Types of studies

Published and unpublished randomised controlled trials and controlled clinical trials will be eligible for inclusion. Cross‐over trials will also be included if participants are randomly allocated to order of intervention (the first period will be analysed as a parallel group trial).

Types of participants

This review will include studies involving participants with existing contracture or those at risk of developing contracture. Participants may be of any age group or either gender. It is anticipated that this review might include participants with:

• neurological conditions (e.g. stroke, multiple sclerosis, spinal cord injury, traumatic brain injury, Guillain Barré syndrome, Parkinson's disease)

• advanced age (e.g. frailty)

• a history of trauma or surgery (e.g. burns, joint replacement surgery)

• underlying joint or muscle pathology and disease processes (e.g. inflammatory arthritis, osteoarthritis)

Participants will be included regardless of sensory or motor impairments.

Participants will be separated according to their diagnoses. There will not be pooling of data across diagnoses.

Studies may recruit participants at risk of contracture (i.e. stretch intervention aimed to prevent the development of contracture) or with existing contracture (i.e. stretch intervention aimed at reducing contracture). It is also possible that studies will recruit a mix of participants.

Types of interventions

This review will include studies of stretch interventions that aim to maintain or increase joint mobility. Interventions may include but need not be limited to: sustained passive stretching, positioning, splinting and serial casting. The intervention can be targeted at any synovial joint. In order to be included, the stretch intervention needs to have been sustained for a minimum of 20 seconds in a lengthened position on more than one occasion. This was selected by author consensus as the minimum plausible time period that stretch can affect joint mobility.

This review will include studies comparing a stretch intervention to no treatment, placebo, usual care or another stretch intervention.

Studies will be excluded if the stretch intervention is described as oscillating throughout range. The following interventions will also be excluded: joint mobilisation (passive accessory movements), joint manipulation, continuous passive motion, passive physiological movements or active movements.

Types of outcome measures

Outcomes will include measures of impairment, activity limitation and participation restriction. To be considered for this review, studies must report on the primary outcome of joint mobility. Outcomes can be measured at any time following the intervention.

Electronic searches

The following electronic databases will be searched using the search strategies in the linked appendices:

• Cochrane CENTRAL Register of Controlled Trials, DARE and HTA (Wiley Interscience), (Appendix 1)

• Cochrane Musculoskeletal Group Specialised Register

• MEDLINE (OVID; 1950 to present), (Appendix 2)

• CINAHL (OVID; 1982 to present), (Appendix 3)

• EMBASE (OVID; 1980 to present), (Appendix 4)

• SCI‐EXPANDED (ISI Web of Knowledge; 1900 to present), (Appendix 5)

• PEDro (http://www.pedro.fhs.usyd.edu.au/) (Appendix 6)

• WHO international clinical trials registry platform (http://www.who.int/trialsearch/)

In the SCI‐EXPANDED database we will search the citations of included studies to identify other potentially relevant studies.

Searching other resources

We will citation track the reference lists of included studies and previous systematic reviews for additional studies. Authors of included studies will be contacted for additional references. We will search clinical trials registers to identify unpublished and ongoing trials, including Current Controlled Trials (www.controlled‐trials.com), National Research Register (www.update‐software.com/national), and Australia New Zealand Clinical Trials Registry (www.anzctr.org.au).

Selection of studies

Two review authors will independently screen the titles and abstracts of the search output to identify potentially relevant studies. The full‐length papers will be retrieved for all potentially relevant studies and used to identify studies that meet the selection criteria. Studies published in languages other than English will be reviewed by colleagues who are fluent in that language. Review authors will not screen studies in which they have been involved. In such instances studies will be screened by two authors who were not involved in conducting the study. Disagreements between the two review authors will be resolved by discussion and, when necessary, arbitrated by a third person.

Data extraction and management

Two review authors will independently extract data from the included studies using pre‐determined data extraction forms. The following data will be extracted:

• aspects of study methodology including type of study, method of randomisation, method of allocation concealment and number lost to follow‐up.

• characteristics of the population including number of subjects, age, gender, health condition or patient group, severity of sensory or motor impairment and whether subjects were at risk of contracture or had existing contracture or a combination of the two.

• characteristics of the intervention and comparisons including details of treatment and control intervention, duration of intervention, frequency of intervention, intensity of intervention, details of co‐interventions, compliance with treatment, joint of interest.

• details of the primary outcomes

  • method used to measure joint mobility (e.g. active or passive measurements of joint mobility, angular or linear measures of joint range of motion, measurements with standardised torque or measurements without standardised torque, and methods used to determine 'end of range' of joint)

  • time period for measurement of joint mobility (e.g. measured immediately, 24 hours or one week after cessation of stretch)

  • mean change scores or mean outcome, and standard deviations of outcomes or change scores of each group. Where these data are not reported they will be calculated from reported data, or estimated from reported data, or authors will be contacted and asked to supply data.

• details of relevant secondary outcomes

• adverse effects and dropouts

Differences in the data extracted by the two review authors will be resolved by discussion and, when necessary, arbitrated by a third person. Data from studies published in languages other than English will be extracted by colleagues who are fluent in that language. Review authors will not extract data on studies in which they have been involved. In such instances data will be extracted by two authors who were not involved in conducting the study.

Assessment of risk of bias in included studies

The risk of bias of the included studies will be also assessed by two independent reviewers. As recommended by the Cochrane Handbook [Chapter 8] (Higgins 2008), the following methodological domains will be assessed:

1. Sequence generation

2. Allocation sequence concealment

3. Blinding of participants, personnel and outcome assessors

4. Incomplete outcome data

5. Selective outcome reporting

6. Other potential threats to validity

Each of these criteria will be explicitly judged using: Yes=(low risk of bias); B=No (high risk of bias); C=unclear (either lack of information or uncertainty over the potential for bias).

Disagreements in the quality rating will be resolved by discussion or, when necessary, arbitrated by an independent third person. The risk of bias of studies published in languages other than English will be assessed by colleagues who are fluent in that language. Review authors will not rate the risk of bias of studies in which they have been involved. In such instances risk of bias will be rated by two authors who were not involved in conducting the study.

Measures of treatment effect

For continuous outcomes with the same units, effects will be expressed as weighted mean differences (WMD) and 95% confidence intervals. For continuous outcomes with different units (e.g. angular or linear measurements of joint range of motion), effects will be calculated using standardised mean differences (SMD). Standardised mean differences will then be converted back to weighted mean differences for better clinical interpretation. For dichotomous outcomes we will express effects as risk ratios and 95% confidence intervals.

Dealing with missing data

Attempts will be made to contact study authors where there is incomplete reporting of data. All analyses will be performed on an intention‐to‐treat basis. However, we anticipate that this may not be possible because outcomes may not be available for participants who dropped out. If study authors do not respond to requests for additional data we will include all available results as feasible. If insufficient data are available for analysis the studies will be included in the review but only with a description of the study details.

Assessment of heterogeneity

If there are at least two clinically homogenous studies (studies which investigate the effect of similar interventions on similar patient groups and report similar outcomes), meta‐analysis will be considered. In such circumstances the I² statistic will be used to quantify heterogeneity of outcomes and determine whether it is appropriate to pool data. If there is low statistical heterogeneity, studies will be combined using a fixed‐effects model. If there is moderate statistical heterogeneity studies will be combined using a random‐effects model. Where heterogeneity is substantial the possible causes of heterogeneity will be explored by performing sensitivity analyses in which individual studies are omitted one at a time, or stratified by particular characteristics (see 'Subgroup analysis' and 'Sensitivity analysis' sections below) or, where appropriate, with meta‐regression. Data will be analysed using Review Manager (RevMan) 5.0.

Assessment of reporting biases

Where possible, data will be analysed from each study on an intention‐to‐treat basis. Small sample bias will be examined with the use of funnel plots.

Subgroup analysis

We will perform several planned comparisons to investigate the following issues:

• the immediate effects of stretch (i.e. effects present less than 24 hours after removal of the last stretch) versus the short‐term effects of stretch (i.e. effect present between 24 hours and one week after removal of the last stretch) versus the medium or long‐term effects of stretch (i.e. effect present more than one week after removal of the last stretch)

• the effects of stretch administered for the prevention of contracture versus the effects of stretch administered for the treatment of contracture

• the effects of stretch administered in different dosages (i.e. total time stretch was administered), and

• the effects of stretch administered in different diagnostic patient groups

If there are sufficient data we will also conduct subgroup analyses to investigate the following issues:

• the effects of stretch administered manually by a therapist versus the effects of stretch that are self‐administered versus the effects of stretch administered with positioning programs versus the effects of stretch administered with a plaster cast versus the effects of stretch administered with a splint

• the effects of stretch when administered to large joints (i.e. shoulder, elbow, hip and knee) versus small joints (i.e. joints about the wrist, ankle, hand and foot), and

• the effects of stretch when outcomes can be influenced by subjects' perceptions of discomfort (e.g. measures of active range of motion, measures of passive range of motion with a non‐standardised measurement torque) versus when outcomes can not be influenced by subjects' perceptions of discomfort (e.g. studies involving unconscious or insensate patients, measurements taken with a standardised torque)

Between‐study comparisons will be interpreted cautiously (i.e. these analyses will be considered exploratory).

Sensitivity analysis

Sensitivity analyses will be conducted to examine the robustness of the meta‐analyses. The sensitivity analyses will examine the effects of randomisation (true versus quasi), allocation concealment (concealed versus non‐concealed), blinding of assessors (blinding versus no blinding) and drop‐out rate (> 15% dropouts versus ≤15% dropouts) on study outcomes.

Grading the quality of the evidence

We will use the GRADE approach to evaluate the quality of the scientific evidence in this systematic review (GRADE Working Group 2004, Schünemann 2006, Guyatt 2008a, Guyatt 2008b). The GRADE approach specifies four levels of quality:
1. high quality: randomised trials or double‐upgraded observational studies
2. medium quality: downgraded randomised trials or upgraded observational studies
3. low quality: double‐downgraded randomised trials or observational studies, and
4. very low quality: triple‐downgraded randomised trials, downgraded observational studies or case series/case reports.

The highest quality rating is for randomised trial evidence. We will, however, downgrade randomised trial evidence to moderate, low, or even very low quality evidence, depending on the presence of the following five factors:
1. limitations in the design and implementation of available studies suggesting high likelihood of bias
2. indirectness of evidence (indirect population, intervention, control, outcomes)
3. unexplained heterogeneity or inconsistency of results (including problems with subgroup analyses)
4. imprecision of results (wide confidence intervals), and
5. high probability of publication bias.

The GRADE approach can also allow lower quality studies to be upgraded in quality based on the following three factors:
1. large magnitude of effect
2. all plausible confounding would reduce a demonstrated effect or suggest a spurious effect when results show no effect, and
3. dose‐response gradient.

Summary of findings tables

Summary of findings tables will be compiled using GRADEpro software to present the main findings of the review in a transparent and simple tabular format. The summary of findings tables will provide key information concerning the quality of evidence, the magnitude of effect of the stretch, and the sum of available data on the main outcomes.