust force of thumb-index grip to
same or different. reach a certain force (500 or 750 g).
Letters tactile recognition: use tactile exploration to Participants completed 20 everyday activities and
recognize 12 capital letters. scoring was scoring based on categories of time taken
to complete tasks.
Participant’s relatives were asked to rate on a visual
analogue scale (VAS) how much participant used
affected hand in daily life.
Carey and SCED 18/30 10 Sensation Study 1: Not measured or reported Not measured
Matyas Texture Discrimination: Tactile Discrimination Test (TDT)
(2005)29 Texture Discrimination: Fabric Matching Test (FMT)
Proprioception: Wrist Position Sense Test (WPST) –
flexion and extension
Proprioception: Ulnar-radial deviation
Study 2:
Texture Discrimination: Fabric Matching Test
(FMT)
Texture Discrimination: Grid Matching Test
(GMT)
Carey and SCED 15/30 3 Sensation Participant 1: Texture discrimination using Grid Not measured or reported Not measured
Matyas Matching Test (GMT)
(2008)31 Participant 2: Proprioception using Wrist Position Sense
Test (WPST).
Participant 3: Texture discrimination using Tactile
Discrimination Test (TDT)
Helliwell SCED 13/30 1 Sensation The Rivermead Assessment of Somatosensory Motor Assessment Scale (MAS): upper limb section. Functional
(2009)32 Performance (RASP): ‘primary’ subtests. Sensory Independence
modalities were tested six times per body part. Measure (FIM)
n, sample size; CT, controlled trial, non-randomized; SCED, single case experimental design; RCT, randomized control trial.
7
8 Clinical Rehabilitation 00(0)
Figure 2. Somatosensory effect sizes: (a) proprioception, (b) touch, (c) object recognition, and (d) combined.
Black circles indicate intervention condition and grey circles indicate control condition. 1 = Yekutiel and Guttman,25
3 = Carey et al.,33 4 = De Diego et al.,34 5 = Byl et al.,30 6 and 7 = Carey et al.26 and Carey and Matyas29 (pooled single
case experimental design data), 8 = Connell et al.,22 9 = Cambier et al.,24 and 10 = Wolny et al.23 Effect size results
from Study 2. Byl et al.28 are not presented as we were unable to calculate these. Results are presented according to
different sensory modalities (i.e. tactile, proprioception); hence, this figure includes multiple representations of findings
from individual studies. These effect size results were calculated from studies with varying quality of study designs.
There was a relative lack of control data in the quality was also found to be relatively low and
included studies from which to compare interven- variable.7,13 Effect sizes must be interpreted with
tion results, and future randomized control consideration of study quality ratings (Table 1 and
trials are required to address this gap.7,13,16,17 Supplemental Appendix 2), which highlight meth-
Consequently, we sought additional somatosensory odological limitations.
control data from the literature to use in our analy-
ses. The calculation of control effect sizes and
Limitations
comparison with intervention effect sizes in this
way is a development to the current literature, par- This review is limited to the investigation of soma-
ticularly when randomized control trials are lim- tosensory retraining with a chronic stroke population.
ited. This method showed that even without sound Participants involved in somatosensory retraining
confidence interval estimates for the effects of studies were typically in the chronic phase of recov-
interventions versus control conditions, it is never- ery (M = 1.8 years post-stroke), were 60 years of age
theless the case that intervention somatosensory (SD = 10.23), and had reasonable communication and
effect sizes were larger than control group soma- perceptual abilities. Furthermore, the sample size dif-
tosensory effect sizes. fered considerably among studies (N = 1–50), and the
total number of participants included in this review
(N = 199) was relatively small.
Arm function Four of the 10 studies included motor retraining
Based on a descriptive summary of data (N = 89 in addition to somatosensory retraining,27,28,30,34
participants), the findings provided limited evi- and three of these studies also included additional
dence to suggest that somatosensory discrimina- therapeutic modalities, such as mirror therapy,
tion retraining may facilitate arm functioning, and guided mental imagery, and constraint induced
future research is needed. Prior systematic reviews practices for the unaffected hand.28,30,34 This has
also found inconclusive results regarding whether implications particularly in relation to the motor
somatosensory treatments change stroke survivors’ outcomes reported and confounds the ability to
arm functioning or general functional status.7,36 separate contributions from somatosensory dis-
Historically, performance and participation out- crimination training alone.
comes have not been widely investigated in soma- The control groups investigated in this review
tosensory treatment programmes,8 which may were heterogeneous in nature (e.g. usual care,
contribute to our present lack of knowledge. expose to somatosensory stimuli without retraining
principles, sham therapy), and this could have
impacted on the comparison of results. We acknowl-
Study quality edge that tactile effect sizes in three groups related
The quality of group study designs was relatively more to detection than discrimination skills,22,24,34
low, with the exception of one randomized control and, in general, somatosensory outcome measures
trial.33 This study did find a significant between- were heterogeneous and lacking normative stand-
group effect.33 The main methodological limita- ards. Given the timeframe of the Connell et al.22
tions of group studies related to a lack of study, results are likely to include the period in
randomization,25,30 concealed allocation,25,28,30 and which maximum spontaneous recovery occurred
blinding of participants and therapists.25,28,30,34 and therefore provided a conservative estimate for
Most group studies involved blinded asses- comparison of interventions that tended to involve
sors,28,30,33,34 although this was not the case for stroke survivors in the chronic stage of recovery.
some studies.26,27,29,31,32 Single case experimental Overall, an inferential meta-analysis could not be
design studies were of a consistent quality that conducted due to lack of information about standard
ranged from 13 to 18 (out of 30) on the RoBiNT errors of change scores and the distribution of
scale.20 In previous systematic reviews, study change scores. In addition, effect size results may be
Turville et al. 11
limited depending on the representativeness of is to enable people to have a more functional arm
standard deviation findings.21 A weakness of para- use after stroke.11 Intervention descriptions lacked
metric standardized effect size indexes, including examples of how retraining was tailored to indi-
the one employed in this study, is that they standard- vidual needs, and they did not specify what skill
ize raw metric effect size against the within group level is required for clinicians to successfully per-
variability. This can create overestimates of the form somatosensory retraining; these components
standardized effect size if sampling is particularly of retraining require further understanding to guide
homogeneous and underestimates if sampling is and advance clinical practice. Studies varied in the
unrepresentatively heterogeneous. In the single case dose of homework they required from participants,
experimental designs, we observed raw data that and further research is required to identify the type,
pretest data did not conform to a Gaussian normal frequency, and acceptability of practice between
distribution,26,29 and other researchers reported devi- formal retraining sessions. We do not know if indi-
ations from normality in their group trial data.25,28,30 vidual differences in variables such as cognition
The data were, therefore, not considered sound and communication influence stroke survivors’
enough to construct the forest plots and conduct het- ability to respond to somatosensory retraining and
erogeneity analysis, sub-group comparison of future research into this topic of individual differ-
pooled effect size, or other inferential statistics of ence is required to further understand the clinical
typical meta-analysis in this review. Furthermore, generalisability of somatosensory retraining.
we had limited data from which to pursue confirma-
tory comparisons via meta-analysis of effect sizes,
Conclusion
including a substantial proportion of effect sizes
across different somatosensory modalities that were The findings of this systematic review suggest that
from the same study, posing the issue of uneven rep- somatosensory discrimination retraining may be an
resentation of particular sources and an issue of efficacious treatment option for at least partial reme-
inflated sample size. diation of somatosensory impairment of the arm fol-
lowing stroke. However, at present, there is limited
evidence to inform how somatosensory discrimina-
Future research tion retraining impacts on stroke survivors’ perfor-
Somatosensory effect sizes need to be quantified mance and participation in daily life, and subjective
relative to controlled conditions via randomized experience of somatosensation, hence limiting inter-
controlled clinical trials of somatosensory discrim- pretation of the clinical significance. The results
ination training, with sufficient sample size and suggest that stroke survivors in the chronic stage of
power. In addition, measures that have adequate stroke recovery may benefit from being offered the
scaling of impairment, are norm-referenced, and treatment option of somatosensory discrimination
define what a clinically meaningful effect is are retraining during stroke rehabilitation if they require
needed to permit direct estimation of the size of the and desire improvement in somatosensory function
reduction in impairment and the proportion of in their hand and arm following stroke.
cases in which large and clinically meaningful
between-group effect sizes are achieved. Clinical Messages
Furthermore, research indicates that somatosen-
sory impairments are related to performance and •• Somatosensory discrimination retrain-
participation outcomes after stroke;37–39 hence, ing improves upper limb somatosen-
future clinical trials would benefit from also meas- sory impairment after stroke.
uring functional arm outcomes to evaluate if stroke •• Evidence of the impact of somatosen-
survivors are receiving functional gains from sory discrimination retraining on arm
somatosensory retraining. This research is critical function and participation is currently
as the long-term goal of somatosensory retraining limited.
12 Clinical Rehabilitation 00(0)
22. Connell LA, Lincoln N and Radford K. Somatosensory patients stable poststroke. Neurorehabil Neural Repair
impairment after stroke: frequency of different deficits 2008; 22(5): 494–504.
and their recovery. Clin Rehabil 2008; 22(8): 758–767. 31. Carey LM and Matyas TA. Effectiveness of sensory dis-
23. Wolny T, Saulicz E, Gnat R, et al. Butler’s neuromobiliza- crimination training when delivered by family members: a
tions combined with proprioceptive neuromuscular facili- pilot study. Brain Impair 2008; 9(2): 140–151.
tation are effective in reducing of upper limb sensory in 32. Helliwell S. Does the use of a sensory re-education pro-
late-stage stroke subjects: a three-group randomized trial. gramme improve the somatosensory and motor function
Clin Rehabil 2010; 24(9): 810–821. of the upper limb in subacute stroke? A single case experi-
24. Cambier D, De Corte E, Danneels LA, et al. Treating sen- mental design. Brit J Occup Ther 2009; 72(12): 551–558.
sory impairments in the post-stroke upper limb with inter- 33. Carey LM, Macdonell R and Matyas TA. SENSe: study
mittent pneumatic compression. Clin Rehabil 2003; 17(1): of the effectiveness of neurorehabilitation on sensation: a
14–20. randomized controlled trial. Neurorehabil Neural Repair
25. Yekutiel M and Guttman E. A controlled trial of the 2011; 25(4): 304–313.
retraining of the sensory function of the hand in stroke 34. de Diego C, Puig S and Navarro X. A sensorimotor
patients. J Neurol Neurosurg Psychiatry 1993; 56(3): stimulation program for rehabilitation of chronic stroke
241–244. patients. Restor Neurol Neurosci 2013; 31(4): 361–371.
26. Carey LM, Matyas TA and Oke LE. Sensory loss in stroke 35. Cohen J. Statistical power analysis for the behavioral sci-
patients: effective training of tactile and proprioceptive dis- ences. Cambridge, MA: Academic Press, 2013.
crimination. Arch Phys Med Rehabil 1993; 74(6): 602–611. 36. Veerbeek JM, van Wegen E, van Peppen R, et al. What
27. Smania N, Montagnana B, Faccioli S, et al. Rehabilitation is the evidence for physical therapy poststroke? A sys-
of somatic sensation and related deficit of motor control in tematic review and meta-analysis. PLoS ONE 2014; 9(2):
patients with pure sensory stroke. Arch Phys Med Rehabil e87987.
2003; 84(11): 1692–1702. 37. Carey LM, Matyas TM and Baum C. Effects of soma-
28. Byl N, Roderick J, Mohamed O, et al. Effectiveness of tosensory impairment on participation post-stroke. Am J
sensory and motor rehabilitation of the upper limb fol- Occup Ther 2018; 72(3): 7203205100.
lowing the principles of neuroplasticity: patients stable 38. Doyle S, Bennett S and Dudgeon B. Upper limb post-
poststroke. Neurorehabil Neural Repair 2003; 17(3): stroke sensory impairments: The survivor’s experience.
176–191. Disabil Rehabil 2014; 36(12): 993–1000.
29. Carey LM and Matyas TA. Training of somatosensory 39. Meyer S, Karttunen AH, Thijs V, et al. How do soma-
discrimination after stroke: facilitation of stimulus gener- tosensory deficits in the arm and hand relate to upper
alization. Am J Phys Med Rehabil 2005; 84(6): 428–442. limb impairment, activity, and participation problems
30. Byl N, Pitsch EA and Abrams GM. Functional outcomes after stroke? A systematic review. Phys Ther 2014; 94(9):
can vary by dose: learning-based sensorimotor training for 1220–1231.