Tuber Locations Associated with Infantile Spasms Map to a Common Brain Network

Abstract

Objective: Approximately 50% of patients with tuberous sclerosis complex develop infantile spasms, a sudden onset epilepsy syndrome associated with poor neurological outcomes. An increased burden of tubers confers an elevated risk of infantile spasms, but it remains unknown whether some tuber locations confer higher risk than others. Here, we test whether tuber location and connectivity are associated with infantile spasms.

Methods: We segmented tubers from 123 children with (n = 74) and without (n = 49) infantile spasms from a prospective observational cohort. We used voxelwise lesion symptom mapping to test for an association between spasms and tuber location. We then used lesion network mapping to test for an association between spasms and connectivity with tuber locations. Finally, we tested the discriminability of identified associations with logistic regression and cross-validation as well as statistical mediation.

Results: Tuber locations associated with infantile spasms were heterogenous, and no single location was significantly associated with spasms. However, >95% of tuber locations associated with spasms were functionally connected to the globi pallidi and cerebellar vermis. These connections were specific compared to tubers in patients without spasms. Logistic regression found that globus pallidus connectivity was a stronger predictor of spasms (odds ratio [OR] = 1.96, 95% confidence interval [CI] = 1.10-3.50, p = 0.02) than tuber burden (OR = 1.65, 95% CI = 0.90-3.04, p = 0.11), with a mean receiver operating characteristic area under the curve of 0.73 (±0.1) during repeated cross-validation.

Interpretation: Connectivity between tuber locations and the bilateral globi pallidi is associated with infantile spasms. Our findings lend insight into spasm pathophysiology and may identify patients at risk. ANN NEUROL 2021;89:726-739.

Figure 1:. PRISMA Flow Chart identifying TACERN research patients with and without infantile spasms.

Patients were identified from the prospective TACERN cohort study of children with TSC. Inclusion criteria included availability of neuroimaging data of sufficient quality to identify tubers, sufficient clinical follow-up to accurately delineate patients who did or did not develop infantile spasms, and the presence of identifiable tubers on neuroimaging so that tuber locations could be used for lesion network mapping. One hundred and twenty-three patients met these criteria, while 19 were excluded, either due to the lack of available neuroimaging or the absence of detectable tubers.

Figure 2 –. Sample tubers distributions from children with and without a history of infantile spasms.

Tubers were segmented from all 123 children using a semi-automated approach and registered to a common brain atlas (MNI 6th gen. atlas). Example tuber distributions are shown in red from the seventy-four children with infantile spasms (A) and the forty-nine children without infantile spasms (B), demonstrating that both groups included children with high tuber burdens (Participants H1 and H2) as well as low tuber burdens (Participants L1 and L2) that could not be visibly distinguished from one another.

Figure 3 –. Increased tuber burden, but not tuber location, is associated with infantile spasms.

Binary tuber distribution masks were summed for: all children with TSC (A), the cohort of children with infantile spasms (B), and the cohort of children without infantile spasms (C). Both visual comparison and quantitative spatial analysis (D), did not identify a particular pattern for tuber distribution in general, nor that distinguished between children with and without a history of infantile spasms. While overall tuber burden was statistically different between the two cohorts (E), there was not a statistically significant difference between cortical lobe involvement (F).

Figure 4 –. Lesion network mapping identified three brains regions with consistent negative functional connectivity to tuber distributions associated with infantile spasms.

The 74 tuber distributions associated with infantile spasms were registered to a standardized MNI brain template (A). Brain regions functionally connected to each tuber distribution were identified using a large-scale functional connectivity database of young adult participants (B). Overlap of these functional connectivity maps identified three brain regions connected to >95% of tuber distributions associated with infantile spasms: the left and right globus palladi and the cerebellar vermis (C). Consistent connected regions were also identified in two independent subsets (D). Of note, overlap of functional connectivity maps from the 49 children without infantile spasms did not reveal any consistently connected regions. Regions where connectivity was specific to infantile spasms were then identified by voxel-wise two-sample t-test between children with infantile spasms and those without. The conjunction of a mask of these significant voxels and the lesion network mapping analysis above generated a map of regions both sensitive and specific for infantile spams, here shown controlling for genetic etiology as a covariate (E). This process was repeated with an alternate large-scale functional connectivity database of 9-year-old participants identifying consistent results (F).

Figure 5 –. Lesion network connectivity with the bilateral globus pallidus and cerebellar vermis predicts locations where tubers are more likely to cause infantile spasms.

The intersection of connectivity with the globus pallidi (Gpi, blue shading) and connectivity with the cerebellar vermis (red shading) defined a specific network of areas (purple shading) predicted to be highly likely to cause infantile spasms if lesioned (A). As a demonstration, the same four patients shown in Figure 2, two with infantile spasms and two without infantile spasms, and with either high or low tuber burden, are again shown here with tuber burden (in red) compared to the identified network (in purple) (B and C). Among patients with infantile spasms (B), it can be seen that tubers are more likely to overlap with the predicted network (blue circles). Conversely, among patients without infantile spasms (C), tubers largely do not overlap with the predicted network.

Figure 6 –. Logistic regression and statistical mediation analyses finds that globus pallidus connectivity is a stronger predictor of infantile spasms than tuber burden and may mediate the observation that increased tuber burden is associated with infantile spasms.

The performance of predicting infantile spasms from patients’ tuber burden and globus pallidus connectivity (Table 2, Model 5) was assessed using repeated stratified k-fold cross validation. First, the 123 TSC patients were divided into five groups, then a logistic model was trained using four of these groups, e.g., 80% of the data, and tested for accuracy on the remaining hold-out group, e.g., 20% of the data. The classification results of 5,000 iterations of this process, with random shuffling within groups, was used to create Receiver Operating Characteristic (ROC) curves (A). The mean ROC curve for this model is shown here (blue), along with a one standard deviation cloud (grey) compared to chance (red). The Area Under the Curve (AUC) was also calculated to be 0.73 (standard deviation = 0.10). A separate statistical mediation analysis (B) identified that the relationship between tuber burden and infantile spasms (path c) is fully mediated by the serial relationship of tuber burden and globus palladi connectivity (path a) and globus palladi connectivity and infantile spasms (path b), with an indirect effect of ab=0.412, 95% CI (0.114,0.872). When this mediation is taken into account, tuber burden itself no longer independently predicts (path c’) infantile spasms. Importantly, the reverse model found no mediation, i.e., tuber burden does not mediate the relationship between globus pallidus connectivity and infantile spasms.