Repetitive gait training early after stroke
What drives improved walking ability after repetitive
training?
It is essential to consider that the task performance in
the context of stroke rehabilitation can improve either
via restitution of impairments or compensation strategies
(15, 61–63). While the included participants potentially
improved their ability to walk, we do not know how
these changes are achieved, as the FAC does not reflect
whether patients “returned towards more normal pat-
terns of motor control” (i.e. restitution), or learned to
“accomplish the goal through a new approach by the use
of intact muscles, joints and effectors” (62, 64). Since
participants improved their ability to walk without nor-
malization of motor control and strength of the paretic
leg (see Table IV), it seems that it is rather through com-
pensation that the included patients improved. Indeed,
compensation is frequently observed in the recovery of
standing balance (65–67) and walking (68) as patients
adopt an asymmetrical pattern to shift the kinetic control
towards the unaffected side, while normalization of
motor control of the paretic leg is almost lacking (69,
70). However, robots provide practice in a symmetri-
cal pattern, which at first sight appears paradoxical. In
future trials, analyses of the quality of the gait pattern,
including inter-limb coordination in spatiotemporal and
kinetic parameters, should be included to provide defini-
tive evidence on mechanisms underlying effectiveness
of early training (64, 71). This knowledge will have
major implications for practice and designing robots
for rehabilitation (i.e. trying to improve impairments or
teaching compensation strategies) (15, 72).
Future directions for robots in rehabilitation
Despite evidence in support of repetitive training, small
effect sizes are found. These are statistically signifi-
cant, but the clinical importance is questionable, e.g.
an improvement of 24 m on the 6-min walk test does
not exceed the minimal clinically important difference
(50 m (73)). We have to consider that interventions
investigated to date are treadmill- or footplate-based.
This means that massed practice of the same action is
provided while the device controls balance via a sup-
porting harness and gait via the pre-set belt speed (74,
75). Consequently, the patient is simply exposed to
repetitive monotonous movement. However, animal
models established that it is not such exposure to mo-
vement, but skill learning, that guides neuroplasticity
(50). This suggests that stroke rehabilitation requires
a whole different concept of RAGT, where the patient
is constantly challenged in engaging environments
and through variable practice (59). The introduction
of mobile exoskeletons might enable the combination
of high-dose practice through robotic assistance with
the challenging nature of over-ground walking, since
85
the patient has to actively initiate each step and control
their balance, meaning that every step taken is treated
as a novel problem to solve (59, 74). However, while
this intervention may be promising, by exposing the
patient to a learning environment, research on the usage
of such devices is just beginning.
The need to change the current scientific approach
Only 15 studies, of which 3 are dependent follow-up
studies, met the inclusion criteria. Those are mostly
phase I or II trials with small sample sizes. Therefore,
this review agrees with Stinear et al. (76), who found
that less than 10% of clinical trials are initiated in
the first 30 days post-stroke (76). A priority shift in
research towards the first weeks is required (15, 62).
This research requires a new approach (62, 77, 78).
Stratification seems important, since a growing body
of evidence suggests that not all patients have the same
potential to recover (6, 79). Using prognostic models
will help to discriminate between these groups and to
identify those patients who are most likely to benefit
(80), e.g. by assessing muscle strength of the paretic
leg when enrolling participants (6). In addition, our
quantitative analysis is based on post-intervention data,
which means that the process of recovery is measured
as a single outcome score assessed on an arbitrary time-
point. Considering that such trials are taking place in
the background of spontaneous neurological recovery, a
time-dependent process responsible for the majority of
improvements on both body function and activity level
(63, 81), recruitment and assessments of participants
should be performed repetitively and at fixed time-
points (77). This allows us to encapsulate the process
rather than simply the outcome of recovery. Besides
that, the majority of trials describe characteristics of
the interventions poorly. A documentation on the dose
in terms of repetitions is a far more accurate outcome
compared with time scheduled for therapy (23, 53) and
would allow us to quantify the treatment contrast bet-
ween groups to analyse a dose-response relationship in
more detail (53). This review highlights the great need
to shift the selection of outcome measures from scales
simply measuring task accomplishment to those mea-
suring the quality of movement, to gather evidence on
how patients improved when engaged in repetitive task
practice (14, 15, 63). Taken together, not only a priority
shift toward the first weeks is required in rehabilitation
research, but also a corresponding shift in methodology
with a need for more precision in our trials (77).
Conclusion
In total, 15 eligible studies were identified, which are
in general pilot studies with small sample sizes. Con-
sequently, well-designed motor rehabilitation trials
J Rehabil Med 51, 2019