Preliminary mapping of the structural effects of age in pediatric bipolar disorder with multimodal MR imaging

Highlights

• Multimodal structural MR imaging of brain maturation rates in pediatric bipolar disorder.

• Group differences found in left frontal pole cortical thickness.

• Group differences found in tensor FA and AD of fronto-temporal-striatal circuit.

• Group differences found multiple imaging metrics of the corpus callosum.

Abstract

This study investigates multimodal structural MR imaging biomarkers of development trajectories in pediatric bipolar disorder. T1-weighted and diffusion-weighted MR imaging was conducted to investigate cross-sectional group differences with age between typically developing controls (N = 26) and youths diagnosed with bipolar disorder (N = 26). Region-based analysis was used to examine cortical thickness of gray matter and diffusion tensor parameters in superficial white matter, and tractography-based analysis was used to examine deep white matter fiber bundles. Patients and controls showed significantly different maturation trajectories across brain areas; however, the magnitude of differences varied by region. The rate of cortical thinning with age was greater in patients than controls in the left frontal pole. While controls showed increasing fractional anisotropy (FA) and axial diffusivity (AD) with age, patients showed an opposite trend of decreasing FA and AD with age in fronto-temporal-striatal regions located in both superficial and deep white matter. The findings support fronto-temporal-striatal alterations in the developmental trajectories of youths diagnosed with bipolar disorder, and further, show the value of multimodal computational techniques in the assessment of neuropsychiatric disorders. These preliminary results warrant further investigation into longitudinal changes and the effects of treatment in the brain areas identified in this study.

Introduction

Research during the past two decades has demonstrated that bipolar disorder (BD) is an escalating problem among children and adolescents. While once thought to be rare, studies have demonstrated that 20–40% of adults diagnosed with BD report that their illness started during childhood or adolescence, rather than adulthood (Leboyer et al., 2005, Baldessarini et al., 2010, Baldessarini et al., 2012). Moreover, several studies have shown increasing numbers of children and adolescents are being diagnosed with BD. For example, the percentage of children and adolescents discharged from United States (U.S.) psychiatric hospitals with a diagnosis of BD has increased, from less than 10% in the mid-1990s to more than 20% in the mid-2000s (Blader and Carlson, 2007). This is not only an inpatient phenomenon, as Moreno et al. found a forty-fold rise in the incidence of outpatient visits made to providers of all specialties for a diagnosis of BD during a similar period, from 25/100,000 in 1993–1994 to 1003/100,000 in 2002–2003 (Moreno et al., 2007). Furthermore, this is not just a diagnostic trend confined to the U.S., as rates of German children under age 19 years old admitted to psychiatric hospitals for BD surged 68.5%—from 1.13/100,000 in 2000 to 1.91/100,000 in 2007, an increase greater than the general trend for mental health disorder admissions (Holtmann et al., 2010). To address the problem of rising numbers of children and adolescents being diagnosed with BD, we need greater understanding of the underlying neural mechanisms which could ultimately improve the specificity of our diagnostic and treatment approaches to pediatric BD. Mechanisms matter, as has been shown from other areas of biomedical research. For example, greater understanding of the mechanisms of childhood leukemia has resulted in better, more specific diagnostic approaches whereby suspicion of clinical symptoms (e.g., easy bruising, weakness, swollen abdomen) is confirmed (or ruled out) by a biomarker (e.g., a complete blood count). A similar approach using biological and behavioral mechanisms to augment clinical diagnosis could improve the specificity of how children are diagnosed as having BD (or not).

Towards that end, structural and functional magnetic resonance imaging (MRI) has been used as a tool to investigate such neural mechanisms, and previous studies have implicated the fronto-temporal-striatal circuit in pediatric BD. Specifically, structural MRI studies have found significantly decreased gray matter volume of the amygdala vs. typically-developing controls (TDC) (Blumberg et al., 2003, DelBello et al., 2004, Chang et al., 2005, Dickstein et al., 2005, Pfeifer et al., 2008, Hajek et al., 2009), as well as some studies showing similar decreases in the left dorsolateral prefrontal cortex (DLPFC) (Dickstein et al., 2005). Functional MRI (fMRI) studies have shown fronto-temporal-striatal alterations in tasks tapping functions including face processing (Dickstein et al., 2007, Pavuluri et al., 2007, Rich et al., 2008, Brotman et al., 2009, Kalmar et al., 2009, Kim et al., 2012, Wiggins et al., 2017), response inhibition (Leibenluft et al., 2007a, Leibenluft et al., 2007b, Singh et al., 2010a, Singh et al., 2010b), response to frustration (Deveney et al., 2013, Rich et al., 2010), cognitive flexibility (Dickstein et al., 2010; Adleman et al., 2011), and resting state connectivity (Stoddard et al., 2015).

Diffusion-weighted MRI (dMRI) is a complementary structural neuroimaging modality used to characterize brain white matter by probing patterns of water molecule motion in tissue. Imaging metrics derived from dMRI datasets provide a unique probe for detecting changes in myelination, fiber organization, and axonal morphometry (Basser and Pierpaoli, 1996, Beaulieu, 2002), and previous dMRI studies have used them to reveal disruptions in tissue microstructure and connectivity that are related to BD. Of the approximately 16 currently published original data dMRI studies of BD youths, all have examined tensor-derived imaging metrics and shown decreased, rather than increased, fractional anisotropy (FA)—a scalar value of the degree of anisotropy of the diffusion process, with zero indicating isotropic diffusion, and one indicating anisotropic/restricted diffusion—and disrupted white matter integrity. This includes decreased FA among BD vs. TDC youths in the anterior corona radiata (Adler et al., 2006, Pavuluri et al., 2009, Lagopoulos et al., 2013), the corpus callosum (Ishida et al., 2017, James et al., 2011, Saxena et al., 2012, Lagopoulos et al., 2013), white matter adjacent to cingulate cortex (Frazier et al., 2007, Gao et al., 2013, Gönenç et al., 2010), white matter areas adjacent to prefrontal cortex (PFC) and orbitofrontal cortex (Adler et al., 2004, Kafantaris et al., 2009). BD youths also have decreased FA in deep white matter regions, including the fornix (Barnea-Goraly et al., 2010) and internal capsule (Lu et al., 2012). Predictive modeling has also shown promising results. For example, in the recent study by Mwangi et al. that trained a support vector machine (SVM) algorithm to predict the diagnostic status of youths with BD (N = 16) vs. TDC (N = 16) youths with 87.5% specificity, 68.75% sensitivity, 78.12% accuracy, and 84.62% positive predictive value (Mwangi et al., 2014). Taken in whole, these studies suggest that decreased FA and disrupted white matter integrity may be involved in the pathophysiology of pediatric BD.

Moving the field forward requires that studies integrate these different MRI modalities, as it can help to more comprehensively evaluate the effect of development on BD vs. TDC alterations across different structures and tissue types. With respect to integrating T₁-weighted and diffusion-weighted MRI, James et al. found decreased gray matter density by voxel based morphometry (VBM) in the left orbitofrontal cortex and decreased FA in the anterior cingulate cortex by tract based spatial statistics (TBSS) among BD (N = 15) vs. TDC (N = 20) youths (James et al., 2011). With respect to potential developmental effects, Toteja et al. found lower FA and higher mean diffusivity (MD; the average total diffusion within a voxel) among BD vs. controls (14 youths and 43 adults in each group), and significantly greater MD among BD participants at all ages than controls (whereas this was not significant for FA) (Toteja et al., 2015).

In the present study, we sought to conduct a preliminary analysis to extend the scope of such integrated studies by performing a more comprehensive analysis of both structural T₁-weighted MRI and dMRI among BD and TDC youths ages 8–17 years old. Based on prior studies, we hypothesized that BD youths would have decreased fronto-temporal-striatal thickness and decreased white matter connectivity among these regions, and that the change of imaging metrics across age is a potentially useful biomarker. To test these hypotheses, we employed novel computational image analyses to characterize regional variation in both cortical thickness and diffusion MR metrics reflecting variation in white matter microstructure and connectivity. Unlike other prior studies, such as those using tract-based spatial statistics (TBSS), the present study unites both structural MRI via cortical thickness with dMRI via multi-compartment tractography modeling (Behrens et al., 2007), which can help resolve complex sub-voxel fiber configurations in white matter and improve the anatomical accuracy and reliability of bundle reconstructions (Cabeen et al., 2016).