Impact of prenatal marijuana exposure on adolescent brain structural and functional connectivity and behavioural outcomes

Abstract

There has been an increase in the number of women using marijuana whilst pregnant. Previous studies have shown that children with prenatal marijuana exposure have developmental deficits in memory and decreased attentiveness. In this study, we assess whether prenatal marijuana exposure is associated with alterations in brain regional morphometry and functional and structural connectivity in adolescents. We downloaded behavioural scores and subject image files from the Adolescent Brain Cognitive Development^SM Study. A total of 178 anatomical and diffusion magnetic resonance imaging files (88 prenatal marijuana exposure and 90 age- and gender-matched controls) and 152 resting-state functional magnetic resonance imaging files (76 prenatal marijuana exposure and 76 controls) were obtained. Behavioural metrics based on the parent-reported child behavioural checklist were also obtained for each subject. The associations of prenatal marijuana exposure with 17 subscales of the child behavioural checklist were calculated. We assessed differences in brain morphometry based on voxel-based and surface-based morphometry in adolescents with prenatal marijuana exposure versus controls. We also evaluated group differences in structural and functional connectivity in adolescents for region-to-region connectivity and graph theoretical metrics. Interactions of prenatal marijuana exposure and graph networks were assessed for impact on behavioural scores. Multiple comparison correction was performed as appropriate. Adolescents with prenatal marijuana exposure had greater abnormal or borderline child behavioural checklist scores in 9 out of 17 subscales. There were no significant differences in voxel- or surface-based morphometry, structural connectivity or functional connectivity between prenatal marijuana exposure and controls. However, there were significant differences in prenatal marijuana exposure-graph network interactions with respect to behavioural scores. There were three structural prenatal marijuana exposure-graph network interactions and seven functional prenatal marijuana exposure-graph network interactions that were significantly associated with behavioural scores. Whilst this study was not able to confirm anatomical or functional differences between prenatal marijuana exposure and unexposed pre-adolescent children, there were prenatal marijuana exposure-brain structural and functional graph network interactions that were significantly associated with behavioural scores. This suggests that altered brain networks may underlie behavioural outcomes in adolescents with prenatal marijuana exposure. More work needs to be conducted to better understand the prognostic value of brain structural and functional network measures in prenatal marijuana exposure.

Keywords: ABCD; functional connectivity; graph networks; prenatal marijuana exposure; structural connectivity.

Graphical Abstract

Figure 1

PME interactions for structural graph networks. Plots of PME–graph network interactions for structural connectivity using a linear model (n = 178). (A) Chart represents a scatter plot with raw data overlayed with an interaction plot based on adjusted data for externalizing problems scores versus local efficiency in the right lateral OFC grouped by exposure. There were significant differences in interactions between PME and control groups (t = −3.69). (B) Chart represents a scatter plot with raw data overlayed with an interaction plot based on adjusted data for externalizing problems scores versus clustering coefficient in the right lateral OFC grouped by exposure. There were significant differences in interactions between PME and control groups (t = −3.68). (C) Chart represents a scatter plot with raw data overlayed with an interaction plot based on adjusted data for total problems scores versus betweenness centrality in the right amygdala grouped by exposure. There were significant differences in interactions between PME and control groups (t = 3.89). A false discovery rate less than 0.05 was considered significant.

Figure 2

PME interactions for rule-breaking scores. Plots of PME–graph network interactions for functional connectivity using a robust linear model (n = 152). (A) Chart represents a scatter plot with raw data overlayed with an interaction plot based on adjusted data for rule-breaking scores versus betweenness centrality in the right fusiform gyrus grouped by exposure. There were significant differences in interactions between PME and control groups (t = 5.02). (B) Chart represents a scatter plot with raw data overlayed with an interaction plot based on adjusted data for rule-breaking scores versus betweenness centrality in the left cuneus grouped by exposure. There were significant differences in interactions between PME and control groups (t = 5.89). A false discovery rate less than 0.05 was considered significant.

Figure 3

Sluggish cognition and betweenness centrality. Plot of PME–graph network interactions for functional connectivity using a robust linear model (n = 152). Chart represents a scatter plot with raw data overlayed with an interaction plot based on adjusted data for sluggish cognition scores versus betweenness centrality in the left calcarine sulcus grouped by exposure. There were significant differences in interactions between PME and control groups (t = 6.68). A false discovery rate less than 0.05 was considered significant.

Figure 4

Sluggish cognition and clustering coefficient. Plots of PME–graph network interactions for functional connectivity using a robust linear model (n = 152). (A) Chart represents a scatter plot with raw data overlayed with an interaction plot based on adjusted data for sluggish cognition scores versus clustering coefficient in the left anterior middle temporal gyrus grouped by exposure. There were significant differences in interactions between PME and control groups (t = 6.29). (B) Chart represents a scatter plot with raw data overlayed with an interaction plot based on adjusted data for sluggish cognition scores versus clustering coefficient in the left anterior inferior temporal gyrus grouped by exposure. There were significant differences in interactions between PME and control groups (t = 3.69). A false discovery rate less than 0.05 was considered significant.

Figure 5

PME interactions for total problems scores. Plot of PME–graph network interactions for functional connectivity using a robust linear model (n = 152). Chart represents a scatter plot with raw data overlayed with an interaction plot based on adjusted data for total problems scores versus clustering coefficient in the posterior cingulate cortex grouped by exposure. There were significant differences in interactions between PME and control groups (t = 4.71). A false discovery rate less than 0.05 was considered significant.