Treadmill interventions with partial body weight support in children under 6 years of age at risk of neuromotor delay
Abstract
This is a protocol for a Cochrane Review (Intervention). The objectives are as follows:
To assess the effectiveness of treadmill interventions in preambulatory infants and children under six years of age who are at risk of neuromotor delay.
Background
Description of the condition
Typical gross motor development
The World Health Organization (WHO) describes the gross motor development of infants as the attainment of six gross motor milestones. These are: (1) sitting without support; (2) crawling on hands and knees; (3) standing with assistance; (4) walking with assistance; (5) standing alone, and (6) walking alone. Approximately 86% of children with typical development attain all six milestones, though the sequence of attainment may vary. For instance, crawling on hands and knees is the most variable milestone; it is observed at different ages during the infant’s development and is sometimes even skipped. While infants are learning these temporary means of locomotion, they are gradually becoming able to support increasing amounts of weight while in a standing position until they eventually begin to walk at around 12 months of age. Attainment of this ultimate milestone has the widest age range at between eight and 18 months of age (WHO 2006) and may depend on various environmental factors, such as sensory or motor stimulation.
Developmental delay
The International Classification of Functioning, Disability and Health for Children and Youth (ICFCY) (WHO 2005) describes developmental delay as retardation in the achievement of developmental milestones. The most plausible cause of the motor delay is an alteration in the typical development and function of the central nervous system. Motor delays in locomotor abilities are defined by standards used in clinical paediatric settings. For example, the onset of independent walking should occur prior to 18 months of corrected age, so the presence of a motor delay would not be considered before this age. Developmental delay in infants is usually diagnosed via routine screening (Case‐Smith 1998) and/or the use of one or more of the following diagnostic tests: (1) norm‐referenced tests; (2) criterion‐referenced tests; (3) brain imaging techniques; (4) kinetic and kinematic analysis using force plates and video motion analysis. Although used for both research and clinical purposes, these tests are typically not good predictors for later outcomes and generally lack sensitivity in detecting small changes in motor development (Heineman 2008). In addition, in the paediatric population the reliability of some of these tests may be affected by the child's emotional state, by daily fluctuations in performance or by the experience of the tester. Due to the continuous developmental changes occurring in the young brain, early diagnostic tests are relatively limited in predicting developmental outcomes (de Graaf‐Peters 2006) and the high level of variation in motor developmental trajectories in healthy children means that care has to be taken when interpreting results from motor assessments (Roze 2010).
Consequences of motor developmental delay
One of the major tasks in gross motor development is locomotion, the ability to move from one place to another (Bly 1995). The failure to attain walking or the late attainment of walking has consequences for the musculoskeletal system. The anatomy of the hip, for instance, needs weight bearing for proper bone growth and correct orientation of the femoral head, as well as for a correct alignment of the spine (Campbell 2006). As well as its importance for subsequent motor skill development, acquiring the ability to locomote is important for infants because of its impact on cognitive, social and emotional skills. Researchers have demonstrated that for infants with typical development, experience with locomotion is associated with the development of a broad array of cognitive skills, including the onset of wariness of heights; the concept of object permanence (objects hidden from sight still exist); a shift from self‐centred to landmark‐based spatial coding strategies; the ability to follow the pointing gestures and gaze of another person, and aspects of social referencing and detour reaching (Bertenthal 1984; Kermoian 1988; Campos 1989; Bertenthal 1990). This suggests that infants are better able to develop spatial cognition and learn about the world around them as they become able to locomote independently. Children who can walk independently show improved active exploration of their environment, as opposed to children who passively observe the environment when being held or carried through space. Rosenbloom 1971 further suggests that the quality of movement may affect subsequent development. He proposes that inefficient locomotion may hamper development by limiting the attention and energy that infants spend on exploration of the environment. Moreover, early locomotor experiences may have a larger impact on the developing brain than similar experiences at a later age due to the brain's high plasticity during the first few postnatal years (Webb 2001; de Graaf‐Peters 2006). Earlier achievement of developmental milestones, in particular independent walking, have also been associated with better intellectual performance in adulthood (Murray 2007). In summary, independent locomotion at early age not only facilitates the infant's motor development, but also impacts other developmental domains and affects quality of life for the child and his or her family.
Population affected
There are various reasons for delays in typical motor development. Disorders affecting motor development during infancy include Down syndrome, spina bifida, cerebral palsy (CP) and a broad range of other neuromuscular disorders (Campbell 2006).
In addition, preterm birth, defined as childbirth occurring at less than 37 weeks or 259 days gestation (Beck 2010), is associated with a series of risk factors that make children vulnerable to delays in their developmental process (Formiga 2011). For instance, children who are born prematurely have higher rates of CP, sensory deficits and/or learning disabilities compared with children born at term (Beck 2010).
The incidence of preterm birth rate is 6.2% in Europe, 10.6% in North America (excluding Mexico) and 6.4% in Australia (Beck 2010) and the incidence of CP is 1.5 to 2 per 1000 live births (Surveillance CP Europe). However, more epidemiological studies are needed to reliably assess the incidence for CP as its causes are not fully understood (Lie 2010). Approximately one in 800 children in the USA are born with Down syndrome, while the incidence in the UK is one in 1000 (Down's Syndrome Association).
Description of the intervention
According to some authors, high levels of motor activity are the key to motor development (Adolph 1998; Damiano 2006). In order to best influence neural plasticity, it is important that any training is performed early in development and that it is specific to the task the child needs to master (Hodgson 1994; Blackman 2002). Intervention studies examining infants developing in a typical and atypical way show that task‐specific training may best facilitate the development of postural control (Hadders‐Algra 1996; Sveistrup 1997; de Graaf‐Peters 2007).This concept of task‐specificity can be considered an evidence‐based concept based on neuroscientific principles (Hodgson 1994).
Although the optimal window of intervention within the motor domain is not clear (Nelson 2000), it is reasonable to think of independent walking as a motor task that needs to be achieved by six years of age if long‐term negative effects are to be minimised.
Locomotor treadmill interventions, with and without partial weight support, have been shown to promote the acquisition of independent walking in children with Down Syndrome (DS). Therefore, it can be used to prevent delay in the onset of walking. Ulrich 2001 found that children with DS who received intensive stepping practice on a treadmill with partial body weight support began walking independently 101 days sooner than their peers who were not exposed to treadmill stepping. In a subsequent study, it was found that children with DS who engaged in high frequency intervention protocols showed better results than children with low frequency intervention protocols (Ulrich 2008).
Locomotor treadmill interventions have also been used for children with cerebral palsy (CP) (Richards 1997; Begnoche 2007; Mattern‐Baxter 2009). For both children with DS and children with CP, treadmill interventions not only enhanced the onset of independent walking but also improved the quality of step type (Looper 2006; Cherng 2007).
Protocols of treadmill interventions described in the literature vary with regard to training speeds, support provided, manual assistance with stepping, and frequency and duration of the intervention. In studies of infants, the majority had training speeds ranging from 0.1 m/s to 0.22 m/s (Davis 1994); whereas, older children were trained at higher speeds of 1.8 m/s (Begnoche 2007). The percentage of body weight used as partial weight support varied across existing studies and was provided either manually (the infant is supported under the arms, with the feet resting on the treadmill surface, bearing as much weight as comfortable) (Ulrich 2001), or with a commercially available pelvic harness or trunk harness, or both (Dodd 2007; Provost 2007). Only a few studies quantified the amount of body weight support provided during training (Schindl 2000; Meyer‐Heim 2007; Provost 2007; Mattern‐Baxter 2009). Training duration ranged between two weeks (Bernitsky‐Beddingfield 2007; de Bode 2007; Provost 2007) and 57 weeks (Ulrich 2001), with some studies including breaks during the training programme (Day 2004; Prosser 2007; Cernak 2008). Frequency of the training sessions varied between studies from two to six training sessions per week (Damiano 2009; Mattern‐Baxter 2009a). Manual facilitation of gait varied from no assistance with leg advancement to assistance from up to three physical therapists (Mattern‐Baxter 2009a).
In summary, the existing scientific literature exhibits wide variation in the parameters of treadmill interventions, indicating a need for systematic establishment of intervention protocols. Furthermore, research found in paediatric populations has used the treadmill for both prevention and rehabilitation purposes. Its use as a preventive tool mainly relates to infants who have no prior walking experience; whereas training in rehabilitation would be directed towards infants or children who, having walked independently, need to retrain that skill after injury/physical dysfunction and/or who need to improve their walking parameters.
How the intervention might work
It is well established that brain plasticity exists and is particularly pronounced in the young nervous system (NS) (Stiles 2000; Stiles 2005). Experience‐dependent and/or activity‐dependent plasticity has been demonstrated in the human NS (Edgerton 1997; Eyre 2003) and postural control intervention studies (Harbourne 2003). The capacity for the NS to reorganize is one of the fundamental mechanisms by which therapeutic interventions may be effective.
The treadmill is one form of intervention used in physical therapy to enhance the locomotor capabilities of patients (Eng 2007; Verschuren 2008); however, most of the scientific knowledge related to this topic comes from animal models or interventions in adult human populations (Sullivan 2007). In fact, the use of treadmill interventions for people with neurological disorders has its roots in animal studies (Eidelberg 1980; Barbeau 1987) where adult cats were able to regain stepping skills after a complete lesion of the spinal cord. The underlying mechanism by which this technique is effective is thought to reside in the regenerating capacity (plasticity) of the central nervous system (CNS) when task‐specific motor practice is provided. Voluntary exercise and treadmill interventions specifically have been utilised in humans and in animal models to promote CNS (including spinal cord) plasticity and functional change (Jones 1999; Cotman 2002; Cotman 2002a). The underlying neuronal mechanisms responsible for such change are thought to be up‐regulation of trophic factors, neurogenesis, synaptogenesis, pre‐ and post‐synaptic modulation and angiogenesis, among others. These plasticity mechanisms are particularly active during early development. These neuroscience principles are the basis of the current motor learning theories (Newell 1991; Kleim 2008).
Plausible positive outcomes from treadmill interventions via CNS plasticity have been proposed in infants with DS and premature infants. Evidence from studies with children who have Down Syndrome indicate statistically significant improvements in a variety of outcome measures including obstacle negotiation and onset of walking. For this population, two main benefits from treadmill interventions implemented during early development have been described. Firstly, it promotes the transition to continuous alternating steps in infants (including typically developing infants (Thelen 1986; Thelen 1991)), which is an important precursor to walking (Ulrich 1992; Ulrich 1995; Ulrich 2001). Secondly, it leads to an acceleration of the onset of independent walking and an improvement of the quality of gait (Ulrich 2001).
Observational studies suggest that infants born prematurely follow similar developmental trajectories to their full‐term peers, although frequently with some delay (Luo 2009; Angulo‐Barroso 2010). The neonatal period of preterm infants is stressful as the immaturity of vital physiological functions makes it difficult for the infant to adapt to the extrauterine situation. This results in vulnerability to delay in motor development and to developmental disorders (Goyen 2002; Pin 2010; Prins 2010; Formiga 2011). The evidence available on the effect of treadmill interventions for this population is almost non‐existent. A case study of a premature infant showed an increase in the number of steps, of which almost 100% were exclusively alternating steps, during the post‐training phase (Bodkin 2003). However, encouraging as these results may seem, evidence of the effectiveness of treadmill interventions remains inconclusive.
Why it is important to do this review
The importance of children attaining independent walking has been well documented. A range of interventions to improve motor development in children is currently used in practice (Riethmuller 2009). However, research on early interventions for children with physical disabilities is very limited, and most studies have methodological limitations (Ziviani 2010).
Treadmill interventions are now being used in rehabilitation to prevent walking problems with children under six years of age. This intervention could have significant benefits in terms of preventing gross motor delays, promoting cognitive and social development and promoting correct biomechanical function during gait. It is important to evaluate the effectiveness of treadmill training as an early intervention method designed to improve motor function and to prevent neuromotor delays in children.
Diagnoses that may result in a delay in the acquisition of walking (DS, CP or spina bifida, among others) have different intrinsic characteristics. Because of this, a differentiation of interventions or parameters specific to the diagnosis may be required, indicating the need to perform subgroup analyses.
There are several existing systematic reviews on treadmill interventions in paediatric populations (Damiano 2009; Mattern‐Baxter 2009a; Mutlu 2009; Willoughby 2009; Molina‐Rueda 2010). However, these reviews reviewed published reports from 1980 to 2008 on treadmill training for children aged up to 21 years. In addition to their reliance on published reports in English, their search strategy did not include terms of specific diagnoses that are known to cause gross motor delay in childhood, and some were limited to children with CP (Mattern‐Baxter 2009a; Mutlu 2009; Willoughby 2009; Molina‐Rueda 2010).
To date, there is no systematic review of treadmill intervention that examines its effectiveness on children before or during the acquisition of independent walking, and that encompasses both prevention and rehabilitation. A systematic review of the literature is needed in order to define the extent of the preventive and rehabilitative effectiveness of treadmill training, and to define optimal training parameters for this intervention.
This review aims to fill this gap and to review all relevant studies, irrespective of publication status or language.
Objectives
To assess the effectiveness of treadmill interventions in preambulatory infants and children under six years of age who are at risk of neuromotor delay.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials, quasi‐randomised controlled trials (that is, where participants are allocated in a way that is not strictly speaking random, such as by alternation or date of birth) and controlled clinical trials.
Types of participants
Children up to six years of age with delays in gait development or the attainment of independent walking (children who cannot walk independently by the age of 18 months), or who are at risk of neuromotor delay (primarily with non‐progressive neurological disorder), however diagnosed.
We will exclude children diagnosed with a condition for which physical activity is contraindicated, for example, infants with genetic degenerative diseases such as neuromuscular dystrophy (and those with diagnoses that preclude independent walking).
Types of interventions
Treadmill intervention of any type, frequency or intensity, aimed at (1) improving gait parameters such as walking speed, endurance, quality of step (how the foot lands on the floor surface) or (2) facilitating onset of independent walking or walking with assistive devices.
Comparison groups will be no treatment or another treatment. Control group treatments may include physical therapy or another intervention designed to improve gait. We will include studies in which treadmill intervention is an adjunctive treatment. We will also report on studies comparing different types of treadmill interventions.
Types of outcome measures
We will accept five types of outcome measures: standardized measures, questionnaires, self‐report data, data from motion analysis systems and coded‐video observations. Based on the ICFCY, we will assess the following outcomes.
Primary outcomes
A. Body functions ‐ Neuromusculoskeletal and movement related functions ‐ Gait pattern functions.
-
Step frequency (number of steps per minute).
-
Step quality (foot doing toe versus flat contact during treadmill stepping).
B. Activities and participation functions.
-
Age of onset of independent walking.
-
Age of onset of walking with assistive devices.
-
Gross motor function.
-
Falls and injuries due to falls.
Secondary outcomes
C. Body functions ‐ Neuromusculoskeletal and movement related functions ‐ Gait pattern functions.
-
Inter‐ and intra‐limb co‐ordination.
D. Activities and participation functions.
-
Infant or child quality of life.
If data permit, we will examine outcomes by intervention type ( preventive or rehabilitative) and by diagnosis (for example CP, DS and other).
We will include the primary outcomes in a 'Summary of findings' table, if data permit.
Search methods for identification of studies
Electronic searches
We will search the following electronic databases with no language restrictions.
MEDLINE
EMBASE
CINAHL
PsycINFO
LILACS (Latin American and Caribbean Health Sciences Literature)
PEDro
Science Citation Index
Conference Proceedings Citation Index
metaRegister of Controlled Trials
National Research Register Archive (UK)
Center Watch Clinical Trials Listing Service (USA)
We will also search Trials Registers ‐ Clinical Trials.gov and WHO ICTRP.
For dissertations, we will perform a search in WorldCat.
We will use the following search strategy, modified as required for each database.
1 Physical Therapy Modalities/
2 "Physical Therapy (Specialty)"/
3 (physiotherap$ or physio therap$ or physical therap$).tw.
4 Exercise Therapy/
5 tread‐mill$.tw.
6 treadmill$.tw.
7 or/1‐6
8 Motor Skills/
9 Motor Skills Disorders/
10 Motor Activity/
11 Psychomotor Disorders/
12 Psychomotor Performance/
13 Movement Disorders/
14 Developmental Disabilities/
15 ((motor or neuromotor or neuro‐motor or psychomotor or psycho‐motor or development$) adj3 (impair$ or skill$ or disorder$ or deficit$ or delay$ or disabilit$)).tw.
16 exp Walking/
17 Gait/
18 Gait Disorders, Neurologic/
19 Gait Ataxia/
20 gait.tw.
21 locomotion/
22 (walk or walking).tw.
23 (ambulation or ambulatory or nonambulation or nonambulatory or non‐ambulation or non‐ambulatory).tw.
24 (locomotor$ or locomotion$).tw.
25 stepping.tw.
26 or/8‐25
27 Disabled Children/
28 down syndrome/
29 cerebral palsy/
30 spinal dysraphism/
31 (down$ syndrome or cerebral pals$ or (spin$ adj3 injur$) or spina bifida).tw.
32 exp infant, low birth weight/ or infant, premature/
33 (low birth weight or pre‐term$ or preterm$ or prematur$).tw.
34 or/27‐33
35 Infant/
36 exp child/
37 (baby or babies or infant$ or child$ or toddler$ or pre‐school$ or preschool$ or schoolchild$).tw.
38 35 or 36 or 37
39 randomized controlled trial.pt.
40 controlled clinical trial.pt.
41 randomi#ed.ab.
42 placebo$.ab.
43 drug therapy.fs.
44 randomly.ab.
45 trial.ab.
46 groups.ab.
47 or/39‐46
48 exp animals/ not humans.sh.
49 47 not 48
50 7 and 26 and 38
51 7 and 34 and 38
52 50 or 51
53 49 and 52
Searching other resources
1. We will identify studies incorporated in previous systematic reviews and other reviews of the subject to consider them for inclusion.
2. We will also read bibliographies of articles identified through the search strategy for additional sources for inclusion.
3. We will evaluate unpublished abstracts and dissertations for possible inclusion.
Data collection and analysis
Selection of studies
We will divide the titles and abstracts yielded by the search strategy into two blocks. Two authors will independently screen the first block of references (KMB and CB), while two other authors will do so with the second block (RA and MV), using the inclusion criteria described above. RA will be the arbiter for KMB and CB, while KMB will fulfil this role for RA and MV, in case of discrepancies. We will then obtain the selected titles in full text to determine their relevance for the review. We will resolve disagreement about eligibility through discussion, and when disagreements cannot be resolved, by seeking advice from another author (MHA). We will record the reasons for excluding trials.
Data extraction and management
Four authors (MV, RA, CB and MG) will independently extract data for each trial using a data extraction form to collect information about the population, intervention, randomisation methods, blinding, sample size, outcome measures, follow‐up duration, attrition and handling of missing data and methods of analysis.
Assessment of risk of bias in included studies
Three authors (CB, MV and RA) will independently assess the risk of bias of each included study using the Cochrane Collaboration’s tool for assessing risk of bias (Higgins 2008). Review authors will independently assess the risk of bias within each included study in relation to the following six domains on the basis of ratings of low risk of bias, high risk of bias and unclear risk of bias: sequence generation; allocation concealment; blinding; incomplete outcome data (including data on attrition and exclusions; differentiating intention‐to‐treat analyses from per‐protocol ("as treated") analyses); selective outcome reporting, and other risks of bias. We will enter these judgements into a 'Risk of bias' table in Review Manager 5.1 (Review Manager 2011) with a brief rationale for the judgements.
Details on the possible sources of bias are below.
Sequence generation
We will describe the method used to generate the allocation sequence in detail so as to assess whether it should have produced comparable groups. We will evaluate whether or not the allocation concealment sequence was adequately generated.
Allocation concealment
We will describe the method used to conceal allocation sequence in sufficient detail to assess whether intervention schedules could have been foreseen before or during recruitment. We will judge whether or not there was adequate allocation concealment.
Blinding
We will describe any measures used to blind outcome assessors so as to assess whether knowledge of the allocated intervention was adequately prevented. It is not possible to blind either those who deliver the therapy (treadmill training) or those infants who receive it, due to the nature of the intervention. Our assessment of risk of bias will take into account the likely bias attributable to the inability to blind participants or personnel in such interventions.
Incomplete outcome data
We will extract and report data on attrition and exclusions, as well as the numbers involved (compared with the total randomised), reasons for attrition or exclusion (where reported or obtained from authors) and any reinclusions in analyses performed by review authors. For each included study, we will assess whether incomplete outcome data were adequately addressed.
Selective reporting
We will make attempts to assess the possibility of selective outcome reporting by investigators. We will evaluate if each study is free from selective outcome reporting by following these judgements:
-
low risk of bias, when all collected data seems to be reported;
-
unclear risk of bias, when it is not clear whether other data was collected and not reported;
-
high risk of bias, when the data from some measures used in the trial are not reported.
Other risks of bias
We will assess the extent to which each study is apparently free of other problems that could put it at high risk of bias, by describing important concerns not addressed in the other domains with the Cochrane Collaboration's 'Risk of bias' tool. We will assess other threats to validity as 'low risk of bias' if the study appears to be free of other sources of bias. Where the risk of bias is unclear from published information, we will attempt to contact the authors for clarification. If that fails, we will then classify that study as at unclear risk of bias.
Measures of treatment effect
We will use the Cochrane Collaboration's Review Manager software (RevMan 5.1) to calculate the adjustments of measures of treatment effects (Review Manager 2011).
Continuous data
We will analyse continuous data if means and standard deviations are reported, can be obtained from primary investigators or can be calculated from the available data. If continuous outcomes are measured identically across studies, we will calculate the mean difference (MD) with 95% confidence interval (CI). If the same continuous outcome (for example, infant's gross motor development level) is measured differently across studies, we will compare standardised mean differences (SMD) with 95% CI across studies (Higgins 2008). Where necessary, we will use formulas to convert F ratios, t‐values and chi‐square values into SMDs (Lipsey 2001), using Hedges g to correct for small sample bias.
Dichotomous data
We will analyse the outcomes of any study reporting binary/dichotomous data by calculation of the risk ratio for the occurrence of an event (rather than a non‐event) for its consistency as a summary statistic and ease of interpretation.
Unit of analysis issues
The authors will take into account the unit of analysis to determine whether: (1) individuals were randomised in groups (i.e. cluster‐randomised trials); (2) results were reported at multiple time points, and (3) individuals simultaneously received multiple interventions.
Cluster‐randomised trials
For trials that use clustered randomisation, we will present results with proper controls for clustering (robust standard errors or hierarchical linear model). If appropriate controls are not used and it is not possible to obtain the full set of each individual participant's data, we will control the data for clustering using the procedures outlined by Higgins 2008. For dichotomous outcome measures, we will divide the number of events and the number of participants per trial arm by the design effect [1 + (1‐m)*r], where m is the average cluster size and r is the intra‐cluster correlation coefficient (ICC). For continuous outcome measures, we will divide the number of participants per trial arm by the design effect, with the mean values unchanged. To determine the ICC, we will use estimates in the primary trials on a study‐by‐study basis. In the case of these values not being reported, we will use external estimates of the ICC that are appropriate to the context of each trial and average cluster size. If they were still not available, we will then use statistical procedures outlined by Higgins 2008.
Cross‐over trials
We will combine the results from cross‐over trials with those of parallel group trials. In cross‐over trials, we will only include the first phase before the point of cross‐over in the analyses. In case of not having the results for the first phase separated from the rest, we will contact the authors to obtain these data.
Multiple time points
When the results are measured at multiple time points, we will only consider baseline measurements and the last time point measurements.
Multiple interventions per individual
If it is found that participants in some trials receive multiple treatments, we will conduct meta‐analysis on those studies separately: the treadmill intervention plus treatment as usual arm will be compared to treatment as usual alone.
Dealing with missing data
We will assess missing data and dropouts in the included studies. We will investigate and report the reasons, numbers and characteristics of dropouts. We will make efforts to contact the authors when further information or data are necessary.
For dichotomous data, we will report the missing data and dropouts for included studies along with the number of participants who are included in the final analyses as a proportion of all participants in each study. We will provide reasons for missing data in a narrative summary. The extent to which the results of the review could be altered by the missing data can be assessed based on consideration of best‐case and worst‐case scenarios (Gamble 2005). The best‐case scenario is the one where all participants with missing outcomes in the experimental condition had good outcomes and all those with the missing outcomes in the control condition had poor outcomes, and the worst‐case scenario is vice versa (Higgins 2008). However, the best‐case and worst‐case scenarios method is too extreme and a more plausible approach is needed. We will use the method suggested by Higgins 2008, which can incorporate specific reasons for missing data and considers plausible event risks among missing participants in relation to risks among those observed.
We will analyse missing continuous data either on an endpoint basis, including only participants with a final assessment, or using last observation carried forward to the final assessment if the last observation carried forward data were reported by the trial authors. If SDs are missing, we will make attempts to obtain these data through contacting trial authors. If SDs are not available from trial authors, we will calculate them from t‐values, confidence intervals or standard errors, where reported in articles (Deeks 1997a; Deeks 1997b). If these additional figures are still not available or obtainable, we will not include the study data in the comparison of interest.
Assessment of heterogeneity
We anticipate finding considerable heterogeneity across studies that might be included in this review. We will assess clinical heterogeneity by comparing the distribution of important participant factors among trials (for example, age, diagnosis), and trial factors (for example, randomisation concealment, blinding of outcome assessment, form of treadmill training, losses to follow‐up). We will describe statistical heterogeneity using I2 (Higgins 2002), a quantity that describes approximately the proportion of variation in point estimates that is due to heterogeneity rather than sampling error). In addition, we will employ a chi2 test of homogeneity to determine the strength of evidence that heterogeneity is genuine.
If an individual study appears to be an outlier, we may carry out sensitivity analysis with and without the study. If the primary studies are judged to be substantially heterogeneous even within these sub‐groupings, we will only give a descriptive analysis, particularly if there is variation in direction of effect.
Assessment of reporting biases
In order to investigate the relationship between effect size and standard error, we will draw funnel plots if sufficient studies are available (i.e., ten or more individuals studies). Asymmetry could be attributable to publication bias, but might also reflect a real relationship between trial size and effect size. if we find such a relationship, we will examine clinical variation of the studies (Higgins 2008, Section 10.4). As a direct test for publication bias, we will compare results extracted from published journal reports with results obtained from other sources, including correspondence.
Data synthesis
We will synthesise the data using RevMan 5.1, the latest version of the Cochrane Collaboration's meta‐analysis software (Review Manager 2011). We anticipate finding small trials, with sparse amounts of data. We propose to use the Mantel‐Haenszel method, the default fixed‐effect method in RevMan 5.1 (Higgins 2008, section 9.4.4.1). This method can pool odds ratios, risk ratios and risk differences.
For continuous variables, we will apply the mean difference approach where data allow.
For dichotomous outcomes, we will also calculate the number needed to treat for an additional beneficial outcome.
When meta‐analysis is inappropriate, we will provide a narrative description of the study results.
Subgroup analysis and investigation of heterogeneity
We will undertake subgroup analysis if clinically different interventions are identified or there are clinically relevant differences between participant groups. We will thus investigate any subgroup differences in order to establish whether there is a single intervention effect, specifically:
-
treadmill 'dose' (total number of training sessions, frequency of training per week or duration of each training session);
-
type of intervention (preventive or rehabilitative);
-
diagnosis (cerebral palsy, Down's syndrome etc.);
-
conditions affecting the neuro‐musculoskeletal system (hypo‐ or hypertonia, spasticity, posture etc.).
Sensitivity analysis
We will conduct sensitivity analysis, where data permit, to determine whether findings are sensitive to restricting inclusion to studies judged to be at low risk of bias. In these analyses, we will re‐evaluate the findings, limiting the inclusion to published studies or to those studies that have a low risk of:
-
selection bias (associated with allocation concealment and sequence generation);
-
performance bias (associated with blinding);
-
attrition bias (associated with completeness of data).
