Clinical and Structural Findings in Patients With Lesion-Induced Dystonia: Descriptive and Quantitative Analysis of Published Cases

Abstract

Background and objectives: Brain lesions are a well-recognized etiology of dystonia. These cases are especially valuable because they offer causal insight into the neuroanatomical substrates of dystonia. To date, knowledge of lesion-induced dystonia comes mainly from isolated case reports or small case series, restricting broader description and analysis.

Methods: Cases of lesion-induced dystonia were first identified from a systematic review of published literature. Latent class analysis then investigated whether patients could be classified into subgroups based on lesion location and body regions affected by dystonia. Regression analyses subsequently investigated whether subgroup membership predicted clinical characteristics of dystonia.

Results: Three hundred fifty-nine published cases were included. Lesions causing dystonia occurred in heterogeneous locations, most commonly in the basal ganglia (46.2%), followed by the thalamus (28.1%), brainstem (22.6%), and white matter (21.2%). The most common form of lesion-induced dystonia was focal dystonia (53.2%), with the hand (49.9%) and arm (44.3%) most commonly affected. Of all cases, 86.6% reported co-occurring neurologic manifestations and 26.1% reported other movement disorders. Latent class analysis identified 3 distinct subgroups of patients: those with predominantly limb dystonias, which were associated with basal ganglia lesions; those with hand dystonia, associated with thalamic lesions; and those with predominantly cervical dystonia, associated with brainstem and cerebellar lesions. Regression demonstrated significant differences between these subgroups on a range of dystonia symptoms, including dystonic tremor, symptom latency, other movement disorders, and dystonia variability.

Discussion: Although dystonia can be induced by lesions to numerous brain regions, there are distinct relationships between lesion locations and dystonic body parts. This suggests that the affected brain networks are different between types of dystonia.

Figure 1. PRISMA Flowchart

Systematic search process. Flowchart adapted from the work of Liberati et al. (2009). There were more cases than published reports because some reports contained multiple cases.

Figure 2. Dystonia Symptom Latency

Histogram (A) shows the number of weeks after the lesion that dystonia symptoms first arose for all cases with latency data. Each bar represents 4 weeks. Figure B shows latency data by lesion etiology. Figure C shows latency by lesion location. Boxes represent 25th, 50th, and 75th percentiles, and dots show values falling outside whiskers (1.5 × interquartile range).

Figure 3. Lesion Location and Dystonic Body Region Latent Class Analyses (LCAs)

LCA investigated whether cases could be classified based on lesion location and body distribution (A), as defined by Albanese et al. (2013), or body part (B). There was a relatively high overlap of the classes within the body distribution categories (A), suggesting a limited ability of the model to classify patients into distinct subgroups. The body part LCA demonstrated clearer separation of classes (B), showing that a high percentage of cases fell within one of the 3 dystonia subgroups: (1) limb, (2) hand, or (3) cervical dystonia. Each of these subgroups showed specific relationships to the brain regions affected by lesions. For example, 75% of the limb class had arm dystonia and 68% of these cases had a BG lesion. Significant differences between classes within each category can be seen by nonoverlapping confidence intervals. BG = basal ganglia; WM = white matter.

Figure 4. Class Membership and Dystonia Symptoms

Logistic and linear regression was used to investigate whether class membership (limb, hand, and cervical dystonia classes) could predict dystonia symptoms or case characteristics. Symptom latency panel shows the percentage of cases above the median latency (11 week). * = p < 0.05.