Autism Biomedical Information Network

MAG-EEG, epilepsy, and neurosurgery for autism

Go to Abstract
Go to Commentary
A recent paper published in PEDIATRICS (a publication of the American Academy of Pediatrics) purports to show that subclinical epilepsy, using an experimental technique, magnetoencephalography, is more common than generally found on conventional EEG (see Abstract below). The authors draw a tenuous analogy with Landau-Kleffner syndrome and subjected autistic children to a neurosurgical procedure with the rationale that by interrupting the epileptigenic pathway and eliminating propagation from the epileptic focus, clinical improvement is likely. They claim that post-surgical testing showed improvement in "autistic features." There are a number of problems with these claims (see Commentary below) but the net result is that this paper creates a demand for neurosurgical treatment on the basis of a dubious rationale for such treatment and poorly-documented claims of benefit.


Citation: Pediatrics 1999 Sep;104(3):405-418

Title: Magnetoencephalographic Patterns of Epileptiform Activity in Children With Regressive Autism Spectrum Disorders.

Authors: Lewine JD, Andrews R, Chez M, Patil AA, Devinsky O, Smith M, Kanner A, Davis JT, Funke M, Jones G, Chong B, Provencal S, Weisend M, Lee RR, Orrison WW, MD Jr, Department of Radiology, University of Utah, Salt Lake City, Utah.

Background: One-third of children diagnosed with autism spectrum disorders (ASDs) are reported to have had normal early development followed by an autistic regression between the ages of 2 and 3 years. This clinical profile partly parallels that seen in Landau-Kleffner syndrome (LKS), an acquired language disorder (aphasia) believed to be caused by epileptiform activity. Given the additional observation that one-third of autistic children experience one or more seizures by adolescence, epileptiform activity may play a causal role in some cases of autism.

Objective: To compare and contrast patterns of epileptiform activity in children with autistic regressions versus classic LKS to determine if there is neurobiological overlap between these conditions. It was hypothesized that many children with regressive ASDs would show epileptiform activity in a multifocal pattern that includes the same brain regions implicated in LKS.

Design: Magnetoencephalography (MEG), a noninvasive method for identifying zones of abnormal brain electrophysiology, was used to evaluate patterns of epileptiform activity during stage III sleep in 6 children with classic LKS and 50 children with regressive ASDs with onset between 20 and 36 months of age (16 with autism and 34 with pervasive developmental disorder-not otherwise specified). Whereas 5 of the 6 children with LKS had been previously diagnosed with complex-partial seizures, a clinical seizure disorder had been diagnosed for only 15 of the 50 ASD children. However, all the children in this study had been reported to occasionally demonstrate unusual behaviors (eg, rapid blinking, holding of the hands to the ears, unprovoked crying episodes, and/or brief staring spells) which, if exhibited by a normal child, might be interpreted as indicative of a subclinical epileptiform condition. MEG data were compared with simultaneously recorded electroencephalography (EEG) data, and with data from previous 1-hour and/or 24-hour clinical EEG, when available. Multiple-dipole, spatiotemporal modeling was used to identify sites of origin and propagation for epileptiform transients.

Results: The MEG of all children with LKS showed primary or secondary epileptiform involvement of the left intra/perisylvian region, with all but 1 child showing additional involvement of the right sylvian region. In all cases of LKS, independent epileptiform activity beyond the sylvian region was absent, although propagation of activity to frontal or parietal regions was seen occasionally. MEG identified epileptiform activity in 41 of the 50 (82%) children with ASDs. In contrast, simultaneous EEG revealed epileptiform activity in only 68%. When epileptiform activity was present in the ASDs, the same intra/perisylvian regions seen to be epileptiform in LKS were active in 85% of the cases. Whereas primary activity outside of the sylvian regions was not seen for any of the children with LKS, 75% of the ASD children with epileptiform activity demonstrated additional nonsylvian zones of independent epileptiform activity. Despite the multifocal nature of the epileptiform activity in the ASDs, neurosurgical intervention aimed at control has lead to a reduction of autistic features and improvement in language skills in 12 of 18 cases.

Conclusions: This study demonstrates that there is a subset of children with ASDs who demonstrate clinically relevant epileptiform activity during slow-wave sleep, and that this activity may be present even in the absence of a clinical seizure disorder. MEG showed significantly greater sensitivity to this epileptiform activity than simultaneous EEG, 1-hour clinical EEG, and 24-hour clinical EEG. The multifocal epileptiform pattern identified by MEG in the ASDs typically includes the same perisylvian brain regions identified as abnormal in LKS. When epileptiform activity is present in the ASDs, therapeutic strategies (antiepileptic drugs, steroids, and even neurosurgery) aimed at its control can lead to a significant improvement in language and autistic features.

Key words: autism, pervasive developmental disorder-not otherwise specified, epilepsy, magnetoencephalography, Landau-Kleffner syndrome.

PMID: 10469763


For the purpose of this commentary, the term "autistic spectrum disorders" or ASDs, includes "autistic disorder" and "pervasive developmental disorder not otherwise specified" (PDDNOS) as specified in the DSM-IV manual.

One of the most confusing and controversial issues in the field of autism is the whole matter of EEGs, regression in autism, and treatment. The confusion arises from conflicting data, unclear definitions (for example, "LKS variant"), and claims for treatment efficacy based on outcome measures of dubious validity. This paper (see Abstract above) raises serious questions of ethics in choosing a treatment, concerns regarding peer review of manuscripts, and individual responsibility, if not legal liability, in the instance of multiple authorship. In particular the following major problems stand out:

  • All subjects are said to have "regressive ASD before the age of 6 years." They all had "normal early development up to 18 to 30 monthsof age" when, by parental report, regression occurred. Of course "normal" is in the eyes of the beholder. It is well-known that reliable diagnosis of autism is especially difficult under the age of 2 years. At this age, some children have not yet developed much expressive language and it is problematic to claim that previously-attained language milestones have regressed. It is possible that many of these young children had subtle features of autism during infancy but that the severity of autistic behaviors had not risen to a level that exceeded a threshold of concern by parents or a threshold of detection by pediatricians. Although Lewine et al conflate autistic disorder and PDDNOS, it is not at all certain that the latter have the same likelihood of regression as that stated for autistic disorder (about 30%).

  • The authors conflate autistic regression and Landau-Kleffner syndrome (LKS). Although there is a broad consensus that LKS is a distinct entity primarily affecting language in children older than 3 to 4 years of age, these authors draw a tenuous analogy between autistic regression and LKS. They presume that a similar electrophysiologic disturbance is the core feature of each. This blurred distinction ignores the fact that there is not even agreement among neurologists as to the significance of EEG changes in LKS, let alone autism. Even Dr. Landau, some 40 years after the initial description of LKS, is uncertain as to the pathophysiologic significance of the EEG changes (Landau WM, Landau-Kleffner Syndrome. An Eponymic Badge of Ignorance. Arch Neurol 1992;49:353). Yet, Lewine et al have made a facile leap from limited data to the conclusion that epileptiform activity is both primary and causative of regression in children with ASDs. Furthermore, they put an epilepsy "spin" on certain behaviors, common in autism, which they describe as "paroxysmal behaviors" (rapid blinking of eyes, holding of hands over the ears, unprovoked crying, or staring spells) and which they equate with seizures. It is well-known that many children with ASDs have unusual "sensory overload" behaviors, especially in relation to lights or ambient noise levels. So-called "unprovoked crying" episodes are often tantrums or distress at a perturbation of "sameness."

  • There are also serious concerns regarding diagnosis and ascertainment. There is imprecision of diagnostic terminology. Some of the imprecision may lie in the fact that entry of patients overlapped the period of DSM-IIIR and the DSM-IV criteria. The DSM-IIIR criteria are known to have been overly sensitive and yielded false positives, especially in the higher-functioning category, PDDNOS. Moreover, they conflate "autism" (probably meaning "autistic disorder") and PDDNOS, without regard to the delineation provided by the DSM-IV criteria.

  • The authors claim that surgical intervention, patterned after that sometimes used for LKS, multiple subpial transections in the cerebral cortical area presumed to be an epileptigenic focus, resulted in "reduction of autistic features" (whatever the term, "autistic features," means). Regrettably, there is exeedingly poor documentation of claims of improvement, using outcome measures that are amenable to statistical tests of significance. For example, higher scores on a follow-up CARS (Childhood Autism Rating Scale) are cited but they fail to state the denominator, i.e., the change in score in relation to the overall potential score (the total possible score is 60). The CARS is an instrument for initial diagnostic assessment and has not been validated for sequential use for the purpose of measuring a treatment effect. It is subject to observer bias when the observer knows that the followup CARS is being done to assess a treatment. It also does not account for a possible placebo effect which is higly likely when children are subjected to a number of observations over a period of time, during which behavioral changes consequent to being observed on multiple occasions can occur without specific linkage to a treatment maneuver. Similar caveats apply to the sequential use of the Peabody Picture Vocabulary Test. Better rating instruments include the Aberrant Behavior Checklist and Vineland Adaptive Behavior scales.

  • Finally, a serious ethical question arises when children are subjected to an experimental neurosurgical procedure without mention, in the paper, of the protocol for obtaining informed consent or Institutional Review Board approval.

The authors describe a novel, and still research-based, method of localizing the site of epileptigenic activity in the brain, which they correlate with MR scans and EEGs, and refer to as magnetoencephalography (MEG). This approach uses magnetic source imaging, measured at the scalp using a whole-head neuromagnetometer, which detects the magnetic flux attributed to the electrophysiologic activity of large neuronal clusters. Computational methods using algorithms provide a topological map of the intracranial source of the electrophysiologic currents responsible for the magnetic flux. This paper should have been limited to a description of MEG and its possible, but not yet proven, utility in more precise mapping of epileptiform foci in autistic disorder and PDDNOS (LKS remains a separate question).

The authors also fail to provide an in-depth discussion of a major conceptual issue: What is the significance of "subclinical epilepsy" in children with ASDs? Is is possible that epileptiform changes on EEG, in the absence of overt seizures, constitute a kind of "neurocognitive epilepsy" (or is such activity merely an epiphenomenon)? This would represent a paradigm shift since the common forms of epilepsy are generally defined in terms of involuntary movement manifestations. Is there is such a thing as altered electrical activity in brain regions not concerned with movement that manifest only as behavioral changes (or, in the instance of LKS, as auditory agnosia)? This question deserved more in-depth discussion.

Return to Overview above

Return to list of commentaries and other autism links

Commentary by Ronald J. Kallen, MD, ©1999, This page last updated on 01/05/00