Methods

Design: Randomised parallel‐group study

Participants

Health condition: Children with spastic cerebral palsy

Sample size: Experimental group: 13, Control group: 12, Other group: 14

Setting, Country: Outpatient clinics, USA

Joint of interest: Ankle

Inclusion criteria:

• Diagnosis of spastic hemiplegia or diplegia

• Between 3‐10 years old

• Ambulate independently without assistive devices

• Ambulate in functional equinus (toe‐toe or heel‐toe pattern)

• Neutral ankle position with full knee extension

Exclusion criteria:

• Previous orthopaedic surgery to tendo‐achilles or sub‐talar joint

• No botulinum toxin injections in previous 6 months

• Hip or knee flexion contractures greater than 10°

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture

Mean age (range): Experimental group: 6 years (3‐8), Control group: 6 years (3‐9), Other group: 6 years (3‐9)

Gender: Experimental group: 54% female, Control group: 50% female, Other group: 57% female

Interventions

Groups included in this review:

Experimental group: Botulinum toxin plus cast

Participants received botulinum toxin injections into gastrocnemius muscle followed by cast for 3 weeks at baseline, 3 months and 6 months. Ankle‐foot orthosis (AFO) worn in between casting periods for 20‐22 h/d

Total stretch time: 24 h x 7 d x 9 weeks = 1512 hours over a 6‐month period

Control group: Botulinum toxin

Participants received botulinum toxin injections into gastrocnemius muscle at baseline, 3 months and 6 months. AFO worn for 20‐22 h/d

Other group: Placebo plus cast

Participants received placebo injections into gastrocnemius muscle followed by cast for 3 weeks at baseline, 3 months and 6 months. AFO worn in between casting periods for 20‐22 h/d

Total stretch time: 24 h x 7 d x 3 weeks = 504 hours over a 3‐week period

Outcomes

Outcomes included in this review:

• Passive ankle dorsiflexion with the knee extended (degrees)

• Triceps surae spasticity (Ashworth)

Other outcomes: Passive ankle dorsiflexion (knee flexed), active ankle dorsiflexion (knee flexed), ankle dorsiflexion at initial contact during gait, peak ankle dorsiflexion during stance, peak ankle dorsiflexion during swing, triceps surae spasticity (Tardieu), walking velocity, stride length, ankle plantarflexion strength, ankle dorsiflexion strength, ankle power generation

Time points included in this review: Outcomes measured at 12 months

Other time points: Outcomes also measured at baseline, 3 months, 6 months and 7.5 months

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote: "...block design randomisation sequence", p 621

Comment: Insufficient detail reported

Allocation concealment (selection bias)

Unclear risk

Comment: Insufficient detail reported. If concealment was used, every third allocation could be determined due to the use of a fixed blocked sequence

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...the children and parents were instructed not to discuss their treatment with the evaluating clinician to ensure that the clinician maintained blinding to the treatment group", p 6

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: 5/39 (13%) dropouts

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

High risk

Quote: "...leading to an early termination of the study before obtaining the projected number of children", p 622

Comment: possible cause of bias introduced by early termination of study

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with stroke

Sample size: Experimental group: 18, Control group: 18

Setting, Country: Inpatient rehabilitation units of 4 metropolitan hospitals, Australia

Joint of interest: Shoulder

Inclusion criteria:

• Experienced first stroke within the previous 20 days

• Had hemiplegia

• Between 50‐80 years old

• At risk of developing contracture as a result of having little or no upper limb function ‐ defined as a score of 0‐4 on item 6 of the MAS

Exclusion criteria:

• Already had a shoulder problem ‐ defined as pain or loss of greater than 20° of intact shoulder ROM in either external rotation or flexion

• Had cognitive problems that precluded them from participating in the positioning programme

Existing contracture, at risk of contracture, or combination of both: Participants were at risk of developing contracture

Mean age (SD): Experimental group: 70 years (7), Control group: 64 years (9)

Gender: Experimental group: 60% female, Control group: 56% female

Interventions

Groups included in this review:

Experimental group: Shoulder positioning and routine care

Participants received 2 x 30‐min sessions of shoulder positioning:
Position 1 ‐ participants in supine, 45° shoulder abduction and maximal external rotation
Position 2 ‐ participants sitting with arm on table with shoulder flexed to 90° and elbow bent at 90°

Participants also received up to 10 min shoulder exercises and routine upper‐limb care

Total stretch time: 30 min x 5 days x 4 weeks = 10 h for each position over a 4‐week period

Control group: Routine care

Participants received up to 10 min shoulder exercises and routine upper‐limb care

Other groups: Nil

Outcomes

Outcomes included in this review:

• Maximum passive shoulder external rotation of the affected limb (degrees)

• Pain experienced during maximal external rotation (yes/no)

• Item 6 MAS (Limits 0‐6; 0 = worse, 6 = better)

Other outcomes: Maximum passive shoulder flexion (affected limb), shoulder contracture in external rotation (as compared to intact limb), shoulder contracture in flexion (as compared to intact limb), pain experienced during maximal flexion

Time points included in this review: Outcomes measured at discharge (or 4 weeks) ‐ whichever was the sooner

Other time points: Outcomes also measured at baseline

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Comment: Insufficient detail reported in paper. Author correspondence revealed that the randomisation sequence was computer generated

Allocation concealment (selection bias)

Low risk

Quote: "...centrally randomized into either the experimental or the control group", p 231

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...an assessor blinded to group allocation carried out measurements", p 231

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: 5/36 (14%) dropouts

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with knee osteoarthritis

Sample size: Experimental group: 17 (33 knees), Control group: 19 (33 knees)

Setting, Country: Outpatient clinic of a large metropolitan hospital, Japan

Joint of interest: Knee

Inclusion criteria:

• Severe unilateral or bilateral knee osteoarthritis established using radiography

• Planning to undergo total knee arthroplasty

Exclusion criteria:

• Could not follow instructions

• Could not lie prone

• Self‐reported severe cardiovascular disease, neurological disease, or lower limb disorders other than knee osteoarthritis

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 72 years (5), Control group: 74 years (6)

Gender: Experimental group: 100% female, Control group: 100% female

Interventions

Groups included in this review:

Experimental group : Home‐based stretch

Participants self‐administered two knee flexion stretches (sitting on the floor and prone)

Total stretch time¹: 5 min x 7 d x 11.6 weeks = 6.7 h over a 3‐month period

Control group : Maintain usual physical activity

Instructed to maintain their current level of physical activity

Other groups : Nil

Outcomes

Outcomes included in this review:

• Knee ROM in supine (degrees)

• Gait speed (m/min)

• Pain during gait (VAS)

Other outcomes:

Knee ROM during gait

Time points included in this review: Outcomes measured at time of admission (end of intervention)

Other time points: Outcomes also measured at baseline

Notes

¹The mean duration of treatment (81 days) was used to estimate the total stretch time for the Experimental group. Also assumed that participants performed 10 repetitions each day, not 10 repetitions of each exercise (20 repetitions).

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote: “…they were randomly allocated to stretching … and control groups”, p 114

Comment: Insufficient detail reported

Allocation concealment (selection bias)

Unclear risk

Quote: “…they were randomly allocated to stretching … and control groups”, p 114

Comment: Insufficient detail reported

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "S‐ROM was measured by a physiotherapist blinded to the participants", p 115

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Unclear risk

Comment: Insufficient detail reported

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes stated were reported

Other bias

High risk

Comment: More than one joint per participant but authors have not adequately accounted for this in the analysis

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with stroke

Sample size: Experimental group: 13, Control group: 13, Other group: 13

Setting, Country: Rehabilitation department in a university hospital, Turkey

Joint of interest: Wrist

Inclusion criteria:

• History of a single stroke

• Wrist Modified Ashworth Scale score ≥1+

Exclusion criteria:

• Cognitive impairment (determined by Mini‐Mental State Examination)

• Behavioural disturbances

• Severe chronic disease likely to interfere with co‐operation

• Cutaneous or joint pathologies in the upper limb preventing splinting

• Previous splinting of the upper limb within the last 8 weeks

• If taking antispasticity medication, dosage change in the last month

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 55 years (12) Control group: 60 years (10), Other group: 52 years (11)

Gender: Experimental group: 46% female, Control group: 42% female, Other group: 38% female

Interventions

Groups included in this review:

Experimental group : Volar splint and home‐based exercise programme

Participants wore each night a custom‐made static volar splint (thermoplastic resin with plastazote on the inner surface) with the hand positioned beyond the angle of 'catch' Participants also did home‐based exercise programme (details below)

Total stretch time: 10 hx 7 d x 5 weeks = 350 h over a 1.25‐month period

Control group : Home‐based exercise programme only

Participants stretched the wrist and finger flexors plus practiced reaching and grasping an object, 10 repetitions of each 3 x d. In addition they were instructed to use their hands as much as possible during daily activities

Other group : Dorsal splint and home‐based exercise programme

Custom‐made static dorsal splint (thermoplastic resin with plastazote on the inner surface) with the hand positioned beyond the angle of 'catch' worn overnight

Total stretch time: 8 h x 7 d x 5 weeks = 280 h over a 1.25‐month period

Outcomes

Outcomes included in this review:

• Passive wrist extension (degrees)

• Spasticity (Modified Ashworth Scale)

Other outcomes:

H latency of flexor carpi radialis, Hmax:Mmax ratio of flexor carpi radialis

Time points included in this review: Outcomes measured at 5 weeks (end of intervention)

Other time points: Outcomes also measured at baseline

Notes

Assumed participants wore the splint for 8 h per night when calculating total stretch time

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: “…subjects were randomly allocated to control and experimental groups by using a simple randomization process (computer‐generated random numbers) after baseline measurements”, p 330

Allocation concealment (selection bias)

Low risk

Quote: “An independent person was responsible for randomization and group assignment”, p 330

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

High risk

Quote: "Measurements associated with electroneuromyography (ENMG) were blinded ...but the others were not", p 331‐2
Comment: Range of motion and spasticity measurements were not blinded

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: 1/39 (3%) dropouts

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes stated were reported

Other bias

Low risk

Comment: Appears to be free of other bias

Methods

Design: Randomised within‐subjects study

Participants

Health condition: Adults with spinal cord injury

Sample size: Experimental group: 20 legs, Control group: 20 legs

Setting, Country: Inpatient rehabilitation unit, Australia

Joint of interest: Ankle

Inclusion criteria:

• Sustained a spinal cord injury within the past 12 months

• Commenced sitting out of bed

• Less than grade 2/5 strength in the lower limbs

Exclusion criteria:

• History of trauma to the pelvis or legs

• Unable to tolerate standing

• Likely to be discharged from hospital within 3 months

• Thought unlikely to co‐operate

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 34 years (15), Control group: 34 years (15)

Gender: Experimental group: 20% female, Control group: 20% female

Interventions

Groups included in this review:

Experimental group: Weight‐bearing and stretch

Participants were stood on a tilt table with a 15° wedge on a high block placed under the experimental foot

Total stretch time: 30 min x 3 d x 12 weeks = 18 h over a 12‐week period

Control group: Non weight‐bearing and non stretch

Participants were stood on a tilt table but with nothing placed underneath the control foot

Other groups: Nil

Outcomes

Outcomes included in this review:

• Passive ankle dorsiflexion (torque controlled)

Other outcomes: Total proximal femur bone mineral density, total proximal femur bone mineral density (% initial), total proximal femur bone mineral density (% loss of control)

Time points included in this review: Outcomes measured at 12 weeks (≥ 24 h after last intervention)

Other time points: Outcomes also measured at baseline

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "...computer‐generated random allocation schedule", p 253

Allocation concealment (selection bias)

Low risk

Quote: "...allocations were placed in sealed, opaque, sequentially numbered envelopes. The envelopes were not opened until after the initial tests had been performed", p 253

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...measurements were taken...by an independent ..", p 253

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: No dropouts

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults post‐radiation therapy for the jaw

Sample size: Experimental group: 9, Control group: 5, Other group: 7

Setting, Country: Oral and maxillofacial surgery clinic, USA

Joint of interest: Mandibular

Inclusion criteria:

• Decreased inter‐incisal opening secondary to radiation therapy

• Maximum inter‐incisal opening of ≤ 30 mm

Exclusion criteria:

• > 5 years since undergoing radiation therapy

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture

Mean age (SD): Experimental group: 51 years (14), Control group: 62 years (9), Other group: 59 years (8)

Gender: Experimental group: 33% female, Control group: 40% female, Other group: 0% female

Interventions

Groups included in this review:

Experimental group: Therabite System plus unassisted exercise (Group 3)

Participants used the Therabite System to sustain a maximum comfortable stretch of the jaw. Also performed 10 cycles/d of unassisted exercise ‐ opening to maximal inter‐incisal distance, closing, then moving maximally to the left and right and protrusively.

Total stretch time: 5 x 30 s x (6‐10 sessions) x 7 d x 10 weeks = 17.5 h‐29.2 h over a 10‐week period

Control group: Unassisted exercise (Group 1)

Participants performed 10 cycles/d of unassisted exercise ‐ opening to maximal inter‐incisal distance, closing, then moving maximally to the left and right and protrusively

Other group: Stacked tongue depressors plus unassisted exercise (Group 2)

Participants used stacked tongue depressors to maximally open the mouth. Also performed 10 cycles/d of unassisted exercise ‐ opening to maximal inter‐incisal distance, closing, then moving maximally to the left and right and protrusively

Total stretch time: 5 x 30 s x (6‐10 sessions) x 7 d x 10 weeks = 17.5 h‐29.2 h over a 10‐week period

Outcomes

Outcomes included in this review:

• Maximal incisal opening (mm)

• Pain rating (scale not reported)

• Subjective well‐being (scale not reported)

Other outcomes: Subjective rating of ROM, lateral jaw movements, protrusive jaw movements

Time points included in this review: Outcomes measured at 10 weeks (end of intervention)

Other time points: Outcomes also measured at baseline, 2 weeks, 4 weeks, 6 weeks and 8 weeks

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote: "...patients were randomly assigned to one of three groups", p 864

Comment: Insufficient detail reported

Allocation concealment (selection bias)

Unclear risk

Comment: Insufficient detail reported

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Unclear risk

Comment: Insufficient detail reported

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: No dropouts reported

Selective reporting (reporting bias)

High risk

Comment: Lateral and protrusive jaw movements, pain, subjective ROM, and subjective well‐being all listed as outcomes in the methods but no data reported

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with ankylosing spondylitis

Sample size: Experimental group: 27, Control group: 12

Setting, Country: Inpatient hospital, UK

Joint of interest: Hip

Inclusion criteria:

• Typical radiological features of ankylosing spondylitis

• No previous hip surgery

Exclusion criteria: Nil reported

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture

Mean age (SD): Experimental group: not reported, Control group: not reported

Gender: Not reported

Interventions

Groups included in this review:

Experimental group: Stretch plus conventional care

Participants received cycles of 3 contract relax stretches to the hip muscles

Total stretch time: not reported

Control group: Conventional care

Participants received active exercises in gymnasium and hydrotherapy pool to increase strength and joint mobility

Other groups: Nil

Outcomes

Outcomes included in this review:

• Hip extension with knee in extension (degrees)

Other outcomes: Hip flexion, hip extension with knee in flexion, single leg abduction, bimalleolar abduction, medial hip rotation, lateral hip rotation

Time points included in this review: Outcomes measured at 15 days (end of intervention), 6 months following end of intervention

Other time points: Outcomes also measured at baseline

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote: "were allocated at random ... in blocks of nine to give two in the treatment group for every one control", p 40

Allocation concealment (selection bias)

Unclear risk

Comment: Insufficient detail reported

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...measurements were recorded by an independent assessor who did not know to which group the patients had been allocated", p 40‐1

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

High risk

Comment: No dropouts for 3‐week data, 7/39 (18%) dropouts at 6 months No data reported for 6 months

Selective reporting (reporting bias)

High risk

Comment: 6‐month joint mobility data were not reported

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with stroke

Sample size: Experimental group: 31, Control group: 16

Setting, Country: Inpatient rehabilitation unit, Switzerland

Joint of interest: Wrist

Inclusion criteria:

• Admitted for intensive rehabilitation

• No previous stroke

• Severe paresis of the upper limb ‐ FMA upper‐extremity motor score ≤ 45 points

• Sufficient comprehension to participate in trial as assessed by speech therapist

Exclusion criteria:

• Traumatic injuries

• Rheumatic co‐morbidities

• Lesion of the peripheral nervous system

• Other lesions of the central nervous system

• Lymphoedema

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 68 years (12), Control group: 64 years (14)

Gender: Experimental group: 60% female, Control group: 67% female

Interventions

Groups included in this review:

Experimental group: Orthosis plus conventional care

Participants were issued a thermoplastic customised wrist splint made following biomechanical principles. The wrist was maintained in a neutral position.

Total stretch time: not reported

Control group: Conventional care

2 sessions of physical therapy/d, 1 session of occupational therapy/d, and, if indicated, neuropsychologic and speech therapy

Other groups: Nil

Outcomes

Outcomes included in this review:

• Wrist ROM (FMA sub‐scale)

• Pain (VAS)

• Modified Ashworth scale

Other outcomes: FMA sub‐scale for ROM of forearm, FMA sub‐scale for ROM of fingers, hand oedema, participant satisfaction with splint

Time points included in this review: Outcomes measured at 13 weeks (end of intervention)

Other time points: Outcomes also measured at baseline

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "...allocation schedule was computer generated", p 1858

Allocation concealment (selection bias)

Low risk

Quote: "concealed in opaque, consecutively numbered sealed envelopes by a person not otherwise involved in the study", p 1858

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "independent blinded assessor however, complete blinding of the assessor to the group assignment proved to be difficult in practice because some patients would spontaneously comment on their splint type", p 1858

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: 4/31 (13%) dropouts at 13 weeks

Selective reporting (reporting bias)

High risk

Comment: Insufficient detail reported to include in meta‐analysis

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults following surgical release for Dupuytren’s contracture

Sample size: Experimental group: 26, Control group: 30

Setting, Country: Hand therapy clinic, New Zealand

Joint of interest: Hand

Inclusion criteria:

• Surgical release of Dupuytren contracture (any surgery type)

• Attended their first postoperative hand therapy appointment within 14 d after surgery

Exclusion criteria:

• K‐wiring of the proximal interphalangeal joint during surgery

• Inability to comply with hand therapy

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture

Mean age (SD): Experimental group: 68 years (8), Control group: 67 years (9)

Gender: Experimental group: 15% female, Control group: 23% female

Interventions

Groups included in this review:

Experimental group : Night extension orthosis plus hand therapy

Participants wore each night a thermoplastic orthosis that was custom‐fabricated (moulded on the dorsum of the hand holding the operated fingers in maximal comfortable extension without placing undue tension on the wound). The orthosis was adjusted to apply greater extension force to the operated fingers if the therapist deemed this necessary. Participants also received hand therapy (details below)

Total stretch time: 8 h x 7 d x 12 weeks = 672 h over a 3‐month period

Control group : Hand therapy alone

Participants received a standard hand therapy programme delivered by an occupational therapist, physiotherapist or hand therapist, which could include active tendon gliding ROM exercises, education, wound care, oedema management, scar management, graded return to usual daily activities, passive stretch with or without heat to increase finger extension and/or flexion, intermittent use of daytime finger‐based dynamic proximal interphalangeal joint extension orthoses, and grip strengthening

Other groups : Nil

Outcomes

Outcomes included in this review:

• Active extension of the little finger (sum of metacarpophalangeal, proximal interphalangeal and distal interphalangeal joints; degrees)

• Disabilities of the Arm, Shoulder and Hand Outcome Measure (DASH)

Other outcomes:

Active extension of each operated finger, active flexion of each operated finger, distal palmar crease of each operated finger, grip strength of left and right hand

Time points included in this review: Outcomes measured at 3 months (end of intervention)

Other time points: Outcomes also measured at before surgery, at the first postoperative hand therapy visit and 6 weeks

Notes

Assumed participants wore the splint for 8 h per night when calculating total stretch time

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "Participants were randomly allocated to 1 of 2 treatment groups ... This occurred at the first postoperative hand therapy appointment by the participant selecting a tag from an envelope with group allocation concealed", p 1286

Allocation concealment (selection bias)

Low risk

Quote: "Participants were randomly allocated to 1 of 2 treatment groups ... This occurred at the first postoperative hand therapy appointment by the participant selecting a tag from an envelope [LH1] with group allocation concealed", p 1286

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Unclear risk

Quote: "1 therapist took nearly all of the measurements. When she was unavailable, 2 other therapists, trained by the first to measure uniformly, filled in", p 1287
Comment: Insufficient detail reported

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: 3/56 (5%) dropouts at 6 weeks and 2/56 (4%) dropouts at 3 months

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

High risk

Comment: Protocol allowed "rescue". Also a unit of analysis issue. Analysed joints

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with acquired brain injury

Sample size: Experimental group: 6, Control group: 4

Setting, Country: Brain injury and geriatric assessment/rehabilitation units of a major metropolitan hospital, Australia

Joint of interest: Wrist and fingers

Inclusion criteria:

• Aged 18‐80 years

• At least 2 months since acquired brain injury

• Moderate stiffness in the wrist and/or hand flexor muscles of the affected upper limb/s with a Modified Ashworth Scale rating of 1+ or 2

• Presence of spasticity in the wrist or finger flexor muscles as indicated by a muscle reactivity rating of at least 2 on the Modified Tardieu Scale

• No soft tissue contracture in wrist or finger flexor muscles as indicated by the Modified Tardieu Scale

Exclusion criteria:

• Cognitive or behavioural deficits that prevented the provision of informed consent

• Cognitive or behavioural deficits that prevented active participation in an upper limb therapy programme

Existing contracture, at risk of contracture, or combination of both: Participants were at risk of developing contracture

Mean age (SD): Experimental group: 40 years (16), Control group: 54 years (6)

Gender: Experimental group: 33% female, Control group: 50% female

Interventions

Groups included in this review:

Experimental group : Splint and standard practice occupational therapy programme

Participants wore an individualised, thermoplastic resting mitt splint designed to approximate the standard resting position (20° wrist extension) but tailored to place each participant’s hypertonic muscle groups on low load, prolonged stretch. The splint was worn for 2‐4 h during the day and overnight. Participants also received an occupational therapy programme (details below)

Total stretch time: 10 h x 90 d (3 months) = 900 h over a 3‐month period

Control group : Standard practice occupational therapy programme only

Participants received a standard practice occupational therapy programme as typically provided to people with upper limb hypertonicity (various combinations of movement training, stretches and functional splinting).

Other groups : Nil

Outcomes

Outcomes included in this review:

• Wrist extension with the fingers extended (degrees)

• Finger flexor spasticity (Modified Tardieu Scale)

Other outcomes:

Wrist extension with the fingers flexed, wrist flexor spasticity, wrist flexor muscle stiffness, finger flexor muscle stiffness.

Time points included in this review: Outcomes measured at 3 months (end of intervention period) and 4 months

Other time points: Outcomes also measured at baseline, 1 month and 2 months

Notes

Assumed participants wore the splint for 10 h per day when calculating total stretch time

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "A random number table was generated by an independent researcher and used to allocate participants to control (no‐splint) and experimental (splint) groups", p 888

Allocation concealment (selection bias)

Unclear risk

Quote: "A random number table was generated by an independent researcher and used to allocate participants to control (no‐splint) and experimental (splint) groups", p 888

Comment: Insufficient detail reported

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "Measures were completed by a blinded assessor", p 888

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

High risk

Comment: 3/10 (30%) dropouts

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes stated were reported

Other bias

High risk

Comment: 3 people were included in ITT analysis but not clear how this was done

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with oral submucous fibrosis

Sample size: Experimental group: 54, Control group: 23

Setting, Country: Hospital, Nepal

Joint of interest: Jaw/mouth

Inclusion criteria:

• Confirmed oral submucous fibrosis by biopsy

• Subjectively reduced oral opening

Exclusion criteria:

• Oral squamous cell carcinoma

• Severely restricted oral opening that required surgical treatment

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture

Mean age (SD): Experimental group: 36 years (15), Control group: 35 years (13), Other group: 44 years (19)

Gender: Experimental group: 30% female, Control group: 30% female, Other group: 10% female

Interventions

Groups included in this review:

Experimental group: Physiotherapy (stacked tongue depressors) plus conventional care

Participants undertook jaw exercises 5 x d in which tongue spatulas were positioned passively between anterior teeth, spatula number determined by comfortable maximal oral opening. The jaws were opened 5 times in each session, and held in position with the teeth resting on the spatulas for 1 min on each occasion. An additional spatula was added every fifth day unless this caused pain in which case the additional spatula was added on the tenth day. Participants also received conventional care.

Total stretch time: 5 min x 5 sessions x 7 d x 17 weeks = 2975 min = 49.6 h over a 17‐week period

Control group: Conventional care

Participants were recommended to cease areca nut use, given dietary advice and received conventional care

Other groups: Hyaluronidase and steroid injections plus conventional care

Participants received bi‐weekly submucosal injections over 4 weeks of hyaluronidase (1500 units) and hydrocortisone (100 mg)

Outcomes

Outcomes included in this review:

• Maximal inter‐incisal opening (mm)

• Mucosal pain (absent, stimulated by eating, spontaneous, constant)

Other outcomes: Reported areca nut use, progressive involvement of oral mucosa

Time points included in this review: Outcomes measured at 4 months (end of intervention)

Other time points: Outcomes also measured at baseline

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "Random numbers were used for assignation", p 221

Allocation concealment (selection bias)

Unclear risk

Comment: Insufficient detail reported

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Unclear risk

Comment: Insufficient detail reported

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

High risk

Comment: 26/54 (48%) dropouts at 4 months

Selective reporting (reporting bias)

High risk

Comment: Insufficient detail reported to include in meta‐analysis

Other bias

High risk

Quote: "patients unable to attend bi‐weekly injection were assigned for physiotherapy with the next subject assigned for injection"; "control and injection enrolment ceased for ethical reasons when sufficient control patients returned, and injection was recognized as having poor outcomes", p 221

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with spinal cord injury

Sample size: Experimental group: 18, Control group: 21

Setting, Country: Acute hospital, Canada

Joint of interest: Shoulder

Inclusion criteria:

• Sustained traumatic spinal cord lesion at or above the C8 level

• Subjects with incomplete lesions were required to have some degree of motor deficit

Exclusion criteria:

• Sustained fracture(s) scapula, clavicle or acromial head at the time of trauma

• Required shoulder immobilisation for any reason following their accident

Existing contracture, at risk of contracture, or combination of both: Participants were at risk of developing contracture

Mean age (SD): Experimental group: 34 years (15), Control group: 44 years (19)

Gender: Experimental group: 11% female, Control group: 10% female

Interventions

Groups included in this review:

Experimental group: Positioning plus conventional care (Group 2)

Participants received 2 sessions of shoulder positioning:
Position 1: participants in supine with their arms placed on padded supporting boards, shoulders abducted to 90° and elbows extended for 30 min
Position 2: participants in supine with their shoulders positioned on pillows in 180° flexion and lateral rotation for 15 min
If the positions were not tolerated, shorter durations were applied and slowly increased. Participants also received full passive movements on their upper limbs (either passive, active assisted, active or resisted), scapula stretches, modalities and medications as required for shoulder pain

Total stretch time: 45 min x 5 d x (2‐16 weeks) = 7.5 h‐60 h over a 2‐16‐week period

Control group: Conventional care (Group 1)

Participants received full passive movements on their upper limbs (either passive, active assisted, active or resisted), scapula stretches, modalities and medications as required for shoulder pain.

Other groups: Nil

Outcomes

Outcomes included in this review:

• Passive shoulder abduction (right arm; degrees)

• Pain during preceding 24 h (right shoulder; VAS)

• Functional Independence Measure

Other outcomes: Passive shoulder abduction (left arm), passive shoulder flexion (right arm), passive shoulder flexion (left arm), passive shoulder medial rotation (right arm), passive shoulder medial rotation (left arm), passive shoulder lateral rotation (right arm), passive shoulder lateral rotation (left arm), pain during preceding 24 h (left shoulder), hours sitting in chair.

Time points included in this review: Outcomes measured at 2 weeks¹

Other time points: Outcomes also measured at baseline, week 1, week 3, week 4, week 5, week 6, week 7, week 8, week 9, week 10, week 11 and week 12

Notes

¹The intervention was ceased early with some participants (from after week 2) while others were treated up until week 12. We included outcomes from week 2 as all participants received at least 2 weeks of stretch

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "...patients were randomly assigned (using a random number generator...)", p 268

Allocation concealment (selection bias)

Low risk

Quote: "…and a system of sealed envelopes", p 268

Comment: Insufficient detail reported in paper. Author correspondence revealed that a system of sequentially‐numbered, sealed, opaque envelopes was used to conceal allocation

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...data were collected by a single therapist at each site who was blinded to the treatment allocation of the patient", p 269

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Unclear risk

Comment: Length of intervention was different for participants, determined by when they were transferred to another facility. Insufficient detail reported to accurately determine dropouts

Selective reporting (reporting bias)

High risk

Comment: Insufficient detail reported to include in meta‐analysis

Other bias

High risk

Quote: "...the trial was terminated with 39 subjects after 3 years of data collection", p 272

Comment 1: possible cause of bias introduced by early termination of study

Comment 2: No standard treatment protocol for participants as they were given varying amounts of treatment dependent on length of stay

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with stroke

Sample size: Experimental group: 10, Control group: 9

Setting, Country: Rehabilitation unit, Netherlands

Joint of interest: Shoulder

Inclusion criteria:

• First ever stroke and maximum of 12 weeks post stroke

• Medial cerebral artery stroke

• No premorbid impairments of the affected arm

• No severe shoulder pain

• No use of anti‐spasticity drugs

• No use of pain‐reducing drugs except for paracetamol

• No planned date of discharge

• Able to give written informed consent

Exclusion criteria:

• Subjects with fair to good recovery of the arm (Brunnstrom stages 4, 5 or 6)

• Severe neglect (score of greater than 3 zeros on letter cancellation test)

• Severe loss of position sense (scores 2 and 3 on thumb finding test)

• Cognitive impairment (less than 23 on Mini‐Mental State Examination)

• Able to prevent contracture by producing voluntary movement (FMA > 18 on the shoulder/elbow/forearm sub‐scales)

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 53 years (10.2)¹, Control group: 52 years (8.8)¹

Gender: Experimental group: 33% female¹, Control group: 63% female¹

Interventions

Groups included in this review:

Experimental group: Positioning plus conventional care

Participant was positioned in supine with arm in maximal shoulder abduction, shoulder external rotation, elbow extension and supination of the forearm that could be tolerated without any pain. The arm was always supported by a pillow and, if necessary, held in position with a sandbag. Participants also received conventional rehabilitation

Total stretch time: 30 min x 2 sessions x 5 d x (5‐10 weeks) = 25 h‐50 h over a 5‐10‐week period

Control group: Conventional care

Participants received conventional rehabilitation

Other groups: Nil

Outcomes

Outcomes included in this review:

• Passive shoulder abduction (degrees)

• Pain (yes/no)

• Spasticity (Ashworth scale)

• Arm motor performance (FMA)

Other outcomes: Passive shoulder flexion, passive shoulder external rotation, passive elbow extension, passive forearm supination, Barthel Index

Time points included in this review: Outcomes measured at 10 weeks (end of intervention).

Other time points: Outcomes also measured at baseline and 5 weeks

Notes

¹Data obtained via correspondence with study author

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote:"...an independent person carried out the randomization procedure. The envelopes were shuffled and drawn blindfolded", p 658

Allocation concealment (selection bias)

Low risk

Quote: "...subjects were randomly assigned to one of the two groups using opaque, sealed envelopes...The envelopes were shuffled and drawn blindfolded", p 658

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...the same two raters, unaware of group allocation and not involved in the treatment of subjects, carried out all the measurements", p 658

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

High risk

Comment: 2/19 (11%) dropouts at 5‐week outcome assessment, 9/19 (47%) dropouts at 10‐week outcome assessment

Selective reporting (reporting bias)

High risk

Comment: Insufficient detail reported on pain to include in meta‐analysis

Other bias

High risk

Quote: "...after nearly two years the trial had to be terminated because of set time limits, leaving only 19 subjects who met all inclusion criteria" p 663

Comment 1: Possible cause of bias introduced by early termination of study

Comment 2: Unclear whether the protocol was for a 10‐week or 5‐week study

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with stroke

Sample size: Experimental group: 14, Control group: 14

Setting, Country: Inpatient rehabilitation unit, Australia

Joint of interest: Shoulder

Inclusion criteria:

• Less than 10 weeks from the onset of stroke

• Score of less than 5 on the upper‐arm function item of the MAS for stroke

• No pre‐morbid shoulder pain

• No premorbid restriction of shoulder movement

• Passive range of shoulder abduction and flexion greater than 90°

• Able to comprehend and use a VAS for pain

Exclusion criteria:

• Subjects with a brainstem stroke

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 58 years (13), Control group: 58 years (11)

Gender: Experimental group: 50% female, Control group: 15% female

Interventions

Groups included in this review:

Experimental group: Shoulder positioning plus conventional care

Participants received 3 x 20 min sessions of shoulder positioning:
Position 1: lying supine, shoulder in maximum tolerable abduction and external rotation, and elbow flexed
Position 2: lying supine, shoulder abduction to 90°, maximum tolerable external rotation, and elbow flexed
Position 3: sitting, shoulder forward flexed 90°, elbow extension, wrist extension, and a cylinder in hand to provide a web space stretch

Participants also received active training of reaching and manipulation tasks

Total stretch time: 3 sessions x 20 min x 5 d x 6 weeks = 30 h over a 6‐week period

Control group: Conventional care

Participants received active training of reaching and manipulation tasks. No formal stretches were applied to the shoulder joint complex

Other groups: Nil

Outcomes

Outcomes included in this review:

• Passive shoulder external rotation (degrees)

• Pain at rest (VAS)

Other outcomes: Active shoulder abduction, pain on dressing

Time points included in this review: Outcomes measured at 6 weeks (end of intervention)

Other time points: Outcomes also measured at baseline

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote:"...random number tables to determine the subject’s group allocation", p 36

Allocation concealment (selection bias)

Low risk

Quote: "...group allocation was completed by a person independent of the recruitment process...the recruiter telephoned another person", p 36

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists.

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...measurements were made by an assessor who was blinded to the subject s group allocation", p 37

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

High risk

Comment: 5/28 (18%) dropouts, with four from experimental group

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with spinal cord injury

Sample size: Experimental group: 7, Control group: 6

Setting, Country: Rehabilitation unit, USA

Joint of interest: Hand

Inclusion criteria:

• Not reported although study involved only people with C6 tetraplegia

Exclusion criteria: Nil reported

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (range): Experimental group: not reported, Control group: not reported, both groups: 26 years (18‐42)

Gender: Experimental group: not reported, Control group: not reported, both groups: 8% female

Interventions

Groups included in this review:

Experimental group: Positional orthosis plus conventional rehabilitation

Participants were issued a short opponens or long opponens orthosis, depending on the strength of their wrist extensors. Both orthoses maintained the distal transverse arch and the thumb web space in 35° of CMC abduction, the metacarpophalangeal joint in full extension, and the interphalangeal joint in slight flexion. Participants also received conventional rehabilitation

Total stretch time: 8 h x 7 d x 12 weeks = 672 h over a 12‐week period

Control group: Conventional rehabilitation

Participants received conventional rehabilitation

Other groups: Nil

Outcomes

Outcomes included in this review:

• Passive metacarpophalangeal (MCP) extension

• Jebsens hand test sub‐item ‐ simulated feeding (seconds)

Other outcomes: Passive MCP flexion, passive proximal interphalangeal (PIP) extension, passive PIP flexion, passive distal interphalangeal (DIP) extension, passive DIP flexion, size of opening the hand when releasing, size of closing the hand with tenodesis, Jebsen hand test ‐ 6 other sub‐items, thumb/finger opposition, palmar abduction, passive lateral prehension grasp, wrist extensor strength

Time points included in this review: Outcomes measured at 12 weeks (end of intervention)

Other time points: Outcomes also measured at baseline, 4 weeks and 8 weeks

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote: "...subjects were randomly assigned", p 140

Comment: Insufficient detail reported

Allocation concealment (selection bias)

Unclear risk

Comment: Insufficient detail reported

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Unclear risk

Comment: Insufficient detail reported

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

High risk

Comment: 4/13 (31%) dropouts

Selective reporting (reporting bias)

High risk

Comment: Not all pre‐stated outcomes were reported

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised cross‐over study

Participants

Health condition: Elderly nursing‐home residents

Sample size: Experimental group: 9, Control group: 9

Setting, Country: Chronic care hospital, Canada

Joint of interest: Knee

Inclusion criteria:

• No plans for discharge within 6 months

• Knee flexion contracture of 10° or greater in at least one leg

• Able to tolerate a bed positioning programme and ongoing assessments without severe pain

Exclusion criteria:

• Behavioural problems that prevented adherence to the programme

• Receiving the medication baclofen at the time of recruitment

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture

Mean age (range): Experimental group: not reported, Control group: not reported, both groups: 82 years (71‐93)

Gender: Experimental group: not reported, Control group: not reported, both groups: 63% female

Interventions

Groups included in this review:

Experimental group: Bed positioning programme (low‐load prolonged knee stretch) ¹

Participants were positioned in supine with their knee extended as much as possible. The position was maintained using bed sheets secured under the mattress

Total stretch time: 40 min x 4 d x 8 weeks = 21.3 h over an 8‐week period

Control group: No intervention ¹

Participants received no intervention

Other groups : Nil

Outcomes

Outcomes included in this review:

• Passive knee extension (degrees)

• Level of pain (rated by assessor)

Other outcomes: Nil

Time points included in this review: Outcomes measuring combined effect after 8 weeks of stretch (both cross‐over periods combined)

Other time points: Outcomes also measured at baseline, 1 week , 2 weeks, 3 weeks, 4 weeks, 5 weeks, 6 weeks, 7 weeks, 8 weeks (end of first cross‐over period), 9 weeks, 10 weeks, 11 weeks, 12 weeks, 13 weeks, 14 weeks, 15 weeks and 16 weeks (end of second cross‐over period)

Notes

¹Only includes details of the first period of the cross‐over

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "...randomly assigned to 2 groups by a random numbers table", p 365

Allocation concealment (selection bias)

Unclear risk

Comment: Insufficient detail reported

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...a single rater blinded to the intervention assessed the participants", p 366

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

High risk

Comment: 6/18 (33%) dropouts

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

High risk

Comment: One participant's group allocation was changed to create even group numbers

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with stroke

Sample size: Experimental group: 17, Control group: 17

Setting, Country: Inpatient rehabilitation hospital, Australia

Joint of interest: Shoulder

Inclusion criteria:

• Admitted for rehabilitation following first time stroke

• No previous history of neurological disease

• Pain in or injury to the affected shoulder

• At least 45° of passive abduction but less than full active flexion in the affected shoulder

Exclusion criteria:

• Complex medical situation

• Not admitted for active rehabilitation

• More than 100 days from time of stroke to admission to rehabilitation

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture

Mean age (SD): Experimental group: 66 years (16), Control group: 67 years (14)

Gender: Experimental group: 41% female¹, Control group: 40% female¹

Interventions

Groups included in this review:

Experimental group: Shoulder positioning plus conventional care

Participants received 2 x 20 min sessions of shoulder positioning:
Position 1: sitting with affected shoulder abducted to 90° and fully supported on the surface of a table with the elbow extended and forearm in neutral
Position 2: lying in supine with affected shoulder abducted to 90° and in the maximal amount of achievable external rotation, elbow flexed and forearm pronated

Participants also received an additional shoulder positioning programme for remainder of days during the intervention period:
In sitting: arm positioned on a custom armrest in 10°‐15° of shoulder abduction and midway between shoulder external and internal rotation
In bed: a pillow was used to support the stroke‐affected shoulder in a position midway between external and internal rotation and not horizontally adducted

Participants also received 30 min upper limb therapy

Total stretch time: 24 h x 30 d² = 720 h over a 30‐d period

Control group: Conventional care

Participants received 30 min upper limb therapy. Participants also used locally fabricated cushion supports for their stroke‐affected upper limb when seated and in bed

Other groups: Nil

Outcomes

Outcomes included in this review:

• Passive shoulder external rotation (degrees)

• Hemiplegic shoulder pain at rest over previous 24 h (VAS)

• Functional independence (Modified Barthel Index)

Other outcomes: Hemiplegic shoulder pain during assessment (Ritchie Articular Index), hemiplegic shoulder pain with movement (VAS), MAS for stroke

Time points included in this review: Outcomes measured at discharge (end of intervention) and 6 months following discharge

Other time points: Outcomes also measured at baseline

Notes

¹Data obtained via correspondence with study author

²Length of intervention was calculated as an average of 30 days for the intervention and control groups

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "...referred to a random number table to identify the predetermined, random allocation", p 279

Allocation concealment (selection bias)

Low risk

Quote: "...once consent was obtained, the primary investigator referred to a random number table to identify the predetermined, random allocation of that participant to either the treatment or comparison group", p 279

Comment: Author correspondence revealed that central allocation was used. The person recruiting participants did not have access to the random number table

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

High risk

Quote: "...a blinded assessor completed the measurement of the dependent variables at admission and discharge from rehabilitation", p 279
Quote: "...follow‐up assessments were completed by the principal investigator", p 163 in follow‐up paper
Comment: Blinded assessor for discharge outcomes. Non‐blinded assessor for 6 month follow‐up outcomes

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: 2/17 (12%) dropouts in control group, no dropouts in experimental group

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

Unclear risk

Comment: It was identified that 38 people would be needed in the power analysis but only 34 were recruited. Author correspondence revealed that the study was stopped due to participant recruitment difficulties

Methods

Design: Randomised within‐subjects study

Participants

Health condition: Adults with spinal cord injury

Sample size: Experimental group: 14 legs, Control group: 14 legs

Setting, Country: 2 spinal injury rehabilitation units, Australia

Joint of interest: Ankle

Inclusion criteria:

• Participating in a rehabilitation programme

• Sustained a spinal cord injury within the preceding year

• Have not more than grade 1 of 5 motor strength around both ankles

• Be willing to cease assisted‐standing and all passive exercises and stretches to their ankles for the duration of the study

Exclusion criteria:

• Pressure sores on their heels that prevented stretching or testing

• Considered unlikely to co‐operate

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 36 years (16), Control group: 36 years (16)

Gender: Experimental group: 0% female, Control group: 0% female

Interventions

Groups included in this review:

Experimental group: Stretch

Participants received a constant stretch on the experimental ankle into dorsiflexion with the knee extended using a purpose‐built device.

Total stretch time: 30 minutes x (5 ‐ 7 days) x 4 weeks = 10 hours to 14 hours over a 4‐week period.

Control group: Non‐stretch

Participants did not receive any type of manual therapy to either ankle nor did they stand or walk.

Other groups: Nil

Outcomes

Outcomes included in this review:

• Ankle angle at 10 Nm torque with the knee extended (degrees)

Other outcomes: Ankle angle at 10 Nm torque with the knee flexed, ankle mobility with knee extended (slope of torque/angle curve), ankle mobility with knee flexed (slope of torque/angle curve), baseline ankle angle with knee extended, baseline ankle angle with knee flexed

Time points included in this review: Outcomes measured at 4 weeks (end of intervention)

Other time points: Outcomes also measured at baseline, 2 weeks and 5 weeks

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "...a computer generated random allocation schedule was determined before the study by an investigator who was not involved in patient recruitment or group allocation", p 1342

Allocation concealment (selection bias)

Low risk

Quote: "...allocations were placed in sealed, opaque, sequentially numbered envelopes by an investigator who was not involved in determining eligibility for the trial. The envelopes were not opened until after the initial tests had been performed", p 1342

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...a blinded therapist was responsible for all measurements", p 1344

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: No dropouts

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised within‐subjects study

Participants

Health condition: Adults with spinal cord injury

Sample size: Experimental group: 16 legs , Control group: 16 legs

Setting, Country: 2 spinal injury rehabilitation units, Australia

Joint of interest: Hip

Inclusion criteria:

• Sustained a spinal cord injury within the past 12 months

• Commenced sitting out of bed following the initial injury

• Less than 110° passive hip flexion with the knee extended

Exclusion criteria:

• More than grade 2/5 motor strength in the muscles around the hips and knees

• Unlikely to remain in the unit for 4 weeks

• History of trauma to the pelvis or upper leg

• Unable to tolerate stretch

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 33 years (15), Control group: 33 years (15)

Gender: Not reported

Interventions

Groups included in this review:

Experimental group: Stretch

Participants received a stretch to the hamstring muscles with a 30 Nm torque using a purpose‐built device. Participants also performed normal activities of daily living

Total stretch time: 30 min x 5 d x 4 weeks = 10 h over a 4‐week period

Control group: Non‐stretch

Participants did not receive any stretches to the hamstring muscles

Participants performed normal activities of daily living

Other groups: Nil

Outcomes

Outcomes included in this review:

• Hip flexion at 30 Nm torque (degrees)

Other outcomes: Nil

Time points included in this review: Outcomes measured at 4 weeks (end of intervention)

Other time points: Outcomes also measured at baseline

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "...a computer‐generated random allocation schedule was produced prior to the study by one of the authors who was not otherwise involved in subject recruitment or allocation", p 178

Allocation concealment (selection bias)

Low risk

Quote: "...to ensure concealment, the same person placed allocations in sealed, opaque, sequentially‐numbered envelopes. The envelopes were not opened until after the initial tests had been performed", p 178

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...measurements were taken...by an independent therapist who was blinded to allocation", p179

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: No dropouts

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised within‐subjects and parallel‐group study

Participants

Health condition: Adults with spinal cord injury, stroke or traumatic brain injury

Sample size: Total: Experimental group: 30 thumbs, Control group: 30 thumbs

Spinal cord injury¹: Experimental group: 19 thumbs, Control group: 20 thumbs

Stroke²: Experimental group: 7 thumbs, Control group: 7 thumbs

Traumatic brain injury³: Experimental group: 4 thumbs, Control group: 3 thumbs

Setting, Country: Community participants, Australia

Joint of interest: Thumb carpometacarpal

Inclusion criteria:

• Sustained a cervical spinal cord injury, traumatic brain injury or stroke that affected one or both upper limbs

• Had a contracture of their thumb web‐space as assessed by clinical examination

Exclusion criteria:

• Had a contracture deemed unlikely to respond to stretch

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture

Mean age (range):Unilateral participants: Experimental group: 58 years (49‐67), Control group: 64 years (50‐71)
Bilateral participants: Experimental group: 47 years (37‐51), Control group: 47 years (37‐51)

Gender: Experimental group: 13% female, Control group: 30% female

Interventions

Groups included in this review:

Experimental group: Thumb splint

Participants' thumbs were stretched by splinting them into abduction. One of two splints was used:
Splint 1: volar splint with a C‐bar to position the thumb into palmar abduction
Splint 2: cone splint used where it was difficult to obtain a good stretch with the thumb C‐bar piece
Splints were reviewed at week 1, week 4 and week 8 after baseline

Participants were also instructed to refrain from self‐administering any other stretch

Total stretch time: 8 h x 7 d x 12 weeks = 672 h over a 12‐week period

Control group: No splint

Participants received no intervention. Participants were instructed to refrain from self‐administering any stretch.

Other groups: Nil

Outcomes

Outcomes included in this review:

• Palmar abduction of the thumb carpometacarpal joint (degrees)

• The effect of the splinting regime on self selected goals (Canadian Outcome Performance Measure)

Other outcomes: Questionnaire on participants’ attitudes towards the effectiveness and convenience of the splinting regime

Time points included in this review: Outcomes measured at 12 weeks (end of intervention)

Other time points: Outcomes also measured at baseline

Notes

¹Spinal cord injury subgroup of Harvey 2006 study; ²Stroke subgroup of Harvey 2006 study; ³Traumatic brain injury subgroup of Harvey 2006 study; data obtained via correspondence with study author

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "...an independent person used a computer to generate the random allocation schedules", p 252

Allocation concealment (selection bias)

Low risk

Quote: "...these were placed in opaque, sequentially numbered envelopes which were sealed and kept off site", p 252

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...the assessors were blinded to participant allocation and participants were asked not to discuss any aspect of the trial with the assessors in order to maintain blinding", p 252

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: 1/60 (2%) dropouts

Selective reporting (reporting bias)

Unclear risk

Comment: Canadian Outcome Performance Measure was discontinued

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised cross‐over study

Participants

Health condition: Adults with brain injury

Sample size: Experimental group: 8¹, Control group: 7¹

Setting, Country: Inpatient rehabilitation hospital, USA

Joint of interest: Elbow and wrist

Inclusion criteria:

• ≥ 8 years old

• Unilateral or bilateral hypertonicity

• Contractures in upper extremities that interfered with function

• ≤ 2 years since injury

• Able to follow simple instructions and participate in self‐care skills

Exclusion criteria:

• Previously treated with casts

• Absent sensation in affected extremity

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture

Mean age (range): Experimental group: 25 years (9‐44), Control group: 32 years (19‐48)

Gender: Not reported

Interventions

Groups included in this review:

Experimental group: Serial casting followed by therapy (Group 1)

Participants wore rigid circular elbow or wrist casts. Casts were re‐applied each 5‐7 d, with 4‐6 casts applied in total. Limbs were positioned 5°‐10° off maximal ROM

Total stretch time: 24 h x 7 d x 4.33 weeks = 728 h over a 4‐week period

Control group: Therapy followed by serial casting (Group 2)

Participants received traditional treatments included passive and active movements, prolonged stretch, splinting, neurophysiological treatment techniques and relaxation techniques

Other groups: Nil

Outcomes

Outcomes included in this review:

• Unidirectional passive joint ROM (degrees)

• Joint angle at which stretch reflex elicited (degrees)

• Observation of performance of functional tasks

Other outcomes: Observation of rapid alternating movements

Time points included in this review: Outcomes measured at 1 month (cross‐over point)

Other time points: Outcomes also measured at baseline, 2 months (end of intervention)

Notes

¹Number of participants who were analysed by the study authors (i.e. these numbers do not include dropouts). Study authors did not report the size of the group allocations at baseline

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

High risk

Quote: "Subjects were alternately assigned", p 220

Allocation concealment (selection bias)

High risk

Quote: "Subjects were alternately assigned", p 220

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "The evaluations were performed by an experienced occupational therapist who was blind to the treatment each patient was receiving", p 220

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

High risk

Comment: 5/20 (25%) dropouts

Selective reporting (reporting bias)

High risk

Comment: Insufficient detail reported to include in meta‐analysis

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults with stroke or stroke‐like brain injury

Sample size: Experimental group: 20, Control group: 20

Setting, Country: Inpatient rehabilitation hospital, Australia

Joint of interest: Wrist

Inclusion criteria:

• Stroke or stroke‐like brain injury (i.e. subarachnoid haemorrhage resulting in hemiplegia, not traumatic head injury or Parkinson's disease)

• 18 years of age or over

• Unable to actively extend the affected wrist past neutral

Exclusion criteria:

• Language, comprehension or reading problems which prevented informed consent

• Co‐existing upper‐limb conditions that directly affected movement

• Not able to participate in upper‐limb rehabilitation

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 61 (21), Control group: 62 (17)

Gender: Experimental group: 70% female, Control group: 35% female

Interventions

Groups included in this review:

Experimental group: Stretch plus usual care

Participants received a weight‐bearing stretch of the arm in sitting, with the shoulder positioned in external rotation, slight abduction and extension, elbow in extension, forearm in supination and wrist and fingers in maximum extension. If unable to do stretch using this method, stretch performed manually or with a stretch board. Participants also received usual upper limb care. No wrist or finger stretches were administered.

Total stretch time: 30 min x 5 d x 4 weeks = 10 h over a 4‐week period

Control group: Usual care

Participants received usual upper limb care. No wrist or finger stretches were administered

Other groups: Nil

Outcomes

Outcomes included in this review:

• Passive wrist extension (degrees)

• Pain at rest at the time of testing (VAS)

• Upper limb activity (composite of 3 items of MAS)

Other outcomes: Nil

Time points included in this review: Outcomes measured at 4 weeks (end of intervention) and 9 weeks (5 weeks after last intervention)

Other time points: Outcomes also measured at baseline and 5 weeks (1 week after last intervention)

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Low risk

Quote: "...computer‐generated randomisation table", p 240

Allocation concealment (selection bias)

Low risk

Quote: "...kept by a person who was remote from the study site and independent of recruitment, and group allocation was revealed by phone call", p 240

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Low risk

Quote: "...outcome measures were collected by therapists...who were blind to group allocation", p 240

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: 2/40 (5%) dropouts at 5 weeks, 3/40 (8%) dropouts at 9 weeks

Selective reporting (reporting bias)

Low risk

Comment: All pre‐stated outcomes were reported

Other bias

Low risk

Comment: Appears free of other bias

Methods

Design: Randomised parallel‐group study

Participants

Health condition: Adults following total knee replacement

Sample size: Experimental group: 27, Control group: 28

Setting, Country: Acute hospital, UK

Joint of interest: Knee

Inclusion criteria:

• Osteoarthritis or rheumatoid arthritis

• Undergoing primary total knee replacement

Exclusion criteria:

• Previous surgery, other than arthroscopy

Existing contracture, at risk of contracture, or combination of both: Participants had existing contracture and were at risk of developing contracture

Mean age (SD): Experimental group: 66 years (14), Control group: 69 years (10)

Gender: Experimental group: 59% female, Control group: 46% female

Interventions

Groups included in this review:

Experimental group: Splint

Participants received a semi‐rigid knee extension splint for the first 48 hours after total knee replacement surgery. Participants also received usual care

Total stretch time: 24 h x 2 d = 48 h over 2 d

Control group: No splint

Participants received no splint after total knee replacement surgery

Participants received usual care.

Other groups: Nil

Outcomes

Outcomes included in this review:

• Knee fixed‐flexion deformity (degrees)

Other outcomes: Knee extension lag, active knee flexion and length of hospital stay

Time points included in this review: Outcomes measured at 2 d (end of intervention) and 3 months (˜ 3 months after last intervention)

Other time points: Outcomes also measured at baseline and 1 week (5 days after last intervention)

Notes

Risk of bias

Bias

Authors' judgement

Support for judgement

Random sequence generation (selection bias)

Unclear risk

Quote:"...patients were randomly assigned to two groups", p 229

Comment: Insufficient detail reported

Allocation concealment (selection bias)

Unclear risk

Quote: "...randomisation was achieved by the closed envelope technique at the time of wound closure, blinding the surgeon to the intended study group until this time", p 229

Comment: Insufficient detail reported

Blinding (performance bias and detection bias)
Therapists

High risk

Comment: Not possible to blind participants or therapists

Blinding of outcome assessors (detection bias) ‐ objective outcomes
All outcomes

Unclear risk

Quote: "...to ensure she would remain blinded to the splint allocation, a second person was trained to take the 48‐h measurements when the splints were still in use", p230
Comment: Second assessor not blinded to splint allocation for outcomes measured at 2 d

Blinding of outcome assessors (detection bias) ‐ self‐reported outcomes
All outcomes

High risk

Comment: Not possible to blind participants

Incomplete outcome data (attrition bias)
All outcomes

Low risk

Comment: 2/55 (4%) dropouts at 3‐month follow‐up

Selective reporting (reporting bias)

High risk

Comment: No data reported for 3‐month follow‐up

Other bias

Unclear risk

Comment: More participants were recruited than original power calculations indicated were necessary. No reason given