Abstract
McDonald CR, Busch RM, Reyes A, Arrotta K, Barr W, Block C, Hessen E, Loring DW, Drane DL, Hamberger MJ, Wilson SJ, Baxendale S, Hermann BP. Neuropsychology. 2022. doi:10.1037/neu0000792
To describe the development and application of a consensus-based, empirically driven approach to cognitive diagnostics in epilepsy research-The International Classification of Cognitive Disorders in Epilepsy (IC-CoDE) and to assess the ability of the IC-CoDE to produce definable and stable cognitive phenotypes in a large, multi-center temporal lobe epilepsy (TLE) patient sample. Neuropsychological data were available for a diverse cohort of 2,485 patients with TLE across seven epilepsy centers. Patterns of impairment were determined based on commonly used tests within five cognitive domains (language, memory, executive functioning, attention/processing speed, and visuospatial ability) using two impairment thresholds (≤1.0 and ≤1.5 standard deviations below the normative mean). Cognitive phenotypes were derived across samples using the IC-CoDE and compared to distributions of phenotypes reported in existing studies. Impairment rates were highest on tests of language, followed by memory, executive functioning, attention/processing speed, and visuospatial ability. Application of the IC-CoDE using varying operational definitions of impairment (≤1.0 and ≤1.5 SD) produced cognitive phenotypes with the following distribution: cognitively intact (30%-50%), single-domain (26%-29%), bi-domain (14%-19%), and generalized (10%-22%) impairment. Application of the ≤1.5 cutoff produced a distribution of phenotypes that was consistent across cohorts and approximated the distribution produced using data-driven approaches in prior studies. The IC-CoDE is the first iteration of a classification system for harmonizing cognitive diagnostics in epilepsy research that can be applied across neuropsychological tests and TLE cohorts. This proof-of-principle study in TLE offers a promising path for enhancing research collaborations globally and accelerating scientific discoveries in epilepsy.Objective:
Method:
Results:
Conclusions:
Commentary
The young woman with temporal lobe epilepsy (TLE) in front of me says “my seizures are not too bad, but I just can’t remember anything, I’ve had to give up my job as a teacher.”
Impaired cognition is a major comorbidity in epilepsy, that can affect quality of life as much as the seizure disorder that drives it. Improving how we define and communicate cognitive phenotypes in epilepsy is critical to how we can help this patient.
Cognitive phenotyping in epilepsy has lagged behind efforts to standardize cognitive deficits in comparison to progress within other areas such as psychiatry (Diagnostic and statistical manual of mental disorders, 5th edition, DSM-V) and dementia (definitions of mild cognitive impairment, MCI). However, in DSM-V, cognitive impairment is only classified as mild or major neurocognitive disorder and measures of MCI are too broad for use in epilepsy. We need an alternative.
The “classic” or “test-driven” model of cognitive phenotyping attempts to define a specific cognitive pattern, such as memory or language dysfunction, in clinicopathological syndromes such as mesial temporal sclerosis (MTS). In practice we see variable or widespread patterns across different epilepsy syndromes, which probably correlates with a widespread disrupted network, supported by functional and structural neuroimaging evidence of altered connectivity or structure both within and outside the temporal lobe in TLE. 1 There are also complex contributions from mood, medications, sleep disruption, and other factors.
McDonald et al show the first fruits of a collaboration between the ILAE Neuropsychology task force and the International Neuropsychological Society—The International Classification of Cognitive Disorders in Epilepsy (IC-CoDE)—to standardize cognitive phenotypes in epilepsy (in this study, TLE) to facilitate research across centers, countries, and ethnic and language boundaries. 2
They use an approach that first classifies cognitive phenotypes within a broad epilepsy syndrome “TLE,” and then test its reproducibility across 7 US-based centers, using IC-CoDE defined cognitive domains and phenotypes, in 2485 patients with TLE who had previously completed neuropsychological testing.
The inclusion criteria were age over 16 years, history of TLE (mesial or lateral), no prior neurosurgery, English speaking, and neuropsychological test battery available. Pathologies were diverse including MTS, cortical dysplasia, tumors, and vascular lesions. Baseline demographics showed that most patients were white, and had an average of 14 years of education, limiting generalizability to other ethnic or deprived populations. The welcome inclusion of nonsurgical TLE allows improved understanding of TLE as a whole, where many previous studies have included only drug-resistant surgical TLE.
Based on prior literature and consensus, a 5 domain cognitive model was proposed—language, memory, executive functioning, visuospatial abilities, and attention/processing speed. Representative abilities (i.e., naming and fluency for language) and a representative test (i.e., Boston naming test [BNT] and category fluency) were chosen for each domain across all centers. The study analyzed impairment rates (percentage abnormal) for each neuropsychological test within the 5 cognitive domains. A domain was deemed affected if at least 2 tests were below abnormal thresholds.
Operational criteria for threshold of abnormality were ≤1 or ≤1.5 standard deviation (SD) from normative means, similar to DSM-V. 3 Which of these two SD cutoffs is more accurate for epilepsy is not known. This approach (known as the Jak/Bondi approach) has also been successful in defining MCI (being good at limiting false positives).
Cognitive deficit rates for each domain below the cutoff of ≤1 SD and ≤1.5 SD were seen, in order of frequency, in language (e.g., range 53%-67% on BNT), memory (e.g., verbal memory/list learning 27%-51%), executive function (e.g., set-shifting 24%-40%), attention (e.g., working memory 19%-29%), and visuospatial (9%-27%).
The cohorts were then grouped into the IC-CoDE cognitive phenotypes, depending on how many domains were affected (1) intact (no domains affected), (2) one or two domain involvement, or (3) generalized impairment (≥3 domains affected). The final outcome was classification into a cognitive phenotype.
Using a ≤1 SD cutoff, 30% had an intact phenotype, 29% single domain impairment (language 48%, memory 27%), 19% bidomain impairment, and 22% generalized impairment. Only 2% showed domain-specific impairment in visuospatial processing.
Using a ≤1.5 SD cutoff, the intact rate was higher at 50% and rates for other domains lower—26% single (language 49%, memory 32%), 14% bidomain, and 10% generalized. The authors concluded that using a 4 domain model (excluding visuospatial) with a ≤1.5 SD cutoff seemed to produce results “closer in proportion” to cluster analysis driven rates in previously published studies. However, this seems optimistic as, for example, the ≤1 SD generalized rate was 22%, and in prior studies was 21.1%, so the reality is somewhere between the 2 cutoff thresholds.
Results were consistent however, across the 4 or 5 domain cohorts, with no statistical difference across sites in both models, despite variability in test batteries and normative datasets, supporting that this model can be used across (at least US based) sites for research.
The study highlights the cognitive variability and complexity in TLE—some are not cognitively affected, while others have generalized deficits—further phenotyping of these groups will help determine what factors contribute to this, by examining clinical, demographic, imaging, and genetic characteristics. In prior studies, a generalized phenotype was associated with a higher lifetime number of seizures (particularly tonic–clonic), early age of use of anti-seizure medications (ASM), or imaging evidence of lower intracranial volume, lower cerebellar volume, reduced connectivity or MTS. 3 The finding that 50% of patients had intact cognition was surprising and suggests that a cognitive deficit is not inevitable.
The study somewhat oversimplifies cognition in epilepsy—cognitive impairment in epilepsy is difficult to compartmentalize. Cognitive deficits are a complex interaction of the underlying pathology, developmental insults, socioeconomic group, seizures, mood, education, and ASM side effects. Patients with intellectual and developmental disabilities may not fit well into the IC-CoDE model. However, the authors emphasize that their data should not be used for clinical application or for baseline rates of cognitive impairment in TLE, which would require validation against clinical imaging and neurophysiological, genetic, and outcome measures. While these cognitive phenotypes are ‘artificial', they do help to harmonize research efforts across centers where different tests, normative values or cultures may exist, and facilitate big data approaches to clinico-genetic analysis of the epilepsies.
We look forward to further applications of the IC-CoDE, including addressing the pediatric population, extratemporal epilepsies, different populations, planned use of modifiers (severity, etiology, and mood), and use of additional cognitive domains such as social cognition and accelerated forgetting. Deeper phenotyping, facilitated by these efforts, should enhance our ability to help our young teacher with TLE and others like her.
