Presented as ASA, October 26, 2015


Caleb H. Ing, M.D., Melanie Wall, Ph.D., Charles DiMaggio, Ph.D., Andrew Whitehouse, Ph.D., Mary Hegarty, M.B.,B.S., Britta von Ungern-Sternberg, M.D., Guohua Li, M.B., Lena Sun, M.D.
Columbia University Medical Center, New York, New York, United States


The safety of anesthetic agents in children has been questioned with some studies reporting an association between early anesthetic exposure and long-term deficit.1 It is however unclear if this association is found in all children or if only specific subgroups of children may be vulnerable. Latent class analysis (LCA) is a statistical method that can be used to identify discrete subgroups of clinical phenotypes.2 We used LCA of specific neurodevelopmental outcomes to characterize subgroups of children potentially vulnerable to anesthesia.


Data were obtained from the Western Australian Pregnancy Cohort Study (Raine) (2868 children born 1989-1992) and results of neuropsychological (NP) assessments at age 10 years, including language (Peabody Picture Vocabulary [PPV], Clinical Evaluation of Language Fundamentals: Receptive [CELF-R] and Expressive [CELF-E]), cognition (Colored Progressive Matrices [CPM], Symbol Digit Modality Test: Oral [SDMT-O] and Written [SDMT-W]), motor function (McCarron Assessment of Neuromuscular Development [MAND]) and behavior (Child Behavior Checklist: Internalizing [CBCL-INT], Externalizing [CBCL-EXT], and Total behavior [CBCL-T]) were assessed. Children were evaluated for anesthesia exposure before age 3, and only those with all available NP test results at age 10 were included in the analysis. Based on LCA of the NP tests, the cohort was divided into mutually exclusive subclasses of cognitive deficit. Using a multivariable polytomous logistic regression model, we determined the strength of association between anesthesia exposure and each latent class, and also adjusted for demographics, perinatal health status, and comorbid disease.


A total of 1444 children were included in the analysis. LCA indicated that a four-class model provided the best fit. Thus, four clinically meaningful groups were identified: 1) Normal: little or no deficits, (n=1135, 78.6%), 2) Behavioral deficits: high probability for behavioral deficits, low probability for other deficits. (n= 151, 10.5%), 3) Language and Cognitive deficits: high probability for language, cognitive, and motor deficits and low probability for behavioral deficits (n=96, 6.6% of cohort) and 4) Severe deficits: high probability for deficits in all NP domains (n=62, 4.3%).(Figure 1) Children with Language and Cognitive deficits had significantly higher odds of prior anesthesia exposure vs Normal children [adjusted odds ratio (aOR), 2.35 (95% CI, 1.36 – 4.07)], while those with Behavioral or Severe deficits were comparable to Normal children [aOR, 0.90 (95% CI, 0.54 – 1.52), and aOR, 0.91 (95% CI, 0.41 – 2.04) respectively] with respect to anesthesia exposure. There were no differences in patient-specific characteristics among the three classes of children with deficits.


Applying LCA to results of NP tests at age 10 years in the Raine cohort, we identified four subgroups of children. Children with isolated deficits in language and cognition were associated with prior exposure to anesthesia, but children with severe deficits, who had language and cognitive in addition to behavioral deficits, were not associated with exposure. Our results suggest that specific subgroups of children may be at risk for developing neuropsychological deficits after anesthesia exposure while other subgroups may not be at risk. Additional studies are needed to better identify demographic and other patient characteristics as risk factors for this vulnerability.