Botulinum neurotoxin-A (BoNT-A) injections are currently the most frequently used clinical intervention for focal spasticity [1,2,3].
Spasticity is a common symptom after various brain and neural injuries,
such as spinal cord injury (SCI) or stroke, referring to an exaggerated
stretch reflex, i.e. stretch hyperreflexia [4, 5].
Spasticity is perceived as an increased joint resistance to movement,
i.e. joint hyper-resistance. BoNT-A injections are used clinically to
reduce muscle activity and hence spasticity [1].
BoNT-A injections reduce muscle activity by inhibiting the release of
acetylcholine at the neuromuscular junction, which chemically denervates
the exposed muscle fibers. BoNT-A effects reduce after 2 to 4 months
due to nerve sprouting and muscle re-innervation [1].
Clinical
evaluation of BoNT-A injections has shown a significant reduction in
joint resistance after 2–8 weeks using the modified Ashworth scale (MAS)
[6,7,8].
With the MAS, currently a common clinical test, clinicians evaluate
overall joint resistance, which can physiologically include tissue
characteristics, and tonic and reflexive muscle activity [5, 9,10,11].
For the MAS, a single passive movement profile is repeatedly applied,
whereas movements with varying characteristics, e.g. slow and fast
velocities, are required to unravel joint resistance contributions.
Therefore, the MAS can clinically only evaluate spasticity indirectly
and cannot distinguish between spasticity and other symptoms as
involuntary background activity, shortened soft tissue, contractures and
muscle fibrosis [4, 12, 13]. Furthermore, the MAS has a questionable reliability, especially when applied at the lower limb [11, 14].
Hence, the clinical effect of BoNT-A injections on spasticity is poorly
understood, while BoNT-A injections are a frequently used clinical
intervention for spasticity.
Quantification of the intrinsic and
reflexive contributions to joint hyper-resistance is essential to
understand the beneficial and adverse effects of BoNT-A injections and
support clinical decision making. BoNT-A injections can, for example,
have side-effects and should ideally only be administered to patients
who suffer from increased reflexive contributions to joint
hyper-resistance [15].
Objective information on both intrinsic and reflexive joint resistance
can support clinical decision making and help evaluate treatment effects
[5].
The intrinsic resistance represents the combination of tissue-related
non-neural and tonic neural contributions to joint resistance [10].
The reflexive resistance, representing the phasic neural contributions,
can be used as measure for spasticity. Model-based processing of
neuromechanical responses can be used to unravel and quantify the
intrinsic and reflexive contributions [10, 16,17,18,19,20].
Furthermore, instrumentation and motorization using robotic devices can
improve precision, consistency and objectivity of the applied movements
and measurements [21,22,23].
Model-based
evaluation of BoNT-A effects on joint hyper-resistance contributions
have been applied using neuromechanical models [24,25,26,27].
These studies showed conflicting results on BoNT-A effects with either
no change or a significant reduction of the reflexive resistance
observed after injection. The neuromechanical modelling approaches used
limited experimental datasets measured over the full passive range of
motion (pROM), similar to current clinical measures. The subsequent
joint resistance estimation primarily relies on a priori knowledge and
simplifying assumptions. As a result, these methodologies are sensitive
to incomplete model definitions and imperfect a priori knowledge [17, 18, 20]. Furthermore, the lack of a gold standard complicates interpretation of the reported conflicting results [5, 28, 29].
Besides the selected model, differences in reported BoNT-A effects may
also be influenced by participant heterogeneity, the experimental setup,
and the assessed joint. Given the conflicting results and lack of a
gold standard, investigating fundamentally different approaches to
assess joint hyper-resistance is of interest to improve understanding of
BoNT-A effects.
An alternative approach to assess BoNT-A effects
on joint hyper-resistance contributions is data-driven modelling.
Data-driven modelling evaluation of BoNT-A effects on joint
hyper-resistance contributions could be executed using system
identification [10, 16, 30, 31].
For example, the parallel-cascade (PC) system identification technique
has shown the ability to discriminate spastic participants from controls
and paretic from non-paretic joints [30, 32].
The PC technique has also shown good group-level responsiveness during
the evaluation of several clinical treatments, like functional
electrical stimulation-assisted walking, Tizanidine and robot-assisted
gait training [33,34,35].
Currently, no system identification results have been reported on
BoNT-A effects. Contrary to neuromechanical modelling, the system
identification techniques previously tested in a clinical setting used
rich experimental datasets measured over only a limited portion of the
pROM [30,31,32,33,34,35].
As intrinsic and reflexive joint resistance depend on joint angle, the
obtained joint resistance estimates do not characterize the full pROM [36].
The
goal of this paper was to distinguish the contribution of intrinsic and
reflexive ankle joint resistance for participants treated with BoNT-A
injections to reduce spasticity. We hypothesized that reflexive joint
resistance decreases 6 weeks after injection, while returning close to
baseline after 12 weeks [24, 25].
Due to the reduced reflexive joint resistance, we also expected the
overall joint resistance to decrease 6 weeks after injection, while
returning close to baseline after 12 weeks [6,7,8].
In absence of a gold standard, the joint resistance contributions were
assessed using multiple joint resistance measures with different
characteristics and limitations. Joint resistance contributions were
estimated using clinical measures (MAS/Tardieu Scale) [9, 37], an instrumented spasticity test (SPAT) [22, 23] and a parallel-cascade (PC) system identification technique [10, 30]. To support validity of the measures used, the linear association between the various outcome measures was investigated.