The predictive value of interictal scalp EEG findings in aiding the detection of malformations of cortical development in temporal lobe epilepsy and impact on surgical planning

Patient Population

We retrospectively identified all adult patients (>18 years of age) who were treated for medically refractory TLE by surgical resection in the Department of Neurosurgery of Northwestern University between January 2007 and December 2017. Surgical procedures performed include temporal lobectomy with or without amygdalohippocampectomy or resection of a specific seizure onset node (topectomy) within the temporal lobe (see Table 1 for breakdown by surgical procedure). Four patients were excluded who had a previous intracranial surgery for epilepsy, as well as those who did not have pre-operative video electroencephalography (VEEG) or epilepsy protocol MRI.

Patients were identified to have temporal lobe epilepsy using diagnostic procedures including surface VEEG with or without the use of sphenoidal electrodes. All patients with diagnosed temporal lobe epilepsy had a pre-operative brain MRI performed in accordance with the epilepsy protocol at Northwestern University. While this protocol has evolved over time, all epilepsy protocol MRIs included axial and coronal T1, T2 and fluid attenuated inversion recovery (FLAIR) sequences at minimum. Tissue samples were obtained at the time of surgery and analyzed by expert neuro-pathologists at Northwestern University. Data collection for this study was approved by the Institutional Review Board (IRB) of Northwestern University (IRB#: STU00205860), and patient records were reviewed in accordance with the guidelines of the IRB. Patient consent was neither required nor sought because only deidentified data were collected.

Study Design

Patient charts including pre-operative VEEG reports, pre-operative epilepsy protocol MRI reports, surgical reports, and surgical pathology reports were reviewed. Classification of EEG abnormalities was performed using reports from pre-operative VEEGs. Recognizing that the interpretation of EEG data can be subjective based on the experience and background of an individual reviewer, and that the language used in the description of EEG findings is based heavily on the reviewer’s training and institutional preferences, interictal abnormalities were classified according to common terms which were used to describe identified abnormalities. The descriptive terms which were identified for inclusion in this study were “continuous slowing,” “continuous sharp waves,” and “rhythmic slowing” as determined by board certified clinical epilepsy neurologists interpreting the VEEG study. A rhythmic interictal abnormality was identified by use of the term “rhythmic” in the body and/or impression of the study report and a continuous abnormality was identified by the term “continuous,” “frequent,” “abundant,” or having a frequency greater than one occurrence every thirty seconds as described in the body and/or impression of the study report. Example EEG images of interictal abnormalities which are related to the descriptor terms described above from the patients in the cohort can be seen in Figure 1. We specifically chose to use the initial impression of the treating epilepsy neurologist as documented in the clinical interpretation of the VEEG study rather than retrospectively reviewing the VEEG in the context of the pathologic findings in order to bolster the clinical utility of this finding on initial review of EEG data.

• Download figure
• Open in new tab

Figure 1.

This figure shows example EEGs corresponding to the use of descriptive terms of A) continuous slowing, B) continuous sharp waves and, C) rhythmic slowing.

Findings from the pre-operative epilepsy protocol MRI reports were classified based on the interpretation of the reading neuroradiologist at the time of image acquisition. The classification scheme included identification of possible focal cortical dysplasia (FCD), hippocampal sclerosis (HS), infarct/lesion, or tumor. Again, the initial impression of the neuroradiologist as documented in the MRI report was selected for inclusion in our analysis.

Finally, findings from the surgical pathology report were classified based on the type of pathology identified. The classification scheme used was presence of FCD, mMCD, HS, temporal gliosis, infarct/encephalomalacia, or tumor. mMCDs were defined by normal cortex with excess ectopic or heterotopic neurons in the molecular layer or subcortical white matter according to the 2004 Palmini or 2011 Blümcke classification systems if not stated directly by the neuropathologist.^{16, 26}

Classification of MRI, EEG, and pathology findings as described above was performed independently by two members of the study group (S.K.B and J.W.F). No discrepancies in classification of findings were identified.

Data are reported as counts for categorical variables and mean ± standard deviation for continuous variables. Microsoft Excel 2013 (Microsoft, Redmond, Washington) was used to conduct all statistical analyses.

Demographics

We identified 34 patients who met inclusion criteria, 21 (60%) of whom were female, with an average age at time of surgery of 40.6 ± 13.2 years. The average age of onset of epilepsy was 20.9 ± 16.1 years and the average duration of seizures was 19.7 ± 16.8 years. A comprehensive review of the surgical procedures performed can be found in Table 1. Six patients had isolated FCD, two patients had isolated mMCD, one patient had both FCD and mMCD, eight patients had FCD co-occurring with another pathology, nine patients had mMCD co-occurring with another pathology, and eight patients had no evidence of MCD on histologic examination.

MRI, Pathology, and VEEG Findings

Table 1 gives the MRI findings for each patient that were identified to be a possible cause of epilepsy along with the surgical pathology findings for each patient. The classification of VEEG reports, according to the interictal findings, compared against each category of histopathological finding can be found in Table 2.

Identification of Malformations of Cortical Development on MRI versus Pathology

Of the five patients identified to have possible MCD on pre-operative MRI, all five had a confirmed diagnosis of FCD on histopathology. However, 10 patients (66.7%) were found to have FCD on pathology that was not identified on pre-operative MRI. None of the 12 confirmed cases of mMCD on pathology were identified on pre-operative MRI.

EEG Findings as a Predictor of Malformations of Cortical Development

The finding of an interictal abnormality, including rhythmic or continuous slowing or sharp waves was seen in 22 of the 34 patients (64.7%) undergoing temporal lobe surgery for refractory epilepsy. A rhythmic or continuous interictal abnormality was seen in 12 (80%) of the 15 patients with FCD and 7 (58.3%) of the 12 patients with mMCD. Overall, a rhythmic interictal abnormality was seen in 19 (73%) of the 26 patients with MCD (FCD or mMCD). A rhythmic or continuous interictal abnormality was seen in 3 (37.5%) of the 8 patients without evidence of MCD.

The finding of a rhythmic or continuous interictal abnormality corresponds to a sensitivity of 80% and specificity of 47.4% for the presence of FCD and a sensitivity of 58.3% and a specificity of 31.8% for the presence of mMCD. The finding of a rhythmic or continuous interictal abnormality on VEEG to predict the presence of any type of MCD (FCD or mMCD) has a sensitivity of 73.1% and a specificity of 62.5%. The positive predictive value for the presence of MCD on pathology given the finding of a rhythmic or continuous interictal abnormality in this cohort was 86.4%.

A continuous interictal abnormality was more common on EEG than was the finding of a rhythmic interictal abnormality with 19 patients having continuous interictal slowing or sharp waves while only 7 patients had rhythmic interictal slowing. The finding of a continuous interictal abnormality alone corresponds to a sensitivity of 65.4% and specificity of 75% for the presence of any type of MCD (FCD or mMCD). The positive predictive value of any MCD in the presence of any continuous interictal abnormality in this cohort was 89.5%.

EEG Findings as Predictor of FCD not Identified on MRI

Ten cases of histopathologically confirmed FCD were not identified using preoperative MRI (66.7%). Of these 10 radiographically-occult FCD, 8 patients had a finding of rhythmic or continuous interictal abnormality on EEG: four patients had continuous slowing, three patients had continuous sharp waves, and one had both continuous sharp waves and rhythmic slowing.

Concordance Between MRI and Pathology

In this study, we sought to evaluate the concordance between radiographic findings of cortical dysplasia and histopathological findings in patients who underwent surgery for refractory TLE. This study found that 66.7% of FCDs found on pathology were missed on MRI. This rate of missed FCD is slightly higher than has been reported in previous studies.^{14, 17-19} This only supports the need to consider the possibility of MCD in all patients living with refractory epilepsy in order to minimize the chances of incomplete diagnosis. This report, together with others, further endorses the subtle nature of MCD on imaging and an inability to rely exclusively on standard radiologic analysis for their detection.

We also find that all 12 cases of histopathologically confirmed mMCD were missed on pre-operative MRI. This is not surprising given the high rate of missed overt FCD. However, according to the Barkovich classification of MCD, these should be included in the FCD spectrum, and therefore considered for their epileptogenic potential.¹³ Few studies have attempted to identify neuroimaging findings consistent with mMCD. In a single study investigating MRI findings in patients with mMCD, 29% of cases of mMCD were associated with lobar hypoplasia and atrophy on MRI and 29% of cases demonstrated blurring of the gray/white matter junction.¹⁷ Unfortunately, both of these features can suggest MCD in general and do not specifically suggest mMCD in particular.

Usefulness of EEG Findings as a Predictor of Malformations of Cortical Development

The second goal of this study was to identify whether the use of descriptive terms identifying rhythmic or continuous interictal abnormalities, as seen on surface EEG, are predictive of MCDs. The percentage of patients with FCD having a rhythmic or continuous interictal abnormality in this study was found to be 80%. This is much larger than the percentage of patients with FCD who were found to have unifocal epileptiform discharges on scalp EEG as reported by Kim et. al. (42.7%) or rhythmic epileptiform discharges as reported by Gambardella et. al. (44%).^{22, 23} This higher rate of interictal findings is accounted for by our inclusion of both epileptiform and non-epileptiform abnormalities, unlike the aforementioned studies which only include epileptiform abnormalities.

In this study, we demonstrate that the description of a rhythmic or continuous interictal abnormality corresponds to a sensitivity of 73.1% and a specificity of 62.5% in detecting either FCD or mMCD. These interictal findings were present in 19 (73%) of the 26 patients with MCD. In comparison, a rhythmic or continuous interictal abnormality was seen in only 3 (37.5%) of the 8 patients without evidence of MCD. These results suggest that, although a finding of rhythmic or continuous interictal slowing on EEG does not definitively determine the presence of MCD, if a patient demonstrates either of these interictal abnormalities on EEG, it should be considered that the patient has an underlying MCD. In particular, the presence of a continuous interictal EEG finding (either slowing or sharp waves) is particularly suggestive of an underlying FCD with a sensitivity of 80%.

Usefulness of EEG Findings as a Predictor of Missed FCD

We find that 80% of patients with missed FCD on pre-operative epilepsy protocol MRI had continuous or rhythmic interictal abnormalities on EEG. This finding, together with an 80% sensitivity in detecting FCD in the setting of a continuous interictal abnormality, suggests that there may be some advantage in considering these interictal EEG findings in the prediction of an underlying FCD that may escape detection on pre-operative MRI. Anticipating the possibility of a radiographically occult MCD may help inform surgical decision-making when considering laser ablation of the mesial temporal structures versus placement of intracranial electrodes to further localize seizure onset in a patient with subtle hippocampal findings on pre-operative MRI and seizures localizing to one temporal lobe on VEEG. Outcome data would suggest that temporal lobectomy yields a slightly more robust seizure freedom rate than does mesial temporal laser ablation but that patients with clear evidence of isolated hippocampal sclerosis perhaps are well-served by laser ablation given its minimally invasive nature.^27-29 This somewhat reduced seizure freedom rate following laser ablation may suggest that in the absence of hippocampal sclerosis (and perhaps the presence of continuous and/or rhythmic interictal EEG findings as described here), occult MCD may contribute to some of the decreased efficacy of such a precise surgical approach.

Limitations and the Evolution of Pathology Classification Systems

Given that terminology and classification of FCDs on histopathology has evolved over the course of this study it can be difficult to compare findings for a given subtype of FCD. Before 2011, the Palmini classification was most commonly used in histopathological classification of MCDs.¹⁶ However, classification of MCDs has shifted over time to the Blümcke classification.²⁶ This change in use of classification systems may have led to differences in histopathologic classification and may account for the more frequent classification of pathology as FCD type 2a in Patients 21-34 in Table 1 who underwent surgery after 2011 when this classification was implemented. Another potential difference in classification can be seen in the pathology findings for Patient 33 which was identified as FCD type 3a and hippocampal sclerosis (HS). Given changes in classification over time, it is possible that this pathologic finding would have been previously described as mMCD and HS or FCD type 2a and HS. The results of this study are not significantly altered by this change in terminology, however, because the results were analyzed according to overall classification of FCD and not by individual FCD type.

Another limitation of this study is that 26 of the 34 cases were found to have an MCD on pathology and of these cases, 12 had dual pathology of MCD and HS. Although an interictal abnormality was seen in only 1 of the 4 patients with isolated HS compared to 10 of the 12 patients with MCD co-occurring with HS--which may suggest that the presence of an MCD is the true underlying cause of interictal abnormalities--because the comparison groups of cases with isolated HS and other pathologies were fewer in number, this makes the specificity of the study findings in identifying MCD less certain. In order to validate the study conclusions, the findings should be applied to additional data sets with larger numbers of non-MCD pathology.