Clinical characteristics of exploration cohort at Centre 1

At Centre 1, 68 individuals with RE (39 female, 36 left-hemispheric focus, median onset seven years, range 1–51 years) met the diagnostic criteria for RE and were included in this study (Table 1). Brain biopsies that confirmed RE diagnosis were performed in 39/68 (57%) cases. All individuals, except for two, were treated with anti-seizure medication (median number of anti-seizure medication taken as monotherapy or in combinations in the past history 6, range 0–15). Fifty-nine subjects (87%) received at least one type of immunotherapy. Seventeen individuals (25%, median age at onset 5 years, range 2–11 years) underwent hemispherotomy, and postsurgical EEG data were available in 12 cases. No cases of incomplete disconnection were identified in the neuroradiological reviews of postoperative MRI.

Table 1 Overview of data analysed in this studyClinical characteristics of validation cohort at Centre 2

The validation cohort from Centre 2 consists of 59 individuals with RE (35 female, 33 left-hemispheric focus, median onset 6 years, range 2–14 years). Brain biopsies were performed in 12/59 (20%) subjects. All individuals were treated with anti-seizure medication (median number of anti-seizure medication 6, range 2–14). Fifty-three subjects (89%) received at least one type of immunotherapy. Thirty-seven individuals (63%, median age at onset 5 years, range 2–14 years) from London underwent hemispherotomy, with postsurgical EEG available in 34 cases. The use of anti-seizure medication (U = 1708.5, P = 0.86) and immunotherapy (χ2(1) = 0.29, P = 0.80) did not differ between both centres. Significantly more individuals were treated surgically at Centre 2 than at Centre 1 (χ2(1) = 16.9, P < 0.001).

EEG findings in exploration study

For the exploration study at Centre 1, a total of 2882.2 EEG recording hours in 531 samples (56/531 post-hemispherotomy) were analysed, with a median of 54.4 EEG recording hours per subject. In sample-wise analysis of presurgical EEG, contralesional interictal epileptiform discharges were detected in 58/226 (26%) EEG samples with interictal epileptiform discharges (Fig. 2c, dark blue bars). Contralesional seizure onset was found in 18/172 (10%) EEG samples with seizure patterns (Fig. 2e, dark blue bars). When adult-onset cases were excluded, contralesional interictal epileptiform discharges were present in 57/170 (66%) EEG samples with interictal epileptiform discharges (Fig. 2c, light blue bars). Contralesional seizure onset was observed in 17/132 (13%) EEG samples with seizure patterns (Fig. 2e, light blue bars). In case-wise analyses, CEA (at least one EEG sample with contralesional interictal epileptiform discharges or contralesional ictal onset) was found in 30/68 (44%) individuals (Fig. 2d, f, dark blue bars). When adult-onset cases were excluded, CEA was observed in 28/56 (50%) individuals (Fig. 2d, f, light blue bars). There was no significant difference regarding total EEG recording hours between cases with CEA and cases without CEA (U = 216, P = 0.66; without adult-onset cases U = 165, P = 0.14). Inter-rater reliability was κ = 0.19 for the detection of abnormal interictal slowing, κ = 0.44 for interictal epileptiform discharges, κ = 0.53 for ictal patterns, and κ = 0.39 (fair agreement) averaged across all categories.

Fig. 2

Summary of sample-wise and case-wise EEG findings. b Schematic overview over sample-wise and case-wise data representation in this study. In case-wise analyses, only those individuals who showed CEA in at least one sample were included (bilateral synchronous refers to synchronous epileptiform patterns across both hemispheres. When asynchronous epileptiform patterns were detected in both hemispheres, they were counted as ipsilesional and contralesional; data add up to > 100%, since single EEG samples may contain independent, and therefore not bilateral synchronous, foci in ipsilesional and contralesional hemispheres): a interictal slowing, c interictal epileptiform discharges, and e onset of ictal pattern. Percentages of individuals in whom contralesional patterns were seen in the respective categories: d interictal epileptiform discharges, f onset of ictal pattern. g Sample-wise mixed-effects logistic regression results are shown for Centre 1(top, dark blue marker), Centre 1 without adult-onset cases (second row, light blue marker), Centre 2 (third row, yellow marker), and across both groups (bottom, black marker). EEG was set as the dependent variable. Age at disease onset, disease duration, sex, and lesional side were included as covariates. The subject was added as a random effect on datasets separately

EEG findings in validation study

For the validation study at Centre 2, a total of 3503.2 EEG recording hours in 156 samples (37/156 post-hemispherotomy) were included, with a median of 46.5 EEG recording hours per subject. In sample-wise analysis of presurgical EEG, contralesional interictal epileptiform discharges were detected in 13/112 (12%) EEG samples with interictal epileptiform discharges (Fig. 2c, yellow bars). Contralesional seizure onset was found in 4/79 (5%) EEG samples with seizure patterns (see Fig. 2e). In case-wise analyses, CEA was found in 8/59 (14%) individuals (Fig. 2d, f, yellow bars). There was no significant difference in total EEG recording hours between cases with CEA and those without CEA (U = 254, P = 0.27). The frequency distributions between both centres did not differ significantly about interictal epileptiform discharges (χ2(2) = 5.82; P = 0.055) and ictal onset (χ2(2) = 2.99; P = 0.22), but concerning abnormal interictal slowing (χ2(2) = 19.60; P < 0.001). The frequency distributions between the cohort from Centre 1 without adult-onset cases (childhood and adolescence cases only, onset < 19 years) and the cohort from Centre 2 only showed a trend-level difference regarding ictal onset (χ2(2) = 5.57, P = 0.062), but were significantly different concerning abnormal interictal slowing (χ2(2) = 13.90, P < 0.001) and interictal epileptiform discharges (χ2(2) = 12.65, P = 0.002).

Mixed effects logistic regression of contralesional epileptiform activity

In a sample-wise mixed-effects logistic regression model across both centres, the presence of CEA was significantly associated with a younger age at onset (OR = 0.9; 95% CI 0.83–0.97; P = 0.006), but not disease duration (OR = 1.0; 95% CI 0.95–1.1; P = 0.81) (Fig. 2g). When analysing interictal and ictal patterns separately, both interictal epileptiform discharges (OR = 0.90; 95% CI 0.84–0.98; P = 0.011) and ictal onset (OR = 0.83; 95% CI 0.64–1.1; P = 0.19) were more likely with younger age at onset; however, the latter effect did not reach statistical significance. From this analysis, however, it cannot be said whether contralesional interictal or ictal patterns ooccur first.

Post-hemispherotomy outcome

At Centre 1, 11/17 (65%; 7/12 [58%] with preoperative CEA, 4/5 [80%] without preoperative CEA) and at Centre 2, 28/37 (76%; 3/5 [60%] with preoperative CEA, 25/32 [78%] without preoperative CEA) were seizure-free after hemispherotomy. ILAE outcomes (1-2-3-4-5-6) at Centre 1 were 7-0-2-2-1-0/12 in cases with preoperative CEA (7 with ILAE outcome 1; 0 with ILAE outcome 2; 2 with ILAE outcome 3;…), and 4-0-1-0-0-0/5 in cases without preoperative CEA. At Centre 2, ILAE outcomes were 3-0-0-2-0-0/5 in preoperative CEA cases and 23-0-5-3-1-0/32 in cases without preoperative CEA. In a mixed-effects logistic regression model across both centres, becoming seizure-free after hemispherotomy was significantly less likely if CEA was present before surgery (OR = 0.69; 95% CI 0.50–0.95; P = 0.029). When analysing interictal and ictal patterns separately, postoperative seizure freedom was significantly less likely if preoperative contralesional interictal epileptiform discharges were present (OR = 0.70; 95% CI 0.50–0.98; P = 0.045), but not associated with preoperative contralesional ictal onset (OR = 0.91; 95% CI 0.57–1.45; P = 0.70). Contralesional interictal epileptiform discharges persisted in 6/12 (50%, Centre 1) and 2/34 (6%, Centre 2) of all hemispherotomy cases with postoperative EEG available.

Results of morphometric MRI analysis

T1-weighted MRI scans of 52 subjects and DTI scans of 14 subjects from Centre 1 were included. We observed significantly (FWE-corrected P < 0.05) lower cortical thickness of the temporoparietal junction and postcentral gyrus (Fig. 3a), as well as higher surface area of the insular cortex, temporoparietal junction and temporal pole of the contralesional hemisphere in individuals who presented at least once with CEA than in individuals who never presented with CEA (Fig. 3b). There was no significant difference between cases with CEA and without CEA included in the surface-based analysis regarding disease duration at MRI (U = 423.5, P = 0.94), number of anti-seizure medication (U = 304, P = 0.69), and immunotherapy (χ2(1) = 0.99, P = 0.32).

Fig. 3

Results of MRI morphometry. a Significant clusters (FWE-corrected P < 0.05) indicating lower cortical thickness in individuals with CEA in at least one EEG sample as compared to individuals without CEA were observed in the temporoparietal junction and postcentral gyrus (blue). b Significant clusters (FWE-corrected P < 0.05) indicating higher cortical surface area in individuals with CEA located in the temporoparietal junction, insular cortex, and temporal pole (red). Significant clusters in (a, b) are visualised on the fsaverage_sym template. c Visualisation of fibres with lower (uncorrected P < 0.05) fibre density in the periinsular and parietal white matter in the contralesional hemisphere of individuals with CEA. d Display of fibres with lower (uncorrected P < 0.05) fibre cross-section in the anterior body of the corpus callosum, the superior longitudinal fascicle, and the cingulum of the contralesional hemisphere, as well as in the pyramidal tract of the lesional hemisphere in individuals with CEA. Thresholded streamlines in (c, d) are shown on the study-specific template; grey values represent the magnitude of the fibre orientation distributions. A: anterior, I: inferior, L: left, P: posterior, R: right, S: superior

Regarding white matter integrity, we observed lower (uncorrected P < 0.05) fibre density in the contralesional parietal and periinsular white matter (Fig. 3c). We found lower fibre cross-section in the anterior body of the corpus callosum, the superior longitudinal fascicle, and the cingulum of the contralesional hemisphere, as well as in the pyramidal tract of the lesional hemisphere in individuals with CEA as compared to individuals without CEA (Fig. 3d). There was no significant difference between cases with CEA and without CEA included in the fixel-based analysis regarding disease duration at MRI (U = 12.5, P = 0.20), the number of anti-seizure medication (U = 27, P = 0.59), and immunotherapy (χ2(1) = 1.75, P = 0.19).

Neuroinflammation

In the semiquantitative approach, inflammation was classified as weak-strong in 4-18 individuals with CEA (4 with weak inflammation; 18 with strong inflammation) and 1-16 individuals without CEA. T-lymphocytes were classified as few-intermediate-strong-perineuronal in 1-5-6-10 individuals with CEA and 0-2-6-9 in individuals without CEA. Microglia were classified as few-intermediate-strong-nodules in 0-4-6-12 individuals with CEA and 0-2-5-10 in individuals without CEA. We did not observe any significant group difference in markers of inflammation (OR = 0.28; 95% CI 0.028–2.78; P = 0.28), T-lymphocytes (OR = 0.25; 95% CI 0.062–2.04; P = 0.25), or microglia count (OR = 0.60; 95% CI 0.096–3.74; P = 0.58) in ipsilesional brain biopsies between individuals who had at least once CEA and individuals who never showed CEA.

Neuropsychological performance

Neuropsychological assessment at Centre 1 was available for 47 people with RE. Individuals with CEA performed significantly worse (strongly impaired, impaired, or borderline) than individuals who never showed CEA concerning intelligence (OR = 5.19; 95% CI 1.28–21.08; P = 0.021) and verbal memory (OR = 10.29; 95% CI 1.97–53.85; P = 0.006). While this equally holds for individuals with either left or right lesional hemispheres regarding intelligence, the finding relating to verbal memory is mainly driven by individuals with lesions in the left hemispheres. A general trend is confirmed in all other cognitive domains (attention, visual memory, visuospatial abilities, and language): More individuals with CEA were impaired within the respective category than individuals without CEA (Fig. 4).

Fig. 4

Neuropsychological deficit in relation to the presence of contralesional epileptiform activity. Odds ratios for neuropsychological deficits when CEA is present were evaluated for six cognitive domains. Odds ratios were calculated for all people with Rasmussen’s encephalitis (n = 47, dark blue markers), as well as for individuals with left lesional hemispheres (n = 25, red markers) and right lesional hemispheres (n = 22, green markers) separately