Stroke rehabilitation study results surprise researchers

Trial found no additive benefit of strength training, but did document efficacy of BSWTT alone

Published in the May 2007 issue of BioMechanics

By Jordana Bieze Foster

If, as it's said, the best laid plans of mice and men often go awry, then apparently stroke research is no exception. A recent trial proved that to be the case.

Given that task-specific training and strength training are both thought to improve gait in post-stroke patients, a multicenter team of researchers designed the Strength Training Effectiveness Post-Stroke (STES) trial1 to assess the additive benefit of combining the two rehabilitation strategies. To the researchers' surprise, however, the 80-patient prospective randomized study found that, in fact, combining body-weight-supported treadmill training (BWSTT) with either locomotor-based or muscle-specific strength training (MSST) was no more effective for improving gait velocity than treadmill training plus a sham intervention.

After further analysis, the investigators concluded that the counterintuitive findings stemmed from a flaw in the study's design. The patients had been randomized to one of four therapy groups, each of which was assigned a pair of training regimens: treadmill training plus locomotor-based strength training (LBST); treadmill training plus MSST; treadmill training plus a "sham" upper extremity training protocol; or LBST plus sham. Patients in each group attended 24 therapy sessions over the course of six weeks, 12 sessions for each of the two regimens assigned to the group. The key, researchers believe, was that exercise regimens changed from one session to the next, which may not have allowed patients enough time to recover between sessions.

"What we had probably done was overtrained," said Katherine J. Sullivan, PT, PhD, an assistant professor of biokinesiology and physical therapy at the University of Southern California in Los Angeles, who was one of three coprincipal investigators presenting the STEPS trial data in February at the annual Combined Sections Meeting of the American Physical Therapy Association.

Gait picks up pace

Despite not illuminating any benefit to task-strength combination therapy, the trial results did demonstrate that BWSTT plus sham therapy was significantly more effective at improving gait velocity than LBST plus sham therapy. The mean change in velocity from baseline after 24 sessions was significantly greater for the BWSTT-sham group than for the LBST-sham group at both comfortable and fast speeds. In addition, the mean velocity-which did not differ significantly between groups at baseline-was significantly higher in the BWSTT-sham group than in the LBST group after 24 sessions and at six months. More than 50% of subjects who received treadmill training were classified at a higher functional walking level by the end of the intervention than at baseline.

These findings offer some validation to supporters of BWSTT, particularly in light of recent meta-analyses that have failed to identify convincing evidence of the therapy's effectiveness. In an October 2005 Cochrane Review analysis, researchers from the University of Sydney, Australia, found no statistically significant differences between treadmill training (with or without body-weight support) and other interventions in 15 trials.2 The Evidence-Based Review of Stroke Rehabilitation, published in 2005 as well, found conflicting evidence that BWSTT improves walking and motor recovery, but also found moderate evidence that it is no better than aggressive braced assisted walking.3

More detailed analysis of the pooled STEPS data also provided insight into the variables that may contribute to improved walking speed in the hemiparetic post-stroke population.

Researchers measured torque generation for ankle plantar flexion and dorsiflexion, knee flexion and extension, and hip flexion and extension; not surprisingly, muscles on the nonparetic side were significantly stronger than those on the paretic side. The mean change in strength from baseline to six weeks reached statistically significant levels for ankle plantar flexion (both limbs), knee extension (both limbs), and hip flexion (paretic limb only).

Both before and after the intervention, muscle strength was found to correlate with walking performance; however, the correlation involved different muscles at each timepoint. At baseline, Berg Balance Scale (BBS) and paretic knee extension torque accounted for more than 90% of the variability in both self-selected and fast walking velocities; BBS, paretic knee extension torque, and nonparetic knee extension torque accounted for 71% of the variability in six-minute walk distance. At the end of the intervention, BBS and paretic hip flexion torque accounted for 67% of variability in self-selected velocity and six-minute walk distance; BBS, paretic hip flexion torque, and paretic knee flexion torque accounted for 69% of the variability in fast velocity.

In addition, the researchers were surprised to find that strength gains did not correlate with walking performance gains. There was only one statistically significant correlation between change in torque values and change in walking performance: change in paretic knee flexion torque was moderately correlated with change in fast walking speed.

"It's a very curious finding that even though at baseline there were strong relationships, there didn't seem to be a change relationship," said David A. Brown, PT, PhD, an assistant professor of physical therapy and human movement services at Northwestern University in Chicago. "Strength improvement alone cannot explain changes in walking performance."

One possible explanation for this result, Brown suggested, is that the magnitude of strength change necessary to significantly impact walking performance may be more than what most of the study subjects were able to achieve.

Ankle and hip analysis

Knowing that improvement in walking performance was dependent on more than just strength gains, the researchers analyzed gait data obtained in a subgroup of the study population to determine the relative contributions of biomechanical parameters at the ankle and hip. Looking at a pooled subgroup consisting of five subjects from each intervention group, they found that those in whom free walking velocity improved by more than 0.06 m/s from baseline (the "high response" group) had significantly greater gains than those in the "low response" group for five biomechanical variables: plantar flexion angle and power in the preswing phase of the gait cycle, hip extension angle and moment in terminal stance phase, and preswing hip flexion power (Table 1). Low-response subjects had only minimal, if any, improvement with respect to those five variables, but did exhibit an increase from baseline for hip flexion in swing phase. In addition, subjects in the high-response group demonstrated a 15% change in soleus intensity, compared with a 0.3% increase in the low-response group; change in walking velocity was not related to increased activity in any other muscles.

Interestingly, the researchers found that initial walking speed was not significantly predictive of change in walking speed. Two baseline variables, ankle plantar flexion torque and lower extremity Fugl-Meyer (FM) score, were moderately predictive of change in walking speed in the 20-person subgroup.

Patients whose FM score improved by more than 25 points also demonstrated gains in walking speed, ankle plantar flexion angle at push-off, and ankle plantar flexion power at preswing that were significantly greater than those whose FM score improved by 25 points or fewer. Those with greater than 25% improvement in ankle plantar flexion torque also had significantly greater gains than those with less change in ankle plantar flexion torque for all gait parameters analyzed; however, subjects with lower levels of ankle strength improvement had significantly greater change in gluteus maximus intensity than those with greater ankle strength improvement (Table 2).

"Subjects with higher capacity made improvements at the ankle and hip. Those with lower capacity had improvements more proximally," said Sara J. Mulroy, PT, PhD, director of the pathokinesiology laboratory at Rancho Los Amigos National Rehabilitation Center in Downey, CA.

The STEPS trial may have fallen short of its original goal, but its data offer new insights about variables that contribute to gait rehabilitation following stroke, and also provide a base from which future research can evolve. The STEPS findings about the efficacy of BWSTT helped secure funding for the ongoing Locomotor Experience Applied Post-Stroke (LEAPS) trial, which will randomize 400 stroke patients to receive early BWSTT (beginning two months post-stroke), late BWSTT (beginning six months post-stroke), or a control regimen. For more information about the LEAPS trial, visit

Jordana Bieze Foster is a freelance writer based in Massachusetts and the former editor of BioMechanics.


1. Brown DA, Mulroy S, Sullivan KJ. Strength-training effectiveness post-stroke: results of the STEPS clinical trial. Presented at the Combined Sections Meeting of the American Physical Therapy Association, Boston, MA; February 2007.

2. Moseley AM, Stark A, Cameron ID, Pollock A. Treadmill training and body weight support for walking after stroke. Cochrane Database Syst Rev 2005;(4):CD002840.

3. Teasell R, Foley N, Salter K, et al. Evidence-based review of stroke rehabilitation, 9th edition., accessed 3/28/07. Copyright 2008 Jordana Foster – 24 Kirkland Dr, Stow, MA – Email: – Fax: (815) 346-5239