A prospective, randomised controlled trial will be carried out. 130 subjects will be randomly allocated into either an experimental group (treadmill walking and partial weight support with one therapist) or a control group (assisted overground walking with one therapist) by a recruiter blinded to the sequence of group allocation. All outcome measures and data analysis will be completed by a researcher who is blinded to participant group allocation. The study has obtained ethical approval from the Human Research Ethics Committees of each of the sites involved in the study.
Stroke patients will be screened and invited to participate if they:
• are within 3 weeks of their first stroke
• are aged between 50 and 85 years of age
• are diagnosed clinically with hemiparesis or hemiplegia of acute onset, and
• are non-ambulatory defined as scoring 0 or 1 on the Motor Assessment Scale for stroke.
Participants will be excluded if they:
• have clinically evident brainstem signs
• have severe cognitive and/or language deficits which preclude them from following instructions in training sessions
• have unstable cardiac status which would preclude participation in a rehabilitation program, or
• have any pre-morbid history of orthopaedic conditions of the lower limbs which would preclude them from relearning to walk.
The presence of sensory loss, neglect and/or spasticity will not be exclusion criteria. However, their severity will be recorded using the Nottingham Sensory Assessment for sensory loss, the line bisection test for neglect, and the Ashworth Scale for spasticity. In addition, information about site and size of lesion will be collected.
Participants will be randomised into an experimental or a control group. We will stratify the randomisation. First, given the potential confounder of site, participants at each site will be randomised separately. Second, at each site, participants will be stratified according to initial level of motor disability since it has been found to affect outcome [16
]. Since all the participants will be unable to walk on admission to the study, sitting balance will be used to stratify the allocation of participants to groups because it has been found to be a useful prognostic indicator of walking outcome [17
]. Therefore, participants will be stratified according to Item 3 (Sitting Balance) of the Motor Assessment Scale for stroke [20
] so that those with a score of 0–3 will be classified as severely disabled and those with a score of 4–6 will be classified as moderately disabled. Within each of the two strata (moderate versus severe level of disability), participants will be allocated randomly to one of two groups, the experimental group or the control group. Random permuted blocks will be used so that after every block (of 6–10 participants), the experimental and control groups will contain equal numbers. In summary, stratification will occur according to site (three sites) and level of disability (two levels). Therefore, there will be 6 strata and participants will be randomised separately within each stratum. The random sequence of group allocation will be concealed from the person recruiting participants.
Both the experimental and the control group will undergo a maximum of 30 minutes per day of walking practice with assistance from one therapist, five days a week until they walk or until discharge from rehabilitation. The total daily time of intervention will be 30 minutes from beginning (ie, from when the participant is in the gym) to end (ie, when the participant is back in the wheelchair). This time therefore includes transferring, putting on aids or setting up equipment, ie, training does not have to be continuous so that rests may be taken. The amount of assistance during walking will be standardised to one therapist, however, additional help will be allowed during setting up walking (ie, getting the participant onto the treadmill for treadmill walking or into standing for overground walking). The rationale for this protocol is based on clinical observation of how much time and how many therapists are currently used in trying to get a non-ambulatory person to walk. Other intervention involving lower limb function, (ie, strengthening exercises, practice loading the affected leg during activities such as sitting, standing up and standing) will be standardised to 60 min per day. No other part of the multidisciplinary rehabilitation program will be controlled. Randomization should ensure that any effect of other interventions will be the same for both groups, therefore, other therapies will not be withheld.
Training for the experimental group will primarily involve walking on a treadmill supported in a body harness. Treadmill training with partial weight support via an overhead harness will be conducted using commercially available systems such as the Spacetrainer (TR Equipment, Tranas, Sweden) and the Lite-Gait (Mobility Research, USA). These systems have an access ramp so that a wheelchair can be wheeled onto the treadmill and an automatic lifter so that the harness can be prefitted in sitting or lying and the patient lifted into standing. In addition, there is good access to the patient's legs and the treadmills can run extremely slowly allowing adequate time to assist the legs to swing through.
There will be guidelines to determine the progression of training both in terms of increasing treadmill speed and reducing weight support. At the start, support from the harness will be as little as possible but up to a maximum of 40% of body weight since Hesse et al [21
] have found this to be the maximum support which does not dramatically alter the kinematic and kinetic features of walking. The actual weight relief will be determined by observation of whether the knee can extend in midstance. If the knee remains flexed, then the affected lower limb muscles are too weak to support the body weight and indicates that more weight relief is required. At the start, the speed of the treadmill will be as fast as comfortable while still maintaining a reasonable step length. If a participant is too disabled to walk on a treadmill moving at 0.1 m/s with the assistance of one therapist, they will walk on the spot practising lifting their feet rhythmically. When participants attain a speed of 0.4 m/s, a reduction in weight support will occur if participants can (i) swing their affected leg through without help, (ii) maintain a straight knee during stance phase without hyperextension, and (iii) maintain an adequate step length (rather than a high cadence) without help. These guidelines have been tested for feasibility and published [8
]. Information describing the specific features of the training session (such as treadmill speed, amount of weight support, distance walked, assistance required) will be recorded to monitor adherence to the guidelines and to be able to describe the intervention accurately.
Training for the control group will involve current practice of assisted overground walking. If a participant is too disabled to walk with the help of one therapist, they will practise stepping forwards and backwards or standing with a knee splint and practising shifting weight from leg to leg. Aids such as knee splints, ankle-foot orthoses, parallel bars, forearm support frames and walking sticks can be utilised as part of training. Training aims to produce independent walking. Therefore, progression of training encompasses both increasing speed and reducing assistance from both aids and the therapist. Information describing the specific features of the training session (such as use of aids, distance walked, assistance required) will be recorded to monitor adherence to the guidelines and to be able to describe the intervention accurately.
The end-point of the training phase of the study will be either the attainment of independent walking or discharge from the rehabilitation unit. The end-point of the follow-up phase of the study will be 6 months after admission to the study.
The initial outcome measures will be:
Proportion of participants achieving independent walking
Independent walking will be operationally defined as 'being able to walk 15 m continuously barefoot across flat ground without any aids'. Participants will be tested once a week in the morning (ie, before the training session). Participants will continue to be tested until they achieve independent walking or are discharged.
Quality of independent walking
Quality of walking will be measured by quantifying parameters such as speed, affected and intact step length, step width, and cadence. These parameters are the result of the timing and magnitude of the angular displacements during walking and are therefore global measures which reflect qualitative aspects of walking. When participants achieve independent walking, their overground walking will be measured by placing markers on the heels of both the unaffected and affected legs, so that step length of both the affected and unaffected leg, walking speed, cadence, and step width can be determined.
The 6-month outcome measures will be:
Proportion of participants achieving independent walking
Quality of independent walking
Two aspects of community participation will be measured. First, mobility status will be assessed using the 6-min walk test, number of falls since discharge, a self-efficacy questionnaire about walking capability, and the Adelaide Activities Index. Second, living arrangements will be assessed by recording the type of residence and the amount of support within the residence.
130 participants will be recruited. The sample size has been calculated to reliably detect a treatment effect size of a 25% increase in proportion of independent walkers with 80% power at a two-tailed significance level of 0.05. For non-ambulatory patients, it takes an average of 3 months to achieve independent walking with about 50% walking independently at six weeks [1
]. We are interested in being able to detect a 25% increase, from 50% to 75%, proportion of non-ambulatory patients walking independently by six weeks. Only an effect such as this is clinically significant enough to warrant a change in the implementation of services which would involve the re-education of physiotherapists and the expense of purchasing a treadmill and overhead harness system. The smallest number of participants to detect this difference between two proportions estimated from independent samples is 65 participants per group, ie, 130 participants in total [23
]. However, since the analysis of the data is survival curve analysis, there is greater power than the calculation would suggest because the effect of dropouts is minimal in this analysis. For example, in a pre-post design trial of 6 weeks intervention, if a subject dies or is lost to follow-up (ie, drops out) at 5 weeks, there is no measurement at the post-test on which to perform an analysis. In this trial, which ascertains once a week whether independent walking has been established, they would be censored at five weeks so that they no longer form part of the total sample. However, this participant's data would be available for the five weeks they participated in the trial thereby minimising the effect of their dropping out.
In addition to, but separate from, the difference in the proportion of participants walking between the groups, there may also be a difference in the quality of walking. On the assumption that 20% of participants may be lost to follow-up and that 80% of participants entering rehabilitation achieve basic independent walking [1
], there are likely to be 84 participants with "quality of walking" data at 6 months. Goldie et al [24
] has suggested that a minimum difference in walking speed worth detecting is 0.2 m/s. The walking speed of a population of stroke patients who have recently completed rehabilitation [25
] on entry to a randomised controlled trial was 0.56 (SD, 0.27) m/s using measurement procedures similar to the present proposal. 60 participants are needed to detect a treatment effect size of 0.2 m/s difference in walking speed between the groups at 6 months with 80% power at a two-tailed significance level of 0.05 therefore 84 participants gives over 90% power.
The proportion of independent walkers and the time to achieve independent walking will be compared between the two groups using the logrank test, in which those who do not achieve independent walking are censored at the time they are discharged. Survival analysis using Cox's regression will be used to compare the times in the two groups while allowing for possible confounding variables, such as other interventions received, and baseline sitting balance.
The five variables that reflect the quality of walking: speed, affected and intact step length, step width and cadence, will be compared between the two groups using Student's t-test, or Wilcoxon's rank-sum test for variables that are clearly not Normally distributed.
The four variables that reflect the mobility status: 6-min distance, number of falls since discharge, self-efficacy questionnaire about walking capability, and the Adelaide Activities Index will be compared between the two groups using Student's t-test, or Wilcoxon's rank-sum test for variables that are clearly not Normally distributed.
The two variables which reflect living arrangements: type of residence and the amount of support within the residence, will be analysed descriptively.
Descriptive data about lesion, neglect, spasticity and sensation will be used in post-hoc multiple regression analysis to examine if these factors affected walking outcome.