Abstract
Reyes A, Hermann BP, Busch RM, Drane DL, Barr WB, Hamberger MJ, Roesch SC, McDonald CR. Brain Commun. 2022;4(6):fcac289. doi:10.1093/braincomms/fcac289
In efforts to understand the cognitive heterogeneity within and across epilepsy syndromes, cognitive phenotyping has been proposed as a new taxonomy aimed at developing a harmonized approach to cognitive classification in epilepsy. Data- and clinically driven approaches have been previously used with variability in the phenotypes derived across studies. In our study, we utilize latent profile analysis to test several models of phenotypes in a large multicentre sample of patients with temporal lobe epilepsy and evaluate their demographic and clinical profiles. For the first time, we examine the added value of replacing missing data and examine factors that may be contributing to missingness. A sample of 1178 participants met the inclusion criteria for the study, which included a diagnosis of temporal lobe epilepsy and the availability of comprehensive neuropsychological data. Models with two to five classes were examined using latent profile analysis and the optimal model was selected based on fit indices, posterior probabilities and proportion of sample sizes. The models were also examined with imputed data to investigate the impact of missing data on model selection. Based on the fit indices, posterior probability and distinctiveness of the latent classes, a three-class solution was the optimal solution. This three-class solution comprised a group of patients with multidomain impairments, a group with impairments predominantly in language and a group with no impairments. Overall, the multidomain group demonstrated a worse clinical profile and comprised a greater proportion of patients with mesial temporal sclerosis, a longer disease duration and a higher number of anti-seizure medications. The four-class and five-class solutions demonstrated the lowest probabilities of a group membership. Analyses with imputed data demonstrated that the four-class solution was the optimal solution; however, there was a weak agreement between the missing and imputed data sets for the four-Class solutions (κ = 0.288, P < 0.001). This study represents the first to use latent profile analysis to test and compare multiple models of cognitive phenotypes in temporal lobe epilepsy and to determine the impact of missing data on model fit. We found that the three-phenotype model was the most meaningful based on several fit indices and produced phenotypes with unique demographic and clinical profiles. Our findings demonstrate that latent profile analysis is a rigorous method to identify phenotypes in large, heterogeneous epilepsy samples. Furthermore, this study highlights the importance of examining the impact of missing data in phenotyping methods. Our latent profile analysis-derived phenotypes can inform future studies aimed at identifying cognitive phenotypes in other neurological disorders.
Commentary
Cognitive dysfunction is a common comorbidity in the epilepsies and is best evaluated with a neuropsychological evaluation. 1 The traditional model of evaluating cognition has relied on the lesion model which links cognitive deficits in specific domains (memory, executive functions, etc) to specific epilepsy syndromes. 2 Examples include memory impairment in temporal lobe epilepsy (TLE), and executive dysfunction in frontal lobe epilepsy. Viewed through this lens, cognitive deficits in focal epilepsy reflected impairment in function in a focal area of the brain where the epilepsy arises from. A similar theme also emerged with the generalized epilepsy syndromes with childhood absence epilepsy linked to impaired attention and juvenile myoclonic epilepsy linked to impaired working memory. As helpful as this model has been in advancing our understanding of cognition in epilepsy, it did not reflect the heterogeneity of patients having these syndromes. It became apparent that despite having TLE some patients had intact memory and cognition, while other patients with TLE frequently had executive dysfunction/multidomain impairment. 3 As our understanding of TLE evolved from a unifocal disease to a network disorder, our analytical techniques also had to evolve.
In the study by Reyes et al, 4 7 epilepsy centers pooled their neuropsychological data to try to tackle this question. They use an analytical technique termed latent profile analysis (LPA) 5 which belongs to the family of person-centered statistical approaches and provides advantages over the techniques used to date; cluster analysis and using prespecified cutoffs for impairment (such as having scores 1 or 1.5 standard deviations below the mean). The goal of LPA is to identify latent subpopulations within a population based on a set of variables and is less prone to misclassification. It allows you to include subjects with missing data, can handle continuous and categorical data, and provides probability values for each subject belonging to the identified subpopulation. The study evaluated subjects with 8 cognitive tests covering 5 cognitive domains: Memory (logical memory immediate and delayed recall), language (Boston naming test, category fluency, letter fluency), mental flexibility (trails making B), processing speed (trails making A), motor dexterity (grooved pegboard). To be included in the analysis, subjects had to have data available on at least 6 out of the 8 tests. The dataset did not include any list-learning tasks or tasks assessing visual memory. Neuroimaging features were limited to the presence or absence of mesial temporal sclerosis.
Ultimately, 1178 subjects were included in the analysis with the following demographics: average age of 37.8 years, 57% female, and 79% non-Hispanic white. Using LPA, the 3-class model was found to be the most parsimonious and clinically meaningful. The identified phenotypes consisted of 1, No impairment; 2, Multidomain impairment; and 3, Language impairment. This 3-class model was the least affected by the missing data although there seemed to be differences between the subjects with missing data versus those without; they tended to be older and have a lower level of education. When assessing the clinical differences between these phenotypes, individuals with the multidomain impairment had the longest disease duration, lowest years of education, highest number of anti-seizure medications (ASMs), and higher rates of mesial temporal sclerosis. Not surprisingly, those with the intact, no impairment, phenotype had the opposite profile. The language phenotype consisted of individuals with impaired performance in the language tasks, and an intermediate duration of epilepsy, educational level, and burden of ASMs.
This study highlights the strengths of multicenter collaborations. As analytical methods advance, larger numbers are needed for us to gain a better understanding of epilepsy and its comorbidities, and we can only achieve this by these types of collaborations. It is also confirmation that there are different cognitive phenotypes within TLE, and these phenotypes are correlated with several clinical variables. These findings will also facilitate individualized treatment as we appreciate what factors create a more resilient brain and allow us to start looking at structural-functional correlations within individuals with these phenotypes. A study evaluating 4 cognitive phenotypes in 70 TLE showed how each phenotype had different white matter network abnormalities. 6
In terms of study limitations, the largest limitation is that the dataset did not have a list learning task because they are more sensitive in detecting verbal memory impairment, so there is likely an underestimation of the degree of memory impairment in this cohort. Single-center studies have identified a memory only or memory and language phenotype. 7 The lack of visual memory tasks is another limitation, this has been a longstanding issue in neuropsychological assessments due to the lack of consensus around the ideal test. The study did not find any effects of laterality on the identified phenotypes, although it is unclear whether this is a true finding or whether they did not have the necessary data points to answer this question. There were also very limited data regarding the TLE etiology, with the only imaging information included being mesial temporal sclerosis, so other common pathologies leading to TLE such as low-grade tumors, malformations of cortical development, and vascular lesions were not assessed. There was also a lack of granularity with regard to the type of ASMs, as the authors mention it could be that medications such as topiramate and zonisamide are a prominent contributor to the language-only phenotype, given that they selectively seem to impact language networks. 8
It is always quite a statement when the sample size of a study is included in the title; 1178! This is certainly the largest study to date examining cognition in TLE. As we start achieving sample sizes of 1000s, we can add more clinical variables and ask interesting questions such as: Does bilingualism or multilingualism play a role? Does exercise provide resilience? What about the effects of marijuana and alcohol? Are there EEG features contributing to these phenotypes such as interictal discharges? Are there optimal rehab approaches that would benefit patients with different phenotypes?
Some questions in the field will only be answered by collaboration and data pooling. This study is an example of the success of this approach, albeit with some limitations. There are patterns we can only appreciate when we look together as a group.
Footnotes
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.