Etiologic classification of infantile spasms in 140 cases: role of positron emission tomography

J Child Neurol. 1996 Jan;11(1):44-8. doi: 10.1177/088307389601100111.


The classification of infantile spasms into symptomatic, cryptogenic, and idiopathic subgroups depends on clinical examination and available diagnostic technology. Positron emission tomography (PET) of glucose utilization is a powerful tool in detecting brain malformations (particularly cortical dysplasia) in infants with spasms. We analyzed etiologic data from 140 such infants, 78 girls and 62 boys, ages 2 months to 4 years 10 months (mean, 17 months). All had been evaluated extensively in one of two major medical centers. It should be emphasized that our referral population is biased toward infants with intractable spasms who fail to show a structural lesion. Seven patients had neurocutaneous syndromes, two had chromosomal abnormalities, two had inborn errors of metabolism, and one each had craniosynostosis or Menkes syndrome. Computed tomography and/or magnetic resonance imaging detected lesions in another 29 infants (20.7%) who did not have a specific disease or syndrome. Without the benefit of PET, the total number of symptomatic cases was 42 (30.0%). One infant, classified as idiopathic, had normal development and PET. In 97 cryptogenic cases, PET uncovered unifocal abnormalities in 30 and multifocal abnormalities in 62. Diffuse PET abnormalities, which did not provide specific etiologic information, were seen in three infants. Another two infants had normal PET scans. Thus, with the benefit of PET, the number of symptomatic cases rose dramatically from 42 (30.0%) to 134 (95.7%). The majority of unifocal and multifocal abnormalities on PET are believed to represent dysplastic lesions.

MeSH terms

  • Brain / diagnostic imaging*
  • Brain / surgery
  • Child, Preschool
  • Female
  • Humans
  • Infant
  • Male
  • Spasms, Infantile / diagnostic imaging*
  • Spasms, Infantile / etiology*
  • Spasms, Infantile / surgery
  • Tomography, Emission-Computed*