829795

research-article2019
CRE0010.1177/0269215519829795Clinical RehabilitationTurville et al.

CLINICAL
Original article REHABILITATION

Clinical Rehabilitation

The effectiveness of somatosensory 1­–13

© The Author(s) 2019
Article reuse guidelines:
retraining for improving sensory sagepub.com/journals-permissions
DOI: 10.1177/0269215519829795
https://doi.org/10.1177/0269215519829795

function in the arm following journals.sagepub.com/home/cre

stroke: a systematic review

Megan L Turville1,2 , Liana S Cahill1,2,3,

Thomas A Matyas1,2,4, Jannette M Blennerhassett5
and Leeanne M Carey1,2

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
Services and Sport, College of Science, Health and
Engineering, La Trobe University, Melbourne, VIC, Australia
2Neurorehabilitation and Recovery, Stroke Division, Florey
Institute of Neuroscience and Mental Health, Melbourne,
VIC, Australia
3Department of Occupational Therapy, School of Allied Health,
Australian Catholic University, Melbourne, VIC, Australia
4School of Psychology and Public Health, College of Science,
Health, and Engineering, La Trobe University, Melbourne,
VIC, Australia
5Department of Physiotherapy, Austin Health, Melbourne,
VIC, Australia
Corresponding author:
Leeanne M Carey, Occupational Therapy, School of Allied
Health, Human Services and Sport, College of Science, Health
and Engineering, La Trobe University, Bundoora, Melbourne,
VIC 3086, Australia.
Email: L.Carey@latrobe.edu.au
2 Clinical Rehabilitation 00(0)
Keywords
Stroke, somatosensation, treatment, intervention, effect size
Received: 25 January 2018; accepted: 2 December 2018
Introduction
International clinical guidelines for stroke vary
regarding recommendations for somatosensory
interventions for the arm and hand,1–4 and therapists
are presently uncertain about clinical decisions
regarding somatosensory treatment.5,6 This field of
stroke rehabilitation is complex as a range of inter-
ventions exists, such as compression, mirror ther-
apy, mobilisations, sensory integrative treatment,
somatosensory retraining programmes, and stimu-
lation methods involving sensory-motor, electrical,
thermal, or magnetic stimulation.7,8 Stimulation
methods are commonly used in clinical settings9,10
and this treatment generally involves the applica-
tion of somatosensory stimuli without active atten-
tion or discrimination.8 In comparison, methods
such as somatosensory retraining programmes
focus on learning-based approaches to facilitate
improvement in sensory discrimination and arm use
as well as neural plastic recovery after stroke.11,12
Uncertainty in the field may also arise from
inconclusive findings reported in previous system-
atic reviews.7,13 The Schabrun and Hillier13 review
reported moderate evidence for stimulation treat-
ment and limited evidence for retraining treat-
ments. On the contrary, the Doyle et al.7 review
reported ‘insufficient’ evidence for any somatosen-
sory treatments, yet identified three separate stud-
ies on mirror therapy, thermal stimulation, and
pneumatic compression that provided ‘prelimi-
nary’ evidence for somatosensory improvement.
These prior systematic reviews7,13 included inter-
ventions that differed in regards to the aim and
approach of treatment. Furthermore, some reviews
included studies that did not specifically treat
somatosensory deficits nor measure somatosen-
sory outcomes.14 This inclusion of studies with
alternative aims and outcomes may have limited
our understanding of somatosensory treatment out-
comes in previous systematic reviews.
On the basis of systematic review evidence, we
do not presently know if somatosensory retraining
programmes are an effective intervention choice
for somatosensory impairment in the arm after
stroke. Somatosensory retraining programmes
involve activities of somatosensory discrimination
and recognition focused on skill learning to pro-
mote adaptive neuroplastic changes post
stroke.11,12,15 Somatosensory discrimination retrain-
ing programmes seem to have promising out-
comes,15–17 yet this intervention type has not been
separately considered in previous systematic
reviews. To clarify evidence in the field, we
focused on systematically reviewing the literature
for this sensory retraining intervention that involves
perceptual learning of somatosensory skills. We
aimed to determine if somatosensory retraining
programmes, with a focus on somatosensory dis-
crimination and recognition, improve somatosen-
sory discrimination skills and arm function in
people with stroke who experience somatosensory
impairment.
Method
This systematic review protocol was registered
with the international prospective register of sys-
tematic reviews (PROSPERO) (reference number:
CRD42015017821).
Search strategy
A systematic search of literature was conducted using
PRISMA guidelines.18 The following databases were
searched up until 29 October 2018: Medline,
Cumulative Index to Nursing and Allied Health
Literature (CINAHL), PsychInfo, Embase, Amed,
Web of Science, Physiotherapy Evidence Database
(PEDro), OT seeker, Cochrane Library, Cochrane
Central Register of Controlled Trials, Stroke engine,
Evidence-based Review of Stroke Rehabilitation, and
Turville et al. 3
Database of Research into Stroke (DORIS). The fol-
lowing websites were also searched: http://www
.opengrey.eu; http://www.controlled-trials.com; and
http://www.ClinicalTrials.gov.
The search strategy was broad to address the
diverse terminology in this field and involved col-
laborative work of two authors (MLT and LMC).
MeSH terms and free text terms were used, along
with truncation and wild card symbols to retrieve
all relevant keywords. Search terms are listed in
Supplemental Appendix 1. To complete word cita-
tion tracking, key authors and references were
entered in Science Citation Index to identify fur-
ther studies. The reference lists of relevant papers
and reviews were hand searched to ensure all appli-
cable research was obtained.
Study selection
Studies were included if they involved (1) a sample
of adults with diagnosis of stroke and measured
somatosensory impairment in the arm; (2) an inter-
vention for retraining somatosensory discrimina-
tion skills in the arm; (3) a primary outcome related
to a somatosensory discrimination modality; and
(4) a study design such as randomized control trial,
controlled clinical trial, or single case experimental
design. When present, outcomes of arm function
were also extracted from included studies.
A study was excluded if the (1) intervention
focused on somatosensory stimulation only; (2)
study did not measure somatosensation discrimina-
tion skills as a primary outcome; (3) intervention
used sensation as a modality to improve movement
in the arm, not sensation; (4) article was written in
a language other than English or not full text; (5)
publication was a descriptive case study/series,
cohort study, review, book chapter, or editorial; or
(6) sample did not have somatosensory impair-
ment. Two review authors (MLT and LSC)
reviewed article titles and abstracts independently
to identify potentially relevant studies. Full text of
identified articles was then retrieved and two
review authors (MLT and LSC) independently
screened studies according to inclusion criteria.
When differences occurred, a third reviewer (LMC)
was consulted for confirmation of criteria.
Data extraction and quality assessment
Data were extracted according to established crite-
ria of key evaluation (e.g. blinding, independent
assessment) and content (e.g. dose, theory) param-
eters. The primary author (MLT) extracted data and
a second author (LSC) verified accuracy and com-
pleteness of extracted data. In the case of missing
data, authors were contacted in an attempt to obtain
and include such data. The quality of studies was
evaluated using the following tools: (1)
Physiotherapy Evidence Database (PEDro-P) scale
for randomized and non-randomized group
designs,19 or (2) Risk of Bias in N-of-1 Trials
(RoBiNT) scale for single case experimental
designs.20 Two authors (MLT and LSC) completed
quality assessment independently, and a third
reviewer (LMC) was consulted if discrepancies
arose.
Data synthesis and analysis
Effect sizes were computed from single groups of
participants (i.e. individual studies) undertaking
either intervention and/or control conditions. To
understand somatosensory changes, effect sizes
were calculated for each somatosensory modality
outcome (i.e. postphase mean result minus pre-
phase mean result, divided by the pretest stand-
ard deviation). We used the pretest standard
deviation to standardize the raw metric effect size
because it provided a representative measure of
variability between individuals in the preinter-
vention state. This index most closely resembles
Glass’ Delta (Glass δ) and enabled computation
of a standardized effect size for each intervention
group.21 This was necessary given that not all
studies possessed control groups, and our aim
was to describe effect sizes from all groups pos-
sessing sufficient data, not just those obtained
from parallel group trials.
We also reviewed the literature to obtain rele-
vant data that would provide a frame of reference
for interpreting intervention group results. For esti-
mates of effects produced by control conditions,
we used results from (1) the control groups of
somatosensory discrimination intervention studies
4 Clinical Rehabilitation 00(0)
with controlled clinical design; (2) other soma-
tosensory intervention studies that did not apply
discrimination retraining, but did provide relevant
control data; and (3) observational studies of arm
somatosensory recovery after stroke. We adopted
this approach given the lack of control data availa-
ble in many (N = 7) of the somatosensory discrimi-
nation studies.
We identified studies that (1) measured recov-
ery of somatosensation post-stroke in a control
group or observational study; (2) assessed soma-
tosensory discrimination in the arm and/or hand in
at least one sensory modality; (3) published mean
and standard deviation data from which effect sizes
could be calculated; (4) included a sample of at
least 10 stroke survivors; and (5) measured change
over a period of weeks to months.
We found three studies to act as control proxies:
(1) the Connell et al.22 study that assessed soma-
tosensory recovery at 0–6 months after stroke in
two rehabilitation facilities (N = 80); (2) the Wolny
et al.23 intervention study on Butler’s neuromobili-
sations, which included a control group (N = 32)
using ‘traditional’ post-stroke therapy (e.g. multi-
disciplinary treatment); and (3) Cambier et al.24
randomized control trial study on intermittent
pneumatic compression, with a control group that
received standard therapy plus sham short wave
therapy to their affected arm (N = 12).
Where possible, we pooled data from single
case experimental design intervention and con-
trol phases to calculate respective effect sizes. To
achieve consistency with published group stud-
ies, we calculated for each case the difference
between start and end of baselines to obtain con-
trol group mean changes. The standard deviation
used to obtain the quasi-Glass δ  was estimated
from a pooled sample of baseline phase pretests
(i.e. the first tests in the time series, which repre-
sented individual differences in preintervention
states of impairment). For intervention effects,
we calculated the change from end of baseline to
end of intervention for each case. These differ-
ences were averaged and the mean change was
then standardized using the standard deviation of
the baseline phase pretest, as for the baseline
change effect.
The majority of studies did not provide suffi-
cient information to compute standard errors for
mean changes, thus confidence intervals for stand-
ardized effect sizes, formal effect size pooling, and
forest plots could not be constructed. Therefore, we
conducted a descriptive meta-analysis based on
point estimates of mean changes.
Finally, narrative analysis was used for addi-
tional outcomes related to arm functioning
because there was insufficient data from which
effect sizes could be calculated. Descriptive sta-
tistics and narrative analysis were employed to
characterize the population sample and interven-
tion characteristics.
Results
A flowchart of search results is presented in Figure 1.
Ten studies were included in this review25–34 and are
presented in Table 1.
Study designs and quality assessment
The study designs and quality assessment are detailed
in Table 1. In general, the quality of study designs
was relatively low for group trials (i.e. less than
5/10), with the exception of one randomized control
trial33 achieving 10/10 on the PEDro-P scale19 (Table
1). Single case experimental design studies (n = 5)
resulted in more consistent quality ratings that ranged
from 13 to 18 (out of 30) on the RoBiNT scale20
(Table 1). For detailed information on quality ratings,
see Supplemental Appendix 2.
Participant characteristics
This review involved 199 stroke survivors (M = 20,
SD = 17.44) and, on average, participants were
59.5  years old (SD = 10.23) and 1.8 years post
stroke (range: 3 weeks–6.2 years). Male partici-
pants were more common than females (65%
males; 35% females), and the sample contained
nearly equal numbers of right-sided (48%) and left-
sided strokes (52%). Typically, participants did not
have significant communication difficulties, previ-
ous neurological conditions, or perceptual difficul-
ties, such as unilateral spatial neglect.
Turville et al. 5

Figure 1.  Flowchart of systematic search results.

 6

Table 1.  Summary of Included Somatosensory Discrimination Retraining Studies.

Group studies

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)
Somatosensory discrimination retraining
intervention
Interventions either focused on retraining soma-
tosensation (6/10 studies), or combined soma-
tosensory and motor retraining (4/10 studies)
(Table 1). Studies reported a range of treatment
principles and the most common were
•• Graded difficulty of a variety of somatosensory
discrimination tasks25,26,28–33
•• Use of attention when exploring sensory fea-
tures of task/objects25,26,28–33
•• Feedback on task performance26–29,31–33
•• Use of vision and unaffected hand to facilitate
relearning of sensation in the affected
hand25,26,29,31–34
•• Focus on functional and/or meaningful
tasks25,28,30,32,34
Other treatment principles referred to include
repetition,28,30–33 client anticipating what to
feel,26,29,31,33 mental rehearsal to reinforce learn-
ing,28,30,34 education regarding neuroplasticity and
treatment principles,28,29,33 and tailoring of the pro-
gramme to individuals’ needs and abili-
ties.25,26,29,32–34 Studies generally used
somatosensory discrimination activities in treat-
ment that differed from those used in assessment
phases;25,27–30,32–34 only two studies assessed the
same tasks used in treatment.26,31
Three studies involved additional therapeutic
modalities, such as mirror therapy, guided mental
imagery of movements or functional activities, rec-
ommended glove use on unaffected hand during
day,28,30 practice of functional activities, and use of
a mitten on the unaffected hand during treatment
sessions.34 In relation to studies that involved a
control intervention, this consisted of ‘non-treat-
ment’ (see ref. 25, p.242), ‘non-specific repeated
exposure to stimuli’ varying in somatosensory
attributes (see ref. 33, p.306), or ‘usual treatment
according to the Bobath concept, without prioritis-
ing therapy of the upper limb’ (see ref. 34, p.362).
In general, treatment involved retraining of
either two or three somatosensory modalities. The
somatosensory modalities of tactile discrimination
and proprioception were the focus of retraining in
all studies.25–34 Object recognition was also
retrained in six of the 10 studies.25,27,28,30,32,33 Two
studies reported a particular focus on haptic activi-
ties, graphesthesia, and use of hand with sensory
interface of object.28,30
On average, somatosensory retraining occurred
for a total of 16 sessions (SD = 6.6), across four to
eight weeks, two to three times per week. Each
treatment session was approximately 70 minutes in
total duration, with an average of 18 total treatment
hours. Studies reported those who delivered the
retraining were trained therapists.26,28–30,32–34 One
study used a therapist to train family members to
deliver the intervention.31 There was minimal
information to further understand the characteris-
tics of treating therapists (i.e. amount of training
and experience). Retraining occurred in an indi-
vidual context with one therapist and one stroke
survivor. Somatosensory discrimination retraining
occurred in the outpatient clinic setting, partici-
pants’ home, or a combination of both, and one
study32 provided retraining in inpatient acute
setting.
Half of the studies reported they specifically
assigned homework to participants, and the recom-
mended homework dosage varied from 15–90 min-
utes per day.27,28,30,32,34 The descriptions of
homework reported either exercises similar to
those performed during supervised somatosensory
or sensorimotor retraining, practice of functional
activities, or wearing a glove on the unaffected
hand.27,28,30,32,34 Studies measuring homework
compliance reported that participants had variable
performance with completing such tasks.27,28,30
Outcome results
Somatosensory discrimination skills. A variety of
somatosensory discrimination measures were used
(Table 1), which were standardized measures with
minor exceptions.25,34
All somatosensory modalities improved with
retraining, and somatosensory effect sizes are dis-
played in Figure 2. Somatosensory intervention
effect sizes ranged from 0.3 to 2.2, with an average
effect size of 0.85 across somatosensory modalities
Turville et al. 9

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.

(Figure 2). Narrative analysis of four studies also Discussion

suggested somatosensory improvements.27,28,31,32
Arm function. Arm use outcomes were present in
six studies (n = 89 participants)27,28,30,32–34 and are
listed in Table 1. Overall, studies reported a posi-
tive trend in results in arm functioning after retrain-
ing. De Diego et al.34 found that participants in the
experimental group showed gains in arm perfor-
mance and use following retraining; however,
results were not significant relative to control
group findings. Byl et al.28 and Byl et al.30 found
significant improvements in fine motor skills, mus-
culoskeletal performance (strength and range of
motion), and functional independence. Similarly,
Smania et al.27 found improvements in upper limb
and functional outcomes following intervention,
with statistical significance varying among the four
participants for motor sequence, reach, and grasp.
Helliwell’s study32 showed a trend towards
improvement in motor ability; however, results are
limited to one participant. Carey et al.33 did not
report on the results for arm functioning.
Somatosensory discrimination skills
The results suggest that somatosensory retraining
may improve stroke survivors’ ability to discrimi-
nate bodily sensations in the arm and hand. The
findings demonstrated that a clear majority of
effect sizes for proprioception and tactile soma-
tosensory domains were greater than 0.5, and all
but one of the effect sizes were larger than the con-
trol intervention effect sizes, at least as point esti-
mates. In quantitative terms, the most common
intervention effect sizes were between 0.75 and 1.6
standard deviations. Effect sizes of 0.80 and larger
suggest a large effect, based on the broad range of
effects seen in samples in the behavioural and
health research literature.35
Unlike previous systematic reviews,7,13 the
present review contains evidence that relates to
interventions specifically designed to treat soma-
tosensory discrimination skills in the arm and hand
after stroke, and involved synthesis of data from
group and single case experimental design studies.
10 Clinical Rehabilitation 00(0)
There was a relative lack of control data in the
included studies from which to compare interven-
tion results, and future randomized control
trials are required to address this gap.7,13,16,17
Consequently, we sought additional somatosensory
control data from the literature to use in our analy-
ses. The calculation of control effect sizes and
comparison with intervention effect sizes in this
way is a development to the current literature, par-
ticularly when randomized control trials are lim-
ited. This method showed that even without sound
confidence interval estimates for the effects of
interventions versus control conditions, it is never-
theless the case that intervention somatosensory
effect sizes were larger than control group soma-
tosensory effect sizes.
Arm function
Based on a descriptive summary of data (N = 89
participants), the findings provided limited evi-
dence to suggest that somatosensory discrimina-
tion retraining may facilitate arm functioning, and
future research is needed. Prior systematic reviews
also found inconclusive results regarding whether
somatosensory treatments change stroke survivors’
arm functioning or general functional status.7,36
Historically, performance and participation out-
comes have not been widely investigated in soma-
tosensory treatment programmes,8 which may
contribute to our present lack of knowledge.
Study quality
The quality of group study designs was relatively
low, with the exception of one randomized control
trial.33 This study did find a significant between-
group effect.33 The main methodological limita-
tions of group studies related to a lack of
randomization,25,30 concealed allocation,25,28,30 and
blinding of participants and therapists.25,28,30,34
Most group studies involved blinded asses-
sors,28,30,33,34 although this was not the case for
some studies.26,27,29,31,32 Single case experimental
design studies were of a consistent quality that
ranged from 13 to 18 (out of 30) on the RoBiNT
scale.20 In previous systematic reviews, study
quality was also found to be relatively low and
variable.7,13 Effect sizes must be interpreted with
consideration of study quality ratings (Table 1 and
Supplemental Appendix 2), which highlight meth-
odological limitations.
Limitations
This review is limited to the investigation of soma-
tosensory retraining with a chronic stroke population.
Participants involved in somatosensory retraining
studies were typically in the chronic phase of recov-
ery (M = 1.8 years post-stroke), were 60 years of age
(SD = 10.23), and had reasonable communication and
perceptual abilities. Furthermore, the sample size dif-
fered considerably among studies (N = 1–50), and the
total number of participants included in this review
(N = 199) was relatively small.
Four of the 10 studies included motor retraining
in addition to somatosensory retraining,27,28,30,34
and three of these studies also included additional
therapeutic modalities, such as mirror therapy,
guided mental imagery, and constraint induced
practices for the unaffected hand.28,30,34 This has
implications particularly in relation to the motor
outcomes reported and confounds the ability to
separate contributions from somatosensory dis-
crimination training alone.
The control groups investigated in this review
were heterogeneous in nature (e.g. usual care,
expose to somatosensory stimuli without retraining
principles, sham therapy), and this could have
impacted on the comparison of results. We acknowl-
edge that tactile effect sizes in three groups related
more to detection than discrimination skills,22,24,34
and, in general, somatosensory outcome measures
were heterogeneous and lacking normative stand-
ards. Given the timeframe of the Connell et al.22
study, results are likely to include the period in
which maximum spontaneous recovery occurred
and therefore provided a conservative estimate for
comparison of interventions that tended to involve
stroke survivors in the chronic stage of recovery.
Overall, an inferential meta-analysis could not be
conducted due to lack of information about standard
errors of change scores and the distribution of
change scores. In addition, effect size results may be
Turville et al. 11
limited depending on the representativeness of
standard deviation findings.21 A weakness of para-
metric standardized effect size indexes, including
the one employed in this study, is that they standard-
ize raw metric effect size against the within group
variability. This can create overestimates of the
standardized effect size if sampling is particularly
homogeneous and underestimates if sampling is
unrepresentatively heterogeneous. In the single case
experimental designs, we observed raw data that
pretest data did not conform to a Gaussian normal
distribution,26,29 and other researchers reported devi-
ations from normality in their group trial data.25,28,30
The data were, therefore, not considered sound
enough to construct the forest plots and conduct het-
erogeneity analysis, sub-group comparison of
pooled effect size, or other inferential statistics of
typical meta-analysis in this review. Furthermore,
we had limited data from which to pursue confirma-
tory comparisons via meta-analysis of effect sizes,
including a substantial proportion of effect sizes
across different somatosensory modalities that were
from the same study, posing the issue of uneven rep-
resentation of particular sources and an issue of
inflated sample size.
Future research
Somatosensory effect sizes need to be quantified
relative to controlled conditions via randomized
controlled clinical trials of somatosensory discrim-
ination training, with sufficient sample size and
power. In addition, measures that have adequate
scaling of impairment, are norm-referenced, and
define what a clinically meaningful effect is are
needed to permit direct estimation of the size of the
reduction in impairment and the proportion of
cases in which large and clinically meaningful
between-group effect sizes are achieved.
Furthermore, research indicates that somatosen-
sory impairments are related to performance and
participation outcomes after stroke;37–39 hence,
future clinical trials would benefit from also meas-
uring functional arm outcomes to evaluate if stroke
survivors are receiving functional gains from
somatosensory retraining. This research is critical
as the long-term goal of somatosensory retraining
is to enable people to have a more functional arm
use after stroke.11 Intervention descriptions lacked
examples of how retraining was tailored to indi-
vidual needs, and they did not specify what skill
level is required for clinicians to successfully per-
form somatosensory retraining; these components
of retraining require further understanding to guide
and advance clinical practice. Studies varied in the
dose of homework they required from participants,
and further research is required to identify the type,
frequency, and acceptability of practice between
formal retraining sessions. We do not know if indi-
vidual differences in variables such as cognition
and communication influence stroke survivors’
ability to respond to somatosensory retraining and
future research into this topic of individual differ-
ence is required to further understand the clinical
generalisability of somatosensory retraining.
Conclusion
The findings of this systematic review suggest that
somatosensory discrimination retraining may be an
efficacious treatment option for at least partial reme-
diation of somatosensory impairment of the arm fol-
lowing stroke. However, at present, there is limited
evidence to inform how somatosensory discrimina-
tion retraining impacts on stroke survivors’ perfor-
mance and participation in daily life, and subjective
experience of somatosensation, hence limiting inter-
pretation of the clinical significance. The results
suggest that stroke survivors in the chronic stage of
stroke recovery may benefit from being offered the
treatment option of somatosensory discrimination
retraining during stroke rehabilitation if they require
and desire improvement in somatosensory function
in their hand and arm following stroke.
Clinical Messages
•• Somatosensory discrimination retrain-
ing improves upper limb somatosen-
sory impairment after stroke.
•• Evidence of the impact of somatosen-
sory discrimination retraining on arm
function and participation is currently
limited.
12 Clinical Rehabilitation 00(0)
Authors’ Note
LMC and TAM were involved in several somatosensory
retraining studies reported on in this review and for this
reason, MLT and LSC were involved in study selection,
data extraction, and quality assessment appraisal.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest
with respect to the research, authorship, and/or publica-
tion of this article.
Funding
The author(s) disclosed receipt of the following financial
support for the research, authorship, and/or publication
of this article: This work was supported by the National
Health and Medical Research Council (NHMRC) of
Australia (grant numbers 191214, 1022694, 1077898,
1134495), and the James S. McDonnell Foundation
(grant number 220020413). MT and LSC were sup-
ported by Australian Government Research Training
Program Scholarships from La Trobe University.
ORCID iD
Megan L Turville https://orcid.org/0000-0002-3938
-6778
Supplemental material
Supplemental material for this article is available online.
References
1. Hebert D, Lindsay MP, McIntyre A, et al. Canadian stroke
best practice recommendations: stroke rehabilitation prac-
tice guidelines, update 2015. Int J Stroke 2016; 11(4):
459–484.
2. Intercollegiate Stroke Working Party. National clinical
guidelines for stroke. 5th ed. London: Royal College of
Physicians, 2016.
3. Stroke Foundation. Clinical guidelines for stroke man-
agement. Melbourne, VIC, Australia: Stroke Foundation,
2017.
4. Winstein CJ, Stein J, Arena R, et al. Guidelines for adult
stroke rehabilitation and recovery: a guideline for health-
care professionals from the American Heart Association/
American Stroke Association. Stroke 2016; 47(6): e98–
e169.
5. Doyle SD, Bennett S and Gustafsson L. Clinical decision
making when addressing upper limb post-stroke sensory
impairments. Brit J Occup Ther 2013; 76(6): 254–263.
6. Doyle S, Bennett S and Dudgeon B. Sensory impairment
after stroke: exploring therapists’ clinical decision mak-
ing. Can J Occup Ther 2014; 81(4): 215–225.
7. Doyle S, Bennett S, Fasoli S, et al. Interventions for sen-
sory impairment in the upper limb after stroke. Cochrane
Database Syst Rev 2010; 6: CD006331.
8. Carey LM, Lamp G and Turville M. The state-of-the-sci-
ence on somatosensory function and its impact on daily
life in adults, older adults and following stroke: a scoping
review. OTJR 2016; 36(2 suppl.): 27S–41S.
9. Pumpa LU, Cahill LS and Carey LM. Somatosensory
assessment and treatment after stroke: an evidence-prac-
tice gap. Aust Occup Ther J 2015; 62(2): 93–104.
10. Doyle S, Bennett S and Gustafsson L. Occupational ther-
apy for upper limb post-stroke sensory impairments: a
survey. Br J Occup Ther 2013; 76(10): 434–442.
11. Yekutiel M. Sensory re-education of the hand after stroke.
London: Whurr, 2000.
12. Carey LM. Touch and body sensations. In: Carey LM
(ed.) Stroke rehabilitation: insights from neuroscience
and imaging. New York: Oxford University Press, 2012,
pp.157–172.
13. Schabrun SM and Hillier S. Evidence for the retraining of
sensation after stroke: a systematic review. Clin Rehabil
2009; 23(1): 27–39.
14. Carey LM, Blennerhassett JM and Matyas TA.
Commentary: evidence for the retraining of sensation
after stroke remains limited. Aust Occup Ther J 2010;
57(3): 200–204.
15. Carey LM. Somatosensory loss after stroke. Crit Rev Phys
Rehabil Med 1995; 7(1): 51–91.
16. Borstad AL and Nichols-Larsen DS. Assessing and treat-
ing higher level somatosensory impairments post stroke.
Top Stroke Rehabil 2014; 21(4): 290–295.
17. Kessner SS, Bingel U and Thomalla G. Somatosensory
deficits after stroke: a scoping review. Top Stroke Rehabil
2016; 23(2): 136–146.
18. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA
statement for reporting systematic reviews and meta-
analyses of studies that evaluate health care interventions:
explanation and elaboration. PLoS Med 6(7): e1000100.
19. Perdices M, Savage S, Tate R, et al. Rater’s manual for
between-group studies (RCTS and NonRCTs): introduc-
tion to the Physiotherapy Evidence Database (PEDro)
Scale for rating methodological Quality, Adapted for
PsycBITE (Pedro-P). Sydney, NSW, Australia: University
of Sydney, 2009.
20. Tate RL, Rosenkoetter U, Wakim D, et al. The Risk of Bias
in N-of-1 Trials (RoBiNT) Scale: an expanded manual
for the critical appraisal of single case reports. Sydney,
NSW, Australia: John Walsh Centre for Rehabilitation
Research, 2015.
21. Grissom RJ and Kim JJ. Effect sizes for research: univari-
ate and multivariate applications. New York: Taylor and
Francis Group, 2012.
Turville et al. 13
22. Connell LA, Lincoln N and Radford K. Somatosensory
impairment after stroke: frequency of different deficits
and their recovery. Clin Rehabil 2008; 22(8): 758–767.
23. Wolny T, Saulicz E, Gnat R, et al. Butler’s neuromobiliza-
tions combined with proprioceptive neuromuscular facili-
tation are effective in reducing of upper limb sensory in
late-stage stroke subjects: a three-group randomized trial.
Clin Rehabil 2010; 24(9): 810–821.
24. Cambier D, De Corte E, Danneels LA, et al. Treating sen-
sory impairments in the post-stroke upper limb with inter-
mittent pneumatic compression. Clin Rehabil 2003; 17(1):
14–20.
25. Yekutiel M and Guttman E. A controlled trial of the
retraining of the sensory function of the hand in stroke
patients. J Neurol Neurosurg Psychiatry 1993; 56(3):
241–244.
26. Carey LM, Matyas TA and Oke LE. Sensory loss in stroke
patients: effective training of tactile and proprioceptive dis-
crimination. Arch Phys Med Rehabil 1993; 74(6): 602–611.
27. Smania N, Montagnana B, Faccioli S, et al. Rehabilitation
of somatic sensation and related deficit of motor control in
patients with pure sensory stroke. Arch Phys Med Rehabil
2003; 84(11): 1692–1702.
28. Byl N, Roderick J, Mohamed O, et al. Effectiveness of
sensory and motor rehabilitation of the upper limb fol-
lowing the principles of neuroplasticity: patients stable
poststroke. Neurorehabil Neural Repair 2003; 17(3):
176–191.
29. Carey LM and Matyas TA. Training of somatosensory
discrimination after stroke: facilitation of stimulus gener-
alization. Am J Phys Med Rehabil 2005; 84(6): 428–442.
30. Byl N, Pitsch EA and Abrams GM. Functional outcomes
can vary by dose: learning-based sensorimotor training for
patients stable poststroke. Neurorehabil Neural Repair
2008; 22(5): 494–504.
31. Carey LM and Matyas TA. Effectiveness of sensory dis-
crimination training when delivered by family members: a
pilot study. Brain Impair 2008; 9(2): 140–151.
32. Helliwell S. Does the use of a sensory re-education pro-
gramme improve the somatosensory and motor function
of the upper limb in subacute stroke? A single case experi-
mental design. Brit J Occup Ther 2009; 72(12): 551–558.
33. Carey LM, Macdonell R and Matyas TA. SENSe: study
of the effectiveness of neurorehabilitation on sensation: a
randomized controlled trial. Neurorehabil Neural Repair
2011; 25(4): 304–313.
34. de Diego C, Puig S and Navarro X. A sensorimotor
stimulation program for rehabilitation of chronic stroke
patients. Restor Neurol Neurosci 2013; 31(4): 361–371.
35. Cohen J. Statistical power analysis for the behavioral sci-
ences. Cambridge, MA: Academic Press, 2013.
36. Veerbeek JM, van Wegen E, van Peppen R, et al. What
is the evidence for physical therapy poststroke? A sys-
tematic review and meta-analysis. PLoS ONE 2014; 9(2):
e87987.
37. Carey LM, Matyas TM and Baum C. Effects of soma-
tosensory impairment on participation post-stroke. Am J
Occup Ther 2018; 72(3): 7203205100.
38. Doyle S, Bennett S and Dudgeon B. Upper limb post-
stroke sensory impairments: The survivor’s experience.
Disabil Rehabil 2014; 36(12): 993–1000.
39. Meyer S, Karttunen AH, Thijs V, et al. How do soma-
tosensory deficits in the arm and hand relate to upper
limb impairment, activity, and participation problems
after stroke? A systematic review. Phys Ther 2014; 94(9):
1220–1231.