Literature Review on Lesion-Symptom Mapping and Deep Brain Stimulation for Poststroke Spasticity: Restoring a Dysfunctional Network? Open Access

Introduction: The present study aimed to define a structural network of stroke-induced and spasticity-related lesions and to relate this network to target sites and reported effects of deep brain stimulation (DBS) to treat poststroke spasticity. Methods: The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed (online suppl. Table 2). We performed two separate systematic literature reviews, collecting data from previously published voxel-based lesion-symptom mapping (VLSM) studies for poststroke spasticity patients searching the Medline database on Pubmed using the keywords “stroke,” “spasticity,” and “lesion mapping” as well as data from previously published cohorts undergoing DBS for poststroke spasticity using the keywords “brain stimulation” and “spasticity.” Data collected from each study included patient demographic characteristics, stroke diagnosis, movement disorder, DBS target, stimulation parameters, complications, and outcomes. Data from VLSM studies were used to calculate coordinate-based activation likelihood clusters, which were then used as seeds for enhanced fiber tracking to analyze affected networks. Results: Data from five studies on voxel-based lesion-symptom mapping for stroke-induced spasticity were included in the analysis. Meta-analytical mapping of stroke-related lesions identified significant clusters located in the basal ganglia-thalamo-cortical network which were predominantly connected to the sensorimotor cortex. We identified eight studies (four retrospective case series, two prospective open-label non-randomized trials, and two prospective double-blind trials) fulfilling our inclusion and exclusion criteria on DBS for spasticity reporting on 107 patients in total. Most studies reported outcomes in patients with cerebral palsy, a condition associated with both stroke-related spasticity and hypertonia-related dystonia, which are difficult to differentiate clinically. Target sites included different parts of the cerebellum and the motor thalamus with overall mixed results. Conclusion: Because all reported effective DBS target sites are situated along the cerebello-thalamo-cortical network, we hypothesize that the therapeutic effect of DBS on spasticity might be induced by resetting a functional imbalance between the basal ganglia-thalamo-cortical and the cerebello-thalamo-cortical networks in patients with a supraspinal etiology of spasticity. However, the results need to be interpreted cautiously due to the inevitable inclusion of stroke-related dystonia.

Spasticity is a neurological condition characterized by spasms, clonus, hyperreflexia, and increased muscle tone which may occur after cerebrovascular accidents and has a considerable impact on the quality of life in patients [1]. With the estimated prevalence of poststroke spasticity being up to 25.3%, it represents a significant clinical and socioeconomic burden, underscoring the need for effective management strategies [2, 3]. Common medical therapy includes the administration of oral or intrathecal GABA agonists, such as baclofen. The pathophysiology of spasticity is only partly understood, with various models being proposed, acknowledging the interplay of both supra- and infraspinal factors [4‒8].

Recent magnetic resonance imaging volumetric studies aimed to find a correlation between stroke location and the development of spasticity based on lesion-symptom mapping models, revealing a rather extended distribution of lesions [9‒13]. According to these studies, the most frequently affected regions included parts of the basal ganglia, caudate nucleus, thalamus, neocortical areas including the insula and premotor cortex, and white matter tracts including the corona radiata and internal capsule [9‒13].

Thus, a more concise area, or well-defined structural network that if affected by stroke induces spasticity has yet to be defined. From a therapeutic point of view, deep brain stimulation (DBS) of the cerebellum and thalamus has emerged as a potentially effective therapy in drug-refractory spasticity. However, due to heterogeneity in treatment techniques, outcome assessments, and the limited sample sizes in these pilot trials, there’s no consensus on where to simulate and which stimulation paradigms to choose. Thus, studies with a larger cohort size are needed to establish more conclusive results. Furthermore, it remains unclear to this point how the effects of DBS relate to networks involved in the pathogenesis of spasticity. Thus, a network approach of spasticity and its related therapies has not yet been addressed.

The present study aimed to define a structural network of stroke-induced and spasticity-related lesions and to relate this network to target sites of neuromodulation. To this end, we first aimed at defining a network of spasticity based on structural connectivity analysis of a lesion sweet spot which we derived from a meta-analysis of stroke lesions that precipitate spasticity, by summarizing and aggregating data of previous studies on lesion-symptom mapping for poststroke spasticity. Second, we aimed at evaluating the effectiveness of DBS for spasticity by performing a comprehensive review of the literature and to summarize the simulation targets that have been chosen for this condition along with its reported clinical effectiveness. By comparing the meta-analytic structural network affected by lesions inducing spasticity with target sites of brain stimulation and their documented therapeutic effect, this study could provide the basis for a better understanding of the pathophysiology of spasticity and identify promising target sites for future studies.

We collected data from previously published voxel-based lesion-symptom mapping (VLSM) studies for poststroke spasticity patients. We performed a systematic literature review using the MEDLINE PubMed database, using the keywords “stroke, spasticity, lesion mapping” to search for cohort studies from inception to June 2023, there was no predefined timeframe. Studies were eligible for inclusion if they (a) were originally published in English, (b) were original articles (clinical trials), (c) had an abstract, and (d) reported on VLSM for poststroke spasticity. We excluded studies with no available abstract, studies published in languages other than English, review articles with no original data, and studies with no reporting on VLSM for poststroke spasticity. We extracted the following data from each article (if available): Descriptive anatomical region associated with spasticity, the corresponding MNI coordinates of the center of mass of each cluster, the z-score of those lesions, and the number of voxels surviving a false discovery rate of p < 0.05. If coordinates were provided in the Talairach space, a transformation into MNI coordinates was conducted on bioimagesuiteweb.github.io/webapp [14]. Additionally, clinical data from patients (information about age, sex, time from stroke, lesion side, handedness/motor dominance, lesion voxels, and stroke pathology) was extracted.

For the meta-analysis of brain lesions that are associated with spasticity, we extracted the MNI coordinates of the center of mass of each voxel cluster as well as the number of spastic patients (=n) reported in the individual studies. All extracted foci as well as the corresponding sample size were used to calculate coordinate-based activation likelihood estimation meta-analysis within GingerALE [15]. The threshold p value was set at 0.0001 and the minimum volume (mm³) was set at 200. The generated NIfTI-files were used for further image processing (see below) and the percentage of gray matter structures each cluster was located in was noted. The NIfTI-files were imported into DSI Studio [16] for normative tractography analysis of a template based on 1,065 subjects derived from the Human Connectome Project [17]. The meta-analytical clusters served as seed regions and tractography was conducted using a deterministic fiber-tracking algorithm [18]. The tracking parameters included a step size of 0.5 mm, fiber length ranging from 50 mm (minimum) to 150 mm (maximum), an angular threshold of 40°, and a tracking threshold of 0.1. Each meta-analytical cluster was used as a seed region to generate 100,000 streamlines to the rest of the brain. Figure 1 illustrates the process of the tractography analysis. The tract-to-region connectome was visualized by color-coding the Brodmann area (BA) into a “heat” color scheme according to the relative ratio of total cortical endpoints to each BA.

Due to the small number of original studies exclusively concerned with DBS in poststroke spasticity, this systematic review also incorporated studies on cerebral palsy (CP). Given its similar or often identical etiology when resulting from perinatal stroke [19], the studies of CP were included to provide a broader perspective on the application of DBS in spasticity management following central nervous system injury. In some studies that met our inclusion criteria, DBS was also evaluated for other neurological conditions, such as primary or acquired dystonia. However, these conditions were excluded from our analysis to focus on spasticity. We collected data from previously published cohorts undergoing DBS for spasticity. To this end, we performed a systematic literature review using the MEDLINE PubMed database, using the keywords “brain stimulation, spasticity” to search for cohort studies from inception to August 2023. Studies were eligible for inclusion if they (a) were originally published in English, (b) were original articles (clinical trials), (c) had an abstract, and (d) reported on outcomes of DBS for poststroke spasticity and/or CP. We excluded studies with no available abstract, published in languages other than English, review articles, studies reporting on spinal cord stimulation, transcranial magnetic stimulation, and studies that reported no outcome regarding spasticity or CP. We extracted the following data from each article (if available): information about age, sex, race/ethnicity, primary diagnosis that led to DBS implantation, DBS implantation site (descriptive), target coordinates, the chronic DBS stimulation settings, reported follow-up time and reported outcome.

We identified a total of five studies on voxel-based lesion-symptom (VLSM) mapping fulfilling our inclusion and exclusion criteria (online suppl. Table 1 [A]; for all online suppl. material, see https://doi.org/10.1159/000545888). The included studies are summarized in Table 1. Out of these five studies we extracted a total of 55 foci of a corresponding number of 172 patients with spasticity for this meta-analysis. This resulted in the identification of 2 clusters, which were significantly associated with spasticity. The clusters and their anatomical extent are shown in Figure 2. Cluster 1 (marked red in Fig. 2) was located in the thalamus (78.2%) and body of the caudate nucleus (21.8%). Large parts of the affected thalamus (34.7%) corresponded to the ventral lateral nucleus (VL). As far as white matter is concerned, the cluster was partly in the central thalamic radiation and the corticospinal tract. Cluster 2 (marked yellow in Fig. 2) was located in the pallidum (76.7% of the neurons allocated to the putamen and 23.3% to the lateral globus pallidus). Furthermore, the cluster covered the central thalamic radiation and the corticospinal tract. Additional information about the clusters is provided in Table 2. Following structural connectivity network analysis, BA 1–3 was predominantly connected to the lesion clusters (53.8% of all fibers), followed by BA 20 (15.9% of all fibers), BA 40 (10.5% of all fibers), BA4 (8.7% of all fibers) and BA13 (5.7% of all fibers). BA 6 and BA 16 were only weakly connected to the lesion clusters (<5%). The results are visualized in Figure 3.

From the 531 articles screened, eight were eligible and included for further analysis (online suppl. Table 1 [B]). A detailed overview of studies reporting on DBS for poststroke spasticity is provided in Table 3. Of the eight included studies, four were retrospective case series [20‒23], two were prospective open-label, non-randomized trials [24, 25] and two were prospective double-blind trials [26, 27]. The studies reported on outcomes of 107 patients in total with an age range of 3–59 years. Regarding the target sites of DBS, four studies targeted different parts of the cerebellum, including the anterior cerebellar lobe [21, 23], superior cerebellar peduncle (SCPs) [22] as well as the dentate nucleus (DNs) [22, 24, 26, 27]. Two studies reported on thalamic DBS [20, 25] with target sites in the VL [20], the ventral oral posterior nucleus (Vop) or the ventral intermediate nucleus (VIM) of the thalamus [25]. The reported stimulation parameters varied considerably and ranged from 1 to 8 V, 60–1,000 μs, and 50–200 Hz. Importantly, there were differences between studies regarding the time frame of stimulation and follow-up times (varying from no reported follow-up time to several years). Whereas all reported studies of thalamic stimulation reported on outcomes after chronic stimulation, cerebellar stimulation occurred only chronically in one study [26], transiently (up to 4–8 weeks) in one study [21] and intermittently with periods lasting several hours in two studies [22, 27].

Regarding included patients clinical profiles, the primary indication for DBS was spasticity following CP in seven studies [20‒23, 25‒27] and spasticity following stroke in two studies [20, 24], with one study excluding severe spasticity (defined as MAS = 4) [24]. Outcome measurements varied considerably between studies varying from clinical examination with no objective measurement [20, 21, 26, 27] to assessing spasticity with validated clinical scales such as the MAS [22, 23, 25] or the Fugl-Meyer Assessment [24]. In studies where an objective outcome measure was reported, the relative change in spasticity varied considerably across studies from −38.3% to +39.42% compared to baseline.

More specifically, reported outcomes for the anterior cerebellar lobe were an evident reduction during stimulation with a slight return of spasticity after stimulation [21] and an average decrease in spasticity of 38.33% in the upper extremities and an average reduction of 32.5% in the lower extremities [23]. For the VL, 57% of the patients (28 out of 49) showed fair to good effects on spasticity [20]. For the SCPs and DNs reported outcomes were a 66.6% reduction in spasticity in the right upper limb and 50% in the left upper limb. On the lower left limb, the reduction in spasticity was 25%, while the right lower limb remained unchanged [22]. For the DNs a 30.6% increase on the Fugl-Meyer scale was reported [24] while other studies observed a modest but inconsistent reduction in spasticity [26, 27]. For the Vop and the VIM, an increase in spasticity of 39.74% in 3 out of 4 patients and a decrease of 21.05% in the other one was reported [25]. All in all cerebellar stimulation showed modest improvements in spasticity but was not consistently effective across all patients, with placebo effect playing a significant role in the outcomes, while thalamic stimulation showed ambiguous effects.

To our knowledge, this is the first study to combine lesion-network meta-analysis of poststroke spasticity and a systematic literature review on DBS for poststroke spasticity to assess their general clinical efficacy and to compare the anatomical relation of different target sites of neuromodulation toward this spasticity-related network. We found several noteworthy findings:

First, our meta-analytical map of stroke lesions associated with spasticity indicated that the significant clusters were all located in the basal ganglia-thalamo-cortical network and were connected predominantly to the sensorimotor cortex and only to a much lesser degree to the premotor area.

Second, DBS stimulation to treat spasticity was predominantly conducted in the cerebellum and different parts of the motor thalamus that predominantly receive cerebellar afferent inputs such as the VIM or ventral lateral posterior nucleus, respectively. Due to the heterogeneity across studies regarding different included clinical patient profiles with varying indications (adult poststroke spasticity, CP, dystonia, MS, etc.), varying reported outcome measures, inconsistencies of stimulation paradigms (chronic versus short-term or intermittent stimulation) and follow-up times, the overall small sample size of individual case series, and a general lack of blinded outcome assessments, no firm conclusions can be drawn about the clinical efficacy of this therapy nor about what stimulation site might be most effective. However, there seems to be a clear trend throughout all included studies that DBS of the cerebello-thalamic network might be effective in some patients (Level 4 evidence). As our lesion-network meta-analysis for poststroke spasticity suggests an involvement of the basal ganglia-thalamo-cortical network predominantly involving the sensorimotor cortex, there is a potential crosslink between the stroke-inducing network and the therapeutic DBS-related network involving the cerebello-thalamo-cortical network as the cerebello-thalamo-cortical network projects predominantly to the sensorimotor region, as visualized in Figure 4.

Meta-analysis of voxel-based lesion-symptom mapping (VLSM) in poststroke spasticity patients revealed two lesion hotspots. Most fibers affected by these lesions have their endpoint in the sensorimotor cortex (BA 1–3, 4) and to a lesser extent in sensory association areas (BA 6, 13, 40).

This is in line with findings by Sherman et al. [4], who observed that ischemic lesions solely located in the pyramidal tract did not lead to noticeable spasticity during clinical examination. Moreover, they noted that a lesion in the medullary pyramid could result in weakness and hyperreflexia, but it was not necessarily linked to spasticity. The authors suggested that there might be other anatomical regions, apart from the pyramidal tract that needed to be affected to induce hyperreflexia and spasticity [4]. Our meta-analysis showed that there were also a large number of fibers affected from the central thalamic radiation, which mostly project to the somatosensory cortex (BA 1–3). Giving further context to that observation, the research findings of Priori et al. indicate that spasticity occurs when cortical lesions also affect non-primary motor areas. In cases where the lesions are located below the motor cortices, spasticity is only induced when these lesions affect the corticoreticulospinal fibers. These fibers connect motor areas to the inhibitory medial bulbar reticular formation and spinal motoneurons, forming inhibitory synapses on the latter. Consequently, it is suggested that a reduction in inhibitory signals to spinal motoneurons may lead to increased facilitation of excitatory inputs, potentially contributing to spasticity [5].

Our meta-analysis of VLSM identified two voxel clusters in which lesions were highly associated with the development of spasticity. Enhanced fiber tracking showed that the most affected network was the basal ganglia-thalamo-cortical network with most afferences connected to the sensorimotor cortex.

Most targets for DBS to treat spasticity included thalamic nuclei and cerebellar structures. Targeted cerebellar structures were the anterior lobe of the cerebellum, the SCP and the DN. A majority of studies demonstrated some reduction in spasticity, irrespective of the stimulated target. Nevertheless, the studies employed diverse scales and methodologies to quantify the outcomes, rendering it very difficult to discern any differences in effect among the various outcomes assessed. All the stimulation sites cover the same network that is the cerebello-thalamo-cortical network which partly overlaps with the basal ganglia-thalamo-cortical network affected by spasticity-inducing lesions [29‒31].

As Nowacki et al. [32] wrote in their article, deep cerebellar nuclei (e.g., the DN) primarily send projections to motor-related thalamic nuclei, including the VIM and ventral lateral posterior nucleus. These nuclei are key relay centers that then project to motor cortical areas, including the primary motor cortex (M1) [32]. As early as 1981, Nieuwenhuys [33] stated that the basal ganglia, specifically the GPi, mainly project to the ventral orale anterior nucleus or ventral lateral anterior nucleus, which in turn projects predominantly to pre-motor areas. Consequently, the two networks overlap in the motor thalamus as both send projections to the ventral lateral (VL) and ventral intermediate (VIM) nucleus of the thalamus, as well as the primary motor cortex (BA 4), since the VL and VIM both project to motor cortical areas [34, 35].

Targeted thalamic nuclei included the VL, Vop and VIM. The thalamic stimulation sites encompass distinct neural networks: the VL is part of the cerebello-basal ganglia-thalamo-motor cortical network, the Vop is associated with the motor control network, cerebellar network, as well as basal ganglia-thalamo-cortical circuit, and the VIM, which is a common target for DBS in cases of refractory tremor, is located within the dentato-thalamo-cortical network. The outcomes were not uniformly consistent; some indicated a decrease, while others demonstrated an increase in spasticity. Furthermore, the studies utilized varied scales and methodologies to measure outcomes, making it challenging to distinguish differences in effect across the various outcomes evaluated. As Cooper et al. [20] noted in 1981 in their article, DBS of the thalamic ventrolateral nucleus, provides some positive effects in regards to reducing spasticity, although the benefits were not quantified and not objectively measured. The same stimulation location was used by Luciano et al. [25] in 2021 with additional stimulation of the ventrointermediate nucleus. In contrast to the work of Cooper et al. [20], their intervention showed contradictory results, with increasing spasticity in a majority of patients.

The network affected by lesions that induce spasticity, specifically the basal ganglia-thalamo-cortical network, and the DBS-targeted cerebello-thalamo-cortical network are distinct networks. However, at the level of the thalamus, there is an overlapping innervation pattern of cerebellar and basal ganglia input structures. Taking all the findings mentioned above into consideration, our hypothesis is that high-frequency stimulation of the cerebello-thalamic network might reduce spasticity, by resetting a functional imbalance between the basal ganglia-thalamo-cortical and the cerebello-thalamo-cortical network in patients with a supraspinal etiology of spasticity.

This study has several limitations. First, the overall sample size in the stimulation setting of DBS and the VLSM poststroke spasticity studies is still relatively small. Second, the level of evidence of the underlying clinical studies is low, with a high risk of bias and missing placebo control group. In addition to that, the wide use of the MAS scale to assess spasticity makes findings susceptible to subjectivity because of the subjective nature of the scale.

Therefore, our evaluation of DBS stimulation sites can only report on what has been tested rather than determine its clinical effectiveness. Moreover, the inquiry into whether there exists an ideal stimulation location for poststroke spasticity still eludes us at this juncture.

Another aspect that warrants attention is that the distinction between spasticity-related rigidity and dystonia-related rigidity has proved to be difficult, especially in the case of people with CP, for whom the two conditions often co-exist [36]. As rigidity seen in clinical evaluations may result from either or both systems and due to the shared nature of the symptoms between the illnesses, diagnosis and management is often challenging [37]. This emphasizes the need for well-designed studies which specifically seek to differentiate spastic and dystonic rigidity.

Despite this, this work aggregated clinical data, assessed the effect of DBS on spasticity and suggested a connectivity profile based on VLSM studies, which set the foundation of a future multicenter, randomized controlled trial analyzing the effects of DBS for poststroke spasticity.

This study supports the concept that stimulation of the cerebello-thalamo-motor cortical network is a promising therapeutical approach for improving spasticity after stroke rehabilitation. Furthermore, this study identified a possible stroke-inducing structural lesion network. According to our findings, we hypothesize that high-frequency stimulation of the cerebello-thalamic network reduces spasticity, by resetting a functional imbalance between the basal ganglia-thalamo-cortical and the cerebello-thalamo-cortical networks in patients with a supraspinal etiology of spasticity. This study therefore lays the groundwork for future large-scale (multicenter, randomized controlled trials) that will investigate the impact of DBS on post-stroke spasticity.

This study is a systematic review and exclusively utilized preexisting data from published studies. As such, no new patient data were collected, and no patient consent was required. Additionally, no ethical approval was necessary as the analysis was based solely on previously published and publicly available data.

None of the authors have any conflicts of interest to declare that are relevant to this study.

This study was conducted without any financial support or external funding.

L.F., C.P., and A.N. contributed to the conception and design of the study. L.F., D.Z. and A.N. contributed to the acquisition and analysis of data, drafting the manuscript, and preparing the figures.

All data generated or analyzed in the course of this study are presented within this article. For additional inquiries, please contact the corresponding author.