This study in young females is one of the largest longitudinal neuroimaging studies of pathological anxiety, focused on understanding alterations in neural pathways relevant to the development of anxiety during childhood. The major finding from this study was derived from the longitudinal data, revealing that within participants, more severe anxiety symptoms were associated with lower whole-brain FA (Table 3 and Fig. 2). This relation manifested in treatment-naïve girls, including girls with subthreshold AD symptoms as well as girls who met criteria for ADs, independent of age and pubertal status. We also performed cross-sectional analyses with data collected at study entry, comparing WM integrity among controls, subthreshold-AD participants, and AD participants. In contrast to the longitudinal finding, this cross-sectional analysis revealed no significant relations between FA and anxiety. Taken together, these findings highlight a dynamic relation between whole-brain WM integrity and anxiety, as well as the importance of a longitudinal within-participant approach for studying developmental psychopathology.
The presence of a significant longitudinal association between anxiety and WM in the context of no significant cross-sectional association is notable (Tables 2 and 3). This suggests that within an individual, whole-brain FA and anxiety fluctuate together over time, regardless of individual differences in FA magnitude. We are unaware of other studies that have concomitantly examined the relation between anxiety symptoms and WM parameters longitudinally in children with pathological anxiety. One longitudinal DTI study in a normative sample of youth (ages 6–18) found that children with higher anxiety/depression symptoms at study intake had slower rates of WM development in multiple WM tracts [27]. Another study in a large sample of youth reported an association between initially assessed internalizing and externalizing symptoms with reduced growth-related increases in global WM [54]. Other work examining WM alterations in relation to pediatric anxiety has been cross-sectional. Four cross-sectional DTI studies have been performed in youth (ages 6–18), examining typically developing youth with trait anxiety as well as children with ADs [25, 26, 28, 29]. Consistent with DTI studies of anxiety in adults [13, 16, 17, 19], these studies have generally reported anxiety-related WM reductions in FA in various regions, including the UF, CING, CC, and IFO. We previously reported a reduction in UF FA in boys with ADs but not girls, also using a cross-sectional approach [25]. The lack of a relation between anxiety and UF FA in girls in our previous study is consistent with the cross-sectional results reported here in preadolescent girls with pathological anxiety. These null effects should be interpreted cautiously. We note that a Bayesian analysis performed on the data from the girls in the previous study did not support evidence for the absence of an effect (see Supplemental Results), whereas a Bayesian analysis performed on the current dataset was supportive of evidence of absence for the lack of an association between UF FA and anxiety (Table 2).
The correlation between global WM microstructure and anxiety symptoms suggests the presence of a diffuse whole-brain WM effect. The consequences of this global effect could be relevant to the inter- and intra-connectivity among brain networks relevant to emotional information processing and integration, aversive stimulus detection, and the interpretation of social behavior, which is consistent with reports of whole-brain WM microstructural alterations related general psychopathology factors and cognitive abilities in youth [24, 55]. While our results suggest a dynamic relationship between WM microstructure and childhood anxiety, the mechanism underlying this association is unclear. Furthermore, it is possible that the association between WM integrity and anxiety symptoms is not causally linked, as other factors could concomitantly impact both of these measures. However, studies in both NHPs and humans suggest that stress can affect WM microstructure [56, 57]. As such, it is plausible that in our sample of girls, the experience of chronically heightened anxiety could result in altered WM microstructure. Specific mechanisms that have been implicated from preclinical studies link adversity to WM microstructure via effects on oligodendrogenesis and myelination [58, 59]. It is also conceivable that WM microstructure plays a role in directly mediating levels of anxiety. Studies manipulating oligodendrocyte function in preclinical models of anxiety could be informative in this regard. Our results provide an impetus to examine the potential utility of WM as a treatment target for early-life anxiety. In contrast to a specific tract, the data support targeting global WM microstructure. Recent work demonstrates that WM microstructure can be enhanced by lifestyle interventions such as exercise, nutrition, and sleep [60–64]. These interventions are well-suited for children and could be easily used in conjunction with pharmacological and/or psychotherapeutic treatment strategies.
Although this study included a relatively large sample, most participants were White (Supplementary Table 4). Expanding the diversity of the participants in these studies to include more BIPOC individuals will be important to enhance generalizability of the findings. Our sample did not include boys, and our longitudinal analysis did not include control participants, limiting our conclusions regarding the association between WM microstructural integrity and anxiety symptoms to only girls with pathological anxiety. While there was some attrition in the follow-up years of the study, the statistical approaches used – linear mixed-effects models in particular – are designed to handle missing data in longitudinal datasets [53]. While we attribute the FA-anxiety relation to whole-brain reductions in FA, it is possible that there are associations between anxiety symptoms and FA in specific WM tracts that were not examined in this study.
In summary, we present one of the largest longitudinal neuroimaging studies of pediatric anxiety, demonstrating that, on an individual level in girls with pathological anxiety, worsening of anxiety symptoms is associated with a global decrease in WM microstructural integrity. Importantly, this relationship is independent of age and puberty. An extensive body of literature has shown that childhood and early adolescence are periods of significant WM growth across the brain [12, 65, 66]. Our present results demonstrate that within this overarching developmental pattern, individual variations in whole-brain WM are dynamically linked to childhood anxiety symptom severity. These findings support future studies investigating the possibility of targeting WM as a modality to aid in the prevention and treatment of childhood anxiety disorders.