Parsing Differences in Amygdala Volume Among Individuals with and without Social and Generalized Anxiety Disorders Across the Lifespan

Abstract

Structural differences in the amygdala (AMG) are implicated in anxiety and observed among individuals with generalized (GAD) and social anxiety (SAD) disorders. Findings have been mixed, perhaps because studies rarely examine differences between GAD and SAD, test comorbidity, or examine age-related differences. We tested AMG volume differences among a sample of adults and youth with/without SAD and GAD. Participants (N = 242; ages 7-60 years) completed an MRI scan, diagnostic interviews, and anxiety symptom measures. Groups were formed from diagnostic interviews: 1) Typically developing (TD; n = 91); 2) GAD (n = 53); 3) SAD (n = 35); and 4) comorbid SAD/GAD (n = 63). We used analysis of covariance with a bonferroni correction to examine group differences in AMG volume. The SAD and comorbid SAD/GAD groups exhibited increased bilateral AMG volume compared to the TD group. GAD and TD groups did not differ from each other in AMG size. The SAD, but not the comorbid SAD/GAD group, displayed greater right AMG size relative to the GAD group. SAD and comorbid SAD/GAD groups did not differ from the GAD group in left AMG volume. SAD and SAD/GAD groups did not exhibit different bilateral AMG size. Linear regression analyses demonstrated that greater social anxiety but not generalized anxiety symptom severity was associated with enlarged AMG volume. Age was not associated with AMG volume and nor did age moderate any group or symptom effects. Future longitudinal studies should examine whether larger AMG volume is a unique biomarker for SAD across the lifespan.

Introduction

Social anxiety disorder (SAD) and generalized anxiety disorder (GAD) are two of the most common anxiety conditions, are linked with increased rates of substance abuse and suicide, and reduced quality of life (Smit et al, 2006; Wittchen et al, 2011). Recent meta-analyses indicate that the age of onset of SAD ranges from 12-14 years old, with a mean onset of 14 (e.g., Legerstee et al., 2018). Conversely, the age of onset of GAD occurs later, with a mean onset of 34 years old, ranging from 30 – 39 years old. However, about 25% individuals develop GAD before age 17 (Legerstee et al., 2018). Crucially, SAD and GAD can exhibit chronic courses when untreated (Essau at el., 2014) and are associated with relatively low rates of spontaneous remission, ranging from 25% to 30% for GAD (Wittchen and Hoyer, 2001; Wittchen et al., 1994; Yonkers et al., 1996) and around 50% for SAD (e.g., Beesdo-Baum et al., 2012; Vriends et al., 2014). Of individuals with GAD and SAD that do experience periods of remission, the probability of recurrence remains relatively high — about 39% for SAD and 45% for GAD (e.g., Bruce et al., 2005). This research has indicated that SAD and GAD can follow chronic trajectories. From a neuroscience perspective, unremitting GAD and SAD symptom trajectories could suggest there are neural mechanisms that may canalize prior to adulthood, which could result in disorder onset and continuity. Research on neurostructural features of SAD and GAD across the lifespan has the potential to identify intrinsic areas of primary disturbance and enhance diagnostic approaches and precision in anxiety treatments.

The amygdala (AMG) is a central limbic structure involved in processing and learning about the emotional significance of environmental stimuli and detecting threat cues, and implicated to have a role in anxiety disorders (e.g., Duval et al., 2015). The AMG is operative in fear conditioning and extinction — central processes underlying the pathophysiology and treatment of anxiety disorders (Shin and Liberzson, 2010). Abnormalities in AMG structure are observed in children with GAD and SAD, as well as adults (Hilbert et al., 2014; Wang et al., 2018). However, AMG neurostructural patterns associated with GAD and SAD have been extremely mixed. Some studies find that youth with GAD exhibit reduced AMG volume relative to healthy controls (Milham et. al., 2005; Mueller et al., 2013), while others observe GAD to be associated with increased AMG volume (De Bellis et al., 2000; Qin et al., 2014). Other work has not detected any AMG volume differences between GAD and healthy controls (e.g., Gold et al., 2017). Findings from adult studies have been similarly inconsistent, documenting both larger and reduced AMG volumes in GAD (Etkin et al., 2009; Hettema et al., 2012; Schienle et al. 2011). Only a few investigations have examined AMG size in relation to SAD. Recent studies demonstrate that young adults with SAD, sub-threshold social anxiety symptoms, and inhibited or socially anxious tendencies exhibit larger AMG volumes (Clauss et al., 2014; Günther et al., 2018; Machado-de-Sousa et al., 2014; Tian et al., 216). Prior work has also found that young children with anxious/inhibited temperament, a risk factor for SAD, exhibit greater AMG volume relative to their less inhibited peers (Barrós-Loscertales et al., 2006; Qin et al., 2014). However, other studies have failed to detect AMG volume differences between healthy controls and those with SAD and social phobia symptoms (Syal et al., 2012), or have documented reduced AMG size among those with SAD (Foell et al., 2019; Irle et al., 2010; Meng et al.,2013).

Animal research has suggested that hyperresponsivity to environmental threats via fear conditioning could lead to increased dendritic arborization and reduced synaptic pruning in the AMG across development (Rosen and Schulkin, 1998; Vyas et al., 2002), which could increase the size of the AMG over time (e.g., Davidson and McEwen 2012) Likewise, fear extinction in mice predicts decreased total number and size of synapses in AMG, and subsequent AMG volume reductions (Heinrichs et al., 2013; Vetere et al., 2013). Relatedly in humans, a robust corpus of neuroscience evidence links SAD with increased AMG reactivity to social stimuli (e.g., Brühl et al., 2014). Increased grey matter AMG volume is positively correlated with AMG reactivity to social stimuli (Clauss et al., 2014; Månsson et al., 2016). Mansson et al. 2016 found that patients with SAD exhibited reductions in AMG volume following Cognitive Behavioral Therapy (CBT), and volume decreases mediated the direct effect of CBT in predicting reduced AMG hyperactivity to social stimuli following treatment. Since SAD is characterized by extreme fears and hypersensitivity to negative evaluation during social situations (American Psychiatric Association, 2013) it seems tenable that a similar etiological biological process could underlie increased AMG size and SAD (Machado-de-Sousa et al., 2014). In contrast, GAD is defined by constant and unfocused worries about ambiguous and negative future outcomes, rather than a specific perceived environmental threat (Duval et al., 2015). Research on associations between GAD and AMG activation patterns is mixed (Hilbert et al., 2014). Thus the relationship between GAD and AMG volume is less clear.

Taken together, the present study sought to enhance understanding of associations between AMG volume, SAD, and GAD in part by addressing limitations in prior work. First, studies rarely compare AMG volume of individuals with SAD relative to those with GAD, or tease apart potential patterns related to their comorbidity (e.g., Gold et al., 2015). Therefore there could be shared or specific structural characteristics associated with each disorder and/or their comorbidity that have yet to be demarcated. Second, although findings across pediatric and adult studies have been compared (e.g., Albaugh et al., 2017; Bas-Hoogendam et al., 2016), to our knowledge none have utilized a lifespan sample comprised of children and adults to test whether anxiety disorder associations with AMG volume are moderated by developmental age. There could be age-related differences in AMG volume that are associated with each disorder. Third, prior studies generally use small overall samples (e.g., full sample size Ns < 30-50; e.g., Montag et al., 2013), which could contribute to inconsistencies due to limited power to detect group differences, particularly if effects are in the small-to-moderate range. Toward addressing these limitations, we combined two pediatric (ages 7-19) and adult (ages 18-60 years) studies of anxiety disorders. Because both the studies were cross-sectional and used similar methodology, data could be harmonized to examine whether neurostructural differences in AMG volume distinguish youth and adults with SAD and/or GAD from their typically developing counterparts without histories of psychopathology, examine whether those with GAD and SAD demonstrate specific or common AMG volumetric features compared to each other, and if neurostructural traits varied when both conditions were present, or appeared to be more strongly associated with one disorder versus the other. Although the study was cross-sectional, we leveraged the wide age range of the sample (ages 7-60 years old) to delineate whether disorder-based AMG volume patterns demonstrated age-related differences across the lifespan. We categorized participants into four discrete groups based on their current diagnosis or healthy psychiatric status: (1) Typically Developing (TD) (i.e., those who have not developed a psychiatric disorder in their lifetime); (2) SAD; (3) GAD; and (4) comorbid SAD/GAD. We expected the three anxiety groups would differ from the TD group in AMG volume. Due to inconsistencies in the literature we did not have specific hypotheses for AMG volume differences between the anxiety disorder groups. We explored potential interactive effects between age and diagnostic group in relation to AMG volume. To explore these relationships dimensionally we also tested linear associations between overall and social anxiety symptom severity in relation to bilateral AMG size, and the moderating effects of age on these linear associations.

Materials and Methods

Participants

Participants included 242 youth and adults (M_age = 19.86, SD = 9.20, R = ages 7- 60 years old) who were enrolled in one of two studies focused on adult and pediatric anxiety disorders and other internalizing psychopathology at the University of Michigan (UM) and University of Illinois at Chicago (UIC). Participants were recruited from flyers, Internet postings, and outpatient psychiatric clinics. Across studies, participants completed self-report anxiety measures, diagnostic interviews, and resting MRI scans, which were all assessed at the same time point. For youth, diagnoses were obtained with the Schedule of Affective Disorders or Schizophrenia for School-Age Children (Kaufman et al., 1997), while for adults the Structured Clinical Interview for DSM-IV Axis-I (First et al., 2002) was administered. Across studies, trained master’s level or doctoral-level clinicians administered the diagnostic interviews. Individuals without any current or lifetime histories of psychopathology were categorized into the typically developing (TD) group (n = 91). Individuals categorized into the GAD group (n = 53) met DSM-IV Axis-I or DSM-5 criteria for current GAD. Participants in the SAD group (n = 35) met DSM-IV Axis-I or DSM-5 criteria for current SAD. The comorbid SAD/GAD group (n = 63) met criteria for current SAD and GAD. In the pediatric study, some of the anxious participants (e.g., those who met criteria for current GAD and/or SAD; n = 75) also met criteria for the following psychiatric disorders: Separation Anxiety Disorder (n = 10); Panic Disorder (n = 6); Obsessive Compulsive Disorder (n = 3); Specific Phobia (n = 13); Post-Traumatic Stress Disorder (PTSD; n = 2); Depressive Disorder (n = 4); Attention-Deficit Hyperactivity Disorder (n = 9); and Oppositional Defiant Disorder (n = 1). In the adult sample, some of the anxious participants (e.g., those who met criteria for current GAD and/or SAD; n = 76) also met criteria for Panic Disorder (n = 22); Obsessive Compulsive Disorder (n = 1); Specific Phobia (n = 14), PTSD (n = 17); Depressive Disorders (n = 39); Substance Abuse (n = 1); Alcohol Abuse (n = 2); Binge Eating Disorder (n = 1); and Agoraphobia (n = 5). Chi-square tests indicated that the proportions of psychiatric comorbidities were not significantly different across the diagnostic groups (Supplementary Material 1). Study exclusions included participants were not permitted to meet criteria for substance/alcohol dependence within the past 6 months or have histories of bipolar or schizophrenia, intellectual disability, and pervasive developmental disorders. The UM and UIC Institutional Review Boards approved study procedures and the study complied with ethical standards of the Helsinki Declaration of 1975, as revised in 2008. Informed consent was obtained from participants after study procedures were explained to them.

Anxiety Symptom Assessments

Youth were administered the Pediatric Anxiety Rating Scale (PARS), which is an interviewer-rated measure of anxiety symptom severity across seven dimensions (e.g., frequency and number of symptoms, overall symptom severity, physical symptom severity, avoidance, and interference at home and outside of the home) (Research Units on Pediatric Psychopharmacology Anxiety Study Group, 2002). Adults completed the Hamilton Anxiety Rating Scale (HAM-A; Hamilton, 1959), which is a clinician-rated scale of current anxiety symptoms. Adults completed the Liebowitz Social Anxiety Scale (LSAS; Liebowitz, 1987) and youth completed the child and adolescent version of the LSAS (LSAS-CA; Masia-Warner et al., 2003). The LSAS and LSAS-CA are self-report measures of social anxiety symptoms. Standardized scores were created for all symptom measures (HAMA-A, PARS, LSAS, LSAS-CA) and z-scores were aggregated to form overall anxiety (HAM-A, PARS) and social anxiety symptom (LSAS, LSAS-CA) composites. This procedure allowed us to examine differences in overall and social anxiety symptom severity across the anxiety groups, and conduct tests of anxiety symptom associations with AMG volume.

Structural MRI data acquisition and processing

At UIC, the structural MRI scans were obtained on a 3 Tesla GE Discovery System (General Electric Healthcare, Waukesha, WI) with an 8-channel head coil. Participants were positioned in a supine position in the scanner, and structural scans were acquired with a 3D BRAVO pulse sequence with the following parameters: flip angle 13°, inversion time 450 msec, field of view 22 x 22 cm, matrix size 256 x 256, slice thickness 1 mm³, 182 slices of the whole brain. At UM, a 3.0 T GE Signa Scanner (General Electric; Milwaukee, Wisconsin, USA) with a GE quad head coil was used to obtain high resolution, T1-weighted volumetric anatomical scans (3D spoiled-gradient echo sequence, 9 ms repetition time, 1.8 ms echo time, 500 ms inversion time, 15° flip angle, 256 x 256 matrix, 256 mm field of view; 124 slices, slice thickness 1.2 mm³). Across both studies and sites, participants were not exposed to any perceptual stimuli during the scans. Due to some differences in MRI data acquisition, study site location was included as a covariate in our statistical analysis. Using FreeSurfer Image analysis suite version 6.0 (http://freesurfer.net/fswiki/FreeSurferWiki), we employed automated procedures for subcortical reconstruction, volumetric segmentation and estimation, and extraction (Dale et al., 1999; Fischl et al., 2002). See Supplementary Material 2 for a detailed description of the structural image analysis

Statistical analyses

Study analyses were conducted in SPSS, Version 25.0 (IBM Corp., Armonk, N.Y., USA). We performed analysis of variance (ANOVA) and chi-square tests to examine differences across groups along demographic and clinical characteristics. We used general linear univariate analysis of covariance (ANCOVA) to examine the group differences in bilateral AMG volume and the two-way interaction between diagnostic group and age. Significant effects were followed up with post-hoc pairwise t-tests using a bonferroni correction to reduce the chance of Type I error. Bonferroni adjusted p-values are reported (e.g., .0125*4). Models statistically adjusted for estimated total intracranial volume, age, gender, site location (UIC versus UM), and study sample (pediatric versus adult study) on bilateral AMG volume¹. We included gender as a covariate in our model due to prior work that has demonstrated that males exhibit larger bilateral AMG volume relative to females (Herting et al., 2018; Uematsu et al., 2012; Wierenga et al., 2017). We conducted multiple linear regressions to examine whether overall and social anxiety symptom severity was associated with bilateral AMG volume while adjusting for the same covariates included in the ANCOVA models. We also tested two-way interactions between anxiety symptoms (overall, social) and age in relation to AMG volume (4 models total). For the regression interaction analyses, we used a bonferroni correction to control for multiple tests and the overall false discovery rate. The alpha criterion was set at .0125 (.05* ¼).

Results

Demographic and clinical characteristics

See Table 1 for all statistical comparisons across demographic and clinical characteristics and the numerical data. The comorbid SAD/GAD group was older than the TD and GAD groups. Although the majority of the sample identified as Caucasian (57%), there were small differences in racial distributions across groups. More participants who identified as African American were in the GAD versus TD group, and a greater number of participants whose racial identity was unknown were in the comorbid SAD/GAD group relative to the TD group. Anxious participants exhibited greater overall anxiety and social anxiety symptoms compared to the TD group, but did not differ from each other in overall anxiety symptoms. The SAD and comorbid SAD/GAD groups demonstrated greater social anxiety symptoms than the GAD group but did not differ from each other. Excluding GAD comorbidity, anxiety groups did not differ in their number of comorbid diagnoses. Within the pediatric sample, there were more TD participants relative to those in the comorbid SAD/GAD group. Within the adult sample, there were more participants in the comorbid SAD/GAD group relative to the TD group. Between studies, there were more adults compared to youth in the comorbid SAD/GAD group.

Table 1.

Demographic and clinical characteristics.

	TD (n = 91)		SAD (n = 35)		GAD (n = 53)		SAD/GAD (n = 63)		Omnibus test
	M	SD	M	SD	M	SD	M	SD	F	p value
Age	18.67a	8.91	19.69a,b	8.21	18.04a	8.78	22.64b	9.55	3.26	.02*
Overall Anxiety (z-score)	−1.13a	.23	.57b	.50	.67b	.54	.73b	.65	220.6^	< .01**
Social Anxiety (z-score)	−.99a	.29	.74b	.69	.07c	.62	.89b	.78	120.9^	< .01**
# comorbid diagnoses	-	-	.83a	1.1	.85a	1.1	1.17a	1.5	1.3	.28
	N	%	N	%	N	%			χ2	p value
Sex (% female)	53a	58.2	17a	48.6	34a	64.2	46a	73.0	6.57	.09
Ethnicity									7.97	.24
Hispanic	9a	9.9	7a	20.0	11a	20.4	14a	22.8
Non-Hispanic	81a	89	27a	77.1	42a	79.2	49a	77.8
Unknown	1a	1.1	1a	2.9	0a	0.0	0a	0.0
Race									32.13	.02*
Caucasian	44a	48.4	19a	54.3	34a	64.2	40a	63.5
African American	21a	23.1	4a,b	11.4	3b	5.7	11a,b	17.5
Asian	19 a	20.9	4a	11.4	5a	9.4	4a	6.3
Native Hawaiian/Pac. Isl.	0a	0.0	0a	0.0	0a	0.0	1a	1.6
American Indian/Ala. Nat.	1a	1.1	1a	2.9	0a	0.0	0a	0.0
More than 1 race	4a	4.4	1a	2.9	4a	7.5	2a	3.2
Other/unknown	2a	2.2	6b	17.1	7a,b	13.2	5a,b	7.9
Study Site									1.70	.64
UIC	66a	72.5	24a	68.6	36a	67.9	49a	77.8
UM	25a	27.5	11a	31.4	17a	32.1	14a	22.2
Within Study Dis.									8.49	.04*
Pediatric	55a	60.4	20a,b	57.1	31a,b	58.5	24b	38.1
Adult	36a	39.6	15a,b	42.9	22a,b	41.5	39b	61.9
Between Study Dis.									8.49	.04*
Adult (v. Pediatric)	36a (55)a	32.1(42.3)	15a(20)a	13.4(15.4)	31a(22)a	19.6 (23. 8)	39a(24)b	34.8(18.5)

Diagnostic group differences in right AMG volume

We detected a significant main effect of diagnostic group on right AMG volume (F (3, 241) = 4.86, p = .003, η²= .06; Figure 1), even while adjusting for estimated total intracranial volume, age, gender, site location, and study sample. The SAD (M = 1888.77, SE = 32.75, t = 3.18, p = .01, d = .64) and comorbid SAD/GAD groups (M = 1849.98, SE = 25.82, t = 2.55, p = .07, d = .42) exhibited greater right AMG volume compared to the TD group (M = 1769.43, SE = 22.10), although the difference between the comorbid SAD/GAD and TD groups was marginal. The GAD (M = 1776.02, SE = 27.26) and TD groups (t = .20, p = 1.0, d = .03) did not differ from each other in right AMG volume. The SAD group (t = 2.73, p = .04, d = .60), but not the comorbid SAD/GAD group (t = 2.07, p = .24, d = .39), displayed increased right AMG volume compared to the GAD group. The SAD and comorbid SAD/ GAD groups did not differ from each other in right AMG volume (t = .96, p = 1.0, d = .20). There was a significant main effect of gender (F (1, 241) = 7.10, p = .008, η²= .03). Males (M = 1860.37, SE = 23.61) exhibited larger right AMG volume than females (M = 1781.73, SE = 18.84). There was a significant effect of study sample on right AMG volume (F (1, 241) = 4.05 p = .05, η²= .02; M_pediatric = 1858.60, SE = 19.92; M_adult = 1783.50, SE = 27.84). There were no main effects of age or site location on right AMG volume (Fs ≤ 1.13, ps ≥ .29, η² < .01). The follow-up two-way interaction analysis between diagnostic group and age on right AMG volume was not significant (F (1, 241) = .03, p = .99, η² = .00).

Fig. 1.

Diagnostic group differences in right and left amygdala volume.

Diagnostic group differences in left AMG

We observed a significant main effect of diagnostic group on left AMG volume (F (3, 241) = 4.04, p < .01, η²= .05; Figure 1). The SAD (M = 1676.80, SE = 29.27, t = 2.70, p = .04, d = .56) and comorbid SAD/GAD groups (M = 1665.07, SE = 23.08, t = 2.80, p = .03, d = .46) exhibited greater left AMG volume than the TD group (M = 1585.99, SE = 19.75). We did not detect significant differences in left AMG volume between the GAD (M = 1605.88, SE = 24.37) and TD groups (t = .68, p = 1.0, d = .12). Neither the SAD (t = 1.92, p = .33, d = .43) nor the comorbid SAD/GAD (t = 1.86, p = .39, d = .35) groups displayed differences in left AMG volume relative to the GAD group, or between each other (t = .33, p = 1.0, d = .07). There was a main effect of gender (F (1, 241) = 16.23, p < .01, η² = .07). Males (M = 1686.59, SE = 21.11) displayed increased left AMG volume compared to females (M = 1580.29, SE = 16.84). There was a main effect of study sample (F (1, 241) = 6.30, p = .01, η²= .03). The pediatric sample (M = 1675.29, SE = 17.81) demonstrated larger left AMG volume than the adult sample (M = 1591.59, SE = 24.89). We did not detect main effects of age or site location on left AMG volume (Fs ≤ 2.03, ps ≥ .16, η² ≤ .01). The follow-up two-way interaction analysis between diagnostic group and age on left AMG volume was not significant (F (1, 241) = .80, p = .50, η² = .01).

Linear associations between anxiety symptoms and AMG volume

We performed multiple linear regressions to examine associations between anxiety symptom severity and AMG volume. Overall and social anxiety symptoms were significantly correlated (r = .72, p < .01). To reduce multicollinearity we examined associations between overall anxiety and social anxiety symptom severity and bilateral AMG size separately. Each model adjusted for the same covariates that were included in the ANCOVA analysis (e.g., study sample, age, gender, study location, estimated total intracranial volume). Overall anxiety symptom severity was not associated with right (β = .07, t = 1.12, p = .27) or left (β = .04, t = .71, p = .48) AMG volume. Social anxiety symptom severity was significantly and positively associated with increased right (β = .12, t = 2.04, p = .04, Figure 2) and left AMG size (β = .11, t = 1.94, p = .05, Figure 3). Next we conducted four separate two-way interaction analyses between anxiety symptoms (overall anxiety, social anxiety) and age in relation to bilateral AMG volume. Across all models there were no main effects of age in relation to AMG volume (βs ≤ .06 , ts ≤ .73 ps ≥ .46). None of the age x symptom two-way interaction effects were statistically significant (βs ≤ .09 , ts ≤ 1.53 ps ≥ .13).

Fig. 2.

Linear association between social anxiety symptom severity and right amygdala volume.

Fig. 3.

Linear association between social anxiety symptom severity and left amygdala volume.

Discussion

The present study is consistent with NIMH’s RDoC initiatives to utilize neuroimaging to examine neural characteristics and mechanisms of common psychiatric disorders, which could have the potential to inform neuroscience-based diagnostic and treatment approaches to mental illness (e.g., Insel et al., 2010). The present study represents several methodological advances in neurostructural research in anxiety disorders. Diverging from prior work that has focused on either pediatric or adult samples, our study utilized a relatively large lifespan sample (ages 7-60) to examine AMG structural differences between individuals with and without SAD and GAD, including those with both conditions, and to test for potential age-related differences in AMG volume across groups. Our main findings were the following: SAD and comorbid SAD/GAD groups displayed enlarged bilateral AMG volumes relative to TD individuals, the SAD group exhibited enlarged right AMG volume relative to the GAD group, and the comorbid group did not differ from the SAD or GAD groups in bilateral AMG size.

We were intrigued by the finding that comorbid SAD and GAD group exhibited larger AMG volume relative to the TD group, particularly since no differences between GAD and TD groups were detected. Therefore, one interpretation is that SAD, versus GAD, could be driving these effects. This finding is novel and has yet to be documented in prior work on anxiety disorders. Additionally, converging with our categorical findings, we found a unique concurrent positive gradient association between increased social anxiety symptom severity and enlarged bilateral AMG volume. These results demonstrated that the group differences in AMG volume that were detected also held dimensionally. Furthermore, these effects were specific to social anxiety as there was no concurrent association between overall anxiety symptom severity and bilateral AMG volume in the sample. One recent systematic review (Bas-Hoogendam et al., 2016) found evidence to suggest that the AMG functional and connectivity patterns that are associated with SAD evince trait-like lifespan stability, meet endophenotype criterion, and represent promising candidate mechanisms underlying SAD vulnerability. One future direction that should be explored is if increased AMG represents a specific biomarker, or endophenotype of SAD. This could be accomplished by examining patterns longitudinally to determine whether AMG volume enlargements are a biomarker of risk for SAD, or a correlate or consequence of the disorder.

In addition, this is the first study to demonstrate right side laterality in AMG volumetric differences between SAD and GAD. Importantly, the right AMG plays a critical role in learning and processing of fear-inducing stimuli. In contrast, the left AMG is associated with the processing of a broader set of affective cues (e.g., happiness, fear, sadness), and implicated in the reward system (Lanteaume et al., 2007; Murray et al., 2009). Therefore, structural abnormalities localized to the right side could be more uniquely associated with SAD as the disorder is characterized by heightened reactivity to threat-specific stimuli (e.g., social situations) whereas GAD is defined by generalized worry about a variety of topics (e.g., health, school, work; Duval et al., 2015). Although these findings are interesting, future investigations utilizing even larger samples are needed to further explore group-based differences, particularly in relation to SAD and GAD comorbidity.

AMG volume did not vary across developmental stages, as we did not detect any main or moderating effects of participant’s age on AMG volume across the statistical models. Research on age-related changes and differences in AMG volume across typically developing and clinical samples has been sparse. Some evidence with typically developing samples has indicated decreases in AMG volume with age (Walhovd et al., 2011), while others have found minimal change beyond early childhood and across participant ages (Good et al., 2001; Uematsu et al., 2012; Walhovd et al., 2011). One recent investigation conducted by Albaugh et al. 2017 found that in a sample of youth ranging from 4-18 years old that greater anxiety symptoms were associated with enlarged AMG volume and these relations were not qualified by age, which could suggest that relations between anxiety and AMG might be independent of developmental phase. Additional research is needed to further elucidate developmental and age-related patterns in AMG volume across individuals with and without anxiety disorders. Future studies may also benefit from focusing on narrower age ranges to delineate AMG volume differences within narrower developmental windows.

Despite the study’s many strengths, it is important that we acknowledge that there were some methodological limitations of the study. Given that the study was cross-sectional and correlational, we cannot conclude that AMG structural abnormalities represent causal mechanisms or vulnerability markers that contribute to SAD risk, or make any predictive claims. We also did not have information about the age of onset for either SAD or GAD, or for the other comorbid disorders. Although the overall sample size was comparatively larger than prior work in this area, the sample size for the SAD group was smaller relative to the other diagnostic groups. There were also differences in age across some of the diagnostic groups. Thus, future investigations should strive for more balanced groups in terms of diagnosis and age. Next, participants were allowed to have comorbid psychiatric disorders, which could have confounded our structural findings. However, the comorbid psychopathology types were statistically proportional across the anxiety groups. Furthermore inclusion of the different comorbid diagnoses in the analyses did not change the results, nor was any comorbid disorder uniquely associated with AMG volume. Finally, the majority of participants identified as Caucasian, which could reduce the generalizability of the detected group-based AMG volume differences to other racial and ethnic populations.

In sum, although the mechanisms that could underlie AMG enlargement in SAD still remain unclear, our study does preliminarily suggest that increased AMG volume is potentially more uniquely associated with SAD compared to GAD, and these structural patterns hold when examining social anxiety symptom severity and AMG relations dimensionally. There are several clinical implications of the study findings for future work. Aligned with recent calls for precision psychiatry and neuroscience-informed approaches to diagnosis and treatment (Frick et. al, 2020; Gabrieli et al., 2015), one potential future direction could be to utilize machine learning or predictive modeling approaches to examine if AMG volume is an objective biomarker that predicts SAD outcomes and treatment response. Future work may also consider using multiple data sets to perform cross-validation analyses to provide stronger evidence that enlarged AMG is uniquely associated with SAD relative to other anxiety disorders. Randomized treatment trials are also needed to delineate if AMG volume changes during treatment relate to decreases in SAD symptoms and predict better remission rates. Studies that seek to address these questions are clinically significant given SAD has a much earlier and narrower age of onset relative to other anxiety disorders, and is associated with a high risk of persistence and recurrence (Copeland et al., 2014; Legerstee et al., 2019). Additionally, cognitive behavioral therapy, the standard treatment for anxiety disorders, has been associated with worse outcomes for SAD (e.g., response rates ranging from 40% to 50%) in children and adults relative to other anxiety disorders (Ginsburg et al., 2011; Hudson et al., 2015; Norton and Price 2007; Rodebaugh et al., 2004). As some have proposed s (e.g., Scaini et al., 2016), current cognitive behavioral treatments may not be adequately tailored to alter the neurobiological mechanisms that are uniquely related to SAD symptomatology, which emphasizes the need for more neuroscience-informed investigations in this area that can enhance SAD prevention and treatment approaches.

Highlights:

• Social anxiety disorder was linked with larger amygdala size across the lifespan.

• Generalized anxiety and healthy individuals did not differ in amygdala size.

• Comorbid social and generalized anxiety was associated with larger amygdala volume.

• Greater social anxiety symptom severity was related to larger amygdala volume.

• Results could enhance clinical prediction models of SAD risk and treatment response.