Introduction
Anxiety disorders are common in the general population, with lifetime prevalence estimates reaching as high as 28% (Baxter et al., Reference Baxter, Scott, Vos and Whiteford2013). Many anxiety disorders have an average onset before age 15 (de Lijster et al., Reference de Lijster, Dierckx, Utens, Verhulst, Zieldorff, Dieleman and Legerstee2017). Thus, research aimed at understanding etiological factors that give rise to anxiety disorders in childhood and early adolescence has important implications for assessments and interventions for these conditions. In line with the idea of fearfulness and anxiety-proneness as distinct trait dispositions (Sylvers et al., Reference Sylvers, Lilienfeld and LaPrairie2011), there is evidence that anxiety disorders entailing excessive fear are distinguishable from those involving nonspecific distress and pervasive negative affect. The former are associated with hyperreactivity to threat cues and avoidance behavior, the latter more with dysphoric mood and depressive symptoms (Watson, Reference Watson2005). The current work focuses on the construct of threat sensitivity as a specific liability for fear-related psychopathology. Specifically, we sought to extend prior twin research demonstrating a heritable basis for threat sensitivity’s association with phobic-fear symptomatology in adulthood to younger-aged individuals (i.e., children aged 9 to 14).
An important referent for the current work is research demonstrating that anxiety disorders entailing excessive fear (specific phobia, social anxiety disorder, panic disorder, agoraphobia) are highly comorbid and can be organized along a coherent “phobic fear’” dimension. This dimension is distinct from, though correlated with, a “dysphoric distress” dimension reflecting symptoms of low mood and pervasive negative affect and cognitions, associated with disorders including major depression, persistent depressive (dysthymic) disorder, generalized anxiety disorder (GAD), and post-traumatic stress disorder (Kotov et al., Reference Kotov, Krueger, Watson, Achenbach, Althoff, Bagby, Brown, Carpenter, Caspi, Clark, Eaton, Forbes, Forbush, Goldberg, Hasin, Hyman, Ivanova, Lynam, Markon, Miller and Zimmerman2017, Reference Kotov, Krueger, Watson, Cicero, Conway, DeYoung, Eaton, Forbes, Hallquist, Latzman, Mullins-Sweatt, Ruggero, Simms, Waldman, Waszczuk and Wright2021). Twin-biometrical modeling studies utilizing both adult and youth twin samples have found that unique additive genetic influences contribute to variance in fear disorder symptomatology and related individual difference measures (Hettema et al., Reference Hettema, Bourdon, Sawyers, Verhulst, Brotman, Leibenluft, Pine and Roberson-Nay2020; Kendler et al., Reference Kendler, Walters, Neale, Kessler, Heath and Eaves1995; Lahey et al., Reference Lahey, Krueger, Rathouz, Waldman and Zald2017). Moreover, research employing psychophysiological measures has shown that patients with fear disorders, but not dysphoric distress disorders, tend to exhibit enhanced potentiation of the startle reflex – a well-established measure of limbic threat reactivity in humans as well as animals (Lang and Davis, Reference Lang and Davis2006) – during exposure to aversive cues (Vaidyanathan et al., Reference Vaidyanathan, Patrick and Cuthbert2009b; see also Vaidyanathan et al., Reference Vaidyanathan, Patrick and Bernat2009a). This finding has been observed across fear-related traits and psychological disorders involving excessive fear (for a review, see (Vaidyanathan et al., Reference Vaidyanathan, Patrick and Bernat2009a), suggesting that these characteristics/conditions may have a basis in common underlying neurophysiological mechanisms along with shared genetic influences. Taken together, these lines of research converge on the notion that fear-related disorders share a common premorbid liability factor that confers specific risk for disorders of this type and is at least partially rooted in neurobiology.
Operating from this perspective, Kramer et al. (Reference Kramer, Patrick, Krueger and Gasperi2012) applied structural equation and biometrical modeling to adult twin data for a number of well-established self-report scale measures of fear and fearlessness as a basis for conceptualizing this trait domain and characterizing it etiologically and neurobiologically. The scales selected for their analysis were ones that prior research had shown to be related to aversive startle potentiation, a physiological-reflex measure of threat sensitivity (Vaidyanathan et al., Reference Vaidyanathan, Patrick and Cuthbert2009b). Kramer et al.’s structural analysis revealed a broad bipolar dimension, termed “trait fear” by the authors, accounting for appreciable variance in all scales. A biometrical analysis revealed that additive genetic influences accounted for 51% of the variance in scores on this broad fear/fearlessness factor. Additionally, Kramer et al. presented evidence that scores on this trait fear dimension were positively related to aversive startle potentiation, establishing linkage with biobehavioral threat sensitivity.
Using data for another adult twin sample, Yancey et al. (Reference Yancey, Vaidyanathan and Patrick2015) showed that increased aversive startle potentiation in persons with a history of one or more fear disorder diagnoses (specific phobia, social anxiety disorder, panic disorder, agoraphobia; cf. Vaidyanathan et al., Reference Vaidyanathan, Patrick and Cuthbert2009b), relative to those with no fear disorder history, was accounted for by elevated trait fear scores. Interestingly, the association between fear disorder history and startle potentiation, and trait fear’s mediating role in this association, was evident only for participants with no history of major depression. The implication is that trait fear shares a component of biological threat sensitivity with fear-related psychopathology that is distinct from internalizing disorders characterized by dysphoric distress.
Subsequent to this, Yancey et al., (2016) demonstrated in the same sample that scores on a scale measure of trait fear covaried with other physiological indicators of threat sensitivity besides startle potentiation (i.e., cardiac, facial electromyographic) and that scores for these different physiological indicators could be combined together with fear-scale scores into a biobehavioral index of dispositional fear. Scores on this biobehavioral fear index were positively predictive of both clinician-assessed fear disorder symptoms and separate measures of threat sensitivity including electrocortical measures. A further study by Venables et al. (Reference Venables, Hicks, Yancey, Kramer, Nelson, Strickland, Krueger, Iacono and Patrick2017) applied twin-biometrical modeling to data for this twin sample and showed a very high genetic correlation (∼ .8) for the biobehavioral fear index with fear disorder symptom scores, compared to moderate and negligible genetic correlations, respectively, with dysphoric and substance disorder symptoms.
The foregoing lines of evidence point to an individual difference dimension of threat sensitivity as a specific liability for fear-related psychopathology. However, research to date supporting this view has been limited to adult samples. Accordingly, there is a need to evaluate threat sensitivity’s association with fear disorder symptomatology in youth and characterize the role of genetic versus environmental influences in this association. We know from prior research that precursors to certain adult personality traits, conceptualized as early temperament characteristics, appear to be prominently heritable and biologically based (Buss and Plomin, Reference Buss and Plomin1984). Sensitivity to fearful or threatening stimuli has been identified as one such characteristic in prominent theories of temperament (Buss and Plomin, Reference Buss and Plomin1986; Goldsmith and Campos, Reference Goldsmith, Campos, Emde and Harmon1982; Kagan, Reference Kagan1997; Kochanska et al., Reference Kochanska, Aksan and Joy2007; Kochanska, Reference Kochanska1997; Tellegen, Reference Tellegen, Tuma and Maser1985). Operating from this temperament-based conceptualization, the current study employed data for a child-aged twin sample to examine the diagnostic correlates of dispositional threat sensitivity and examine the role of genetic versus environmental influences in its relationship with fear disorders compared to other disorders. We first assessed the internal psychometric properties of an established measure of threat sensitivity, the 20-item Trait Fear scale (TF-20; Kramer et al., Reference Kramer, Patrick, Hettema, Moore, Sawyers and Yancey2020), in terms of its item-level structure, internal consistency, and reliability in our child-aged study sample. Next, we evaluated the specificity of the TF-20’s relations with fear disorder symptomatology, relative to dysphoric-distress and externalizing disorder symptomatology. Finally, we examined its underlying latent genetic and environmental structure in relation to internalizing disorders in youth, disaggregating the shared versus specific etiological influences across these two domains (i.e., trait scores and symptom dimensions) using twin-biometrical modeling.
Method
Participants
Data for the current study derived from the Virginia Commonwealth University Juvenile Anxiety Study, a twin study of anxiety and related phenotypes in White children aged 9–14 (692 individuals from 121 monozygotic [MZ] and 225 dizygotic [DZ] pairs). Twin pairs were recruited through the Mid-Atlantic Twin Registry (MATR; Lilley and Silberg, Reference Lilley and Silberg2013) and enrolled at either the VCU site or the National Institute of Health (NIH) site. Zygosity was determined using parental responses to standard questions about physical appearance for all 346 twin pairs and DNA testing for a subset of 13 pairs, with results for the two methods converging perfectly in that subset of participants (for details, see Carney et al., Reference Carney, Moroney, Machlin, Hahn, Savage, Lee, Towbin, Brotman, Pine, Leibenluft, Roberson-Nay and Hettema2016). A variety of self- and parent-report measures were collected via REDCap (Research Electronic Data Capture; Harris et al., Reference Harris, Taylor, Thielke, Payne, Gonzalez and Conde2009). A masters or doctoral level trained clinician administered the Kiddie Schedule for Affective Disorders and Schizophrenia-Present & Lifetime Version (KSADS-PL; Kaufman et al., Reference Kaufman, Birmaher, Brent, Rao, Flynn, Moreci, Williamson and Ryan1997) in relation to each twin through a face-to-face interview with the parent(s).
Following the initial visit, a portion of participants (n = 241) were re-assessed to evaluate test-retest reliability. Six of these with intervals exceeding 90 days between visits were excluded from reliability analyses, resulting in retest data for 235 children, for whom the mean time between visits was 25.9 days (SD = 10.19; range = 11–70). The subset of participants who returned for visit 2 did not differ significantly from the overall visit 1 sample in terms of sex, age, or zygosity (ps = .65, .80, and .82, respectively).
The Institutional Review Boards at VCU and NIH approved the study, and all participants provided informed consent (parents) and assent (children) prior to testing. Twins and parents were monetarily compensated for their participation in the study.
Measures
20-Item trait fear inventory (TF-20)
The TF-20 (Kramer et al., Reference Kramer, Patrick, Hettema, Moore, Sawyers and Yancey2020) is an abbreviated version of a 44-item self-report scale developed to index the broad fear/fearlessness (dispositional threat sensitivity) dimension identified by Kramer et al. (Reference Kramer, Patrick, Krueger and Gasperi2012). Each item is rated on a 4-point Likert scale, from 0 (false) to 3 (true). Items indexing fearlessness were reversed scored, and a sum score was computed for the scale as a whole – with possible total scores ranging from 0–60 and higher scores representing greater fearfulness. Kramer et al. (Reference Kramer, Patrick, Hettema, Moore, Sawyers and Yancey2020) reported that scores on this scale correlated robustly with fear disorder symptomatology, at a level (r = .44) that exceeded its associations with dysphoric/distress and substance use symptomatology (rs = .29 and −.16).
Other measures
Other child self-report measures administered to participants included: the Revised Fear Survey Schedule for Children (FSSC-R; Ollendick, Reference Ollendick1983), the Screen for Child Anxiety Related Disorders (SCARED; Birmaher et al., Reference Birmaher, Brent, Chiappetta, Bridge, Monga and Baugher1999), Carver and White’s (Reference Carver and White1994) Behavioral Inhibition System/Behavioral Activation System (BIS/BAS) scales, the Affective Reactivity Index (ARI; Stringaris et al., Reference Stringaris, Goodman, Ferdinando, Razdan, Muhrer, Leibenluft and Brotman2012), the Short Mood and Feelings Questionnaire (SMFQ; Angold et al., Reference Angold, Costello, Messer, Pickles, Winder and Silver1995), the Junior Eysenck Personality Questionnaire (JEPQ; Eysenck, Reference Eysenck1965), and the Child Anxiety Sensitivity Index (CASI; Silverman et al., Reference Silverman, Fleisig, Rabian and Peterson1991). Parent-report measures included: the Behavioral Inhibition Questionnaire (BIQ; Gensthaler et al., Reference Gensthaler, Möhler, Resch, Paulus, Schwenck, Freitag and Goth2013), the Inventory of Callous-Unemotional Traits (ICU; Kimonis et al., Reference Kimonis, Frick, Skeem, Marsee, Cruise, Munoz, Aucoin and Morris2008), and the Child Behavior Checklist (CBCL; Achenbach and Rescorla, Reference Achenbach and Rescorla2001), along with parent-adapted versions of the SCARED, SMFQ, and ARI (see “Descriptions of Study Measures” section of the online Supplement for information regarding these measures, and Supplemental Table A for Ms/SDs and reliability coefficients [α, test-retest] for each).
Statistical analyses
Exploratory structural equation modeling (ESEM), implemented via MPlus version 7.31 (Muthén and Muthén, Reference Muthén and Muthén1998–2015) with adjustment for potential clustering of responses within families to account for non-independence of twin observations, was used to characterize the factor structure of the TF-20 (Kramer et al., Reference Kramer, Patrick, Hettema, Moore, Sawyers and Yancey2020). We also conducted reliability and validity analyses of the TF-20 using the R software package (R Core Team, 2015). Test-retest reliability of TF-20 sum scores between visit 1 and visit 2 was evaluated using Pearson’s product-moment correlation. In addition, we estimated internal consistency reliability of scores for each assessment visit via Cronbach’s α, using the psych package in R. To assess convergent and discriminant validity, we computed correlations between sum scores for the TF-20 at visit 1 and scores for the various other construct-relevant measures administered at that initial visit. We also examined criterion-related validity in terms of polyserial correlations between TF-20 sum scores and current and past psychiatric diagnoses from the KSADS. To verify that findings from these reliability and validity analyses were not overly affected by twin dependency, we compared results for all participants with results for a subsample consisting of one twin randomly selected from each pair.
To further evaluate the criterion-related validity of the TF-20, we examined its relations with correlated factors corresponding to phobic fear and dysphoric distress, derived from a confirmatory factor analysis (CFA; conducted in Mplus version 8.6) of child-report scores for the following scales: FSSC-R, SCARED, and BIS as indicators of phobic fear, and ARI, SMFQ, and JEPN as indicators of dysphoric distress. Regression analysis was used to quantify relations of TF-20 scores with each factor of this model when controlling for its overlap with the other factor. Additionally, we added TF-20 sum scores to the model as a fourth indicator of the latent fear factor to evaluate its loading on this factor. Twin interdependency was accounted for in these analyses by adjusting for clustering of responses within families, and participant sex and age were included as covariates to account for any potential influence of these variables.Footnote 1
Within twin pairs, we applied biometrical structural equation modeling (SEM; Neale and Cardon, Reference Neale and Cardon1992) to decompose the observed variance of TF-20 scores into additive genetic, common environmental, and unique environmental effects. Additive genetic effects (A) reflect the combined contributions of different gene alleles to an observed trait. Common environment effects (C) comprise nongenetic influences that make twins more similar to each other compared to the general population. Unique environmental effects (E) represent influences that contribute to observed differences between co-twins and include measurement error. Using this modeling approach, we estimated the heritability of the TF-20 (i.e., proportion of variance in scores attributable to additive genetic influences) and the extent to which its observed covariance with child-reported fear measures (FSSC-R, SCARED, BIS) could be accounted for by overlapping genetic versus environmental influences. A common pathway model was then used to estimate the etiologic overlap of the TF-20 with the aforementioned fear and dysphoria latent factors.Footnote 2
We conducted these twin biometric modeling analyses using the OpenMx package (Neale et al., Reference Neale, Hunter, Pritikin, Zahery, Brick, Kirkpatrick, Estabrook, Bates, Maes and Boker2016) in R. Customary alternative models consisting of ACE, AE, CE, and E were specified. Variance components for each of these twin models were estimated using the direct symmetric approach recommended by Verhulst et al. (Reference Verhulst, Prom-Wormley, Keller, Medland and Neale2019), in which variance-covariance matrices for variance components are estimated directly, without constraints. Participant sex and age were included as covariates to account for any potential influence of these variables. Model parameters reflecting genetic and environmental sources of variance were sequentially constrained to zero to determine the best-fitting, most parsimonious model. Comparative fit was evaluated using the likelihood ratio χ2 and Akaike information criterion (AIC) statistics.Footnote 3
Results
Sample characteristics and basic statistics for TF-20 scores
The mean age of participants in the analytic sample at visit 1 was 11.23 (SD = 1.43; range = 9–14.5), with 52.4% being female. The mean TF-20 scores at visit 1 and visit 2 were 29.8 (SD = 10.5; range = 2–59) and 28.6 (SD = 12.1; range = 0–59), respectively. Figure 1 displays the distribution of TF-20 sum scores in the visit 1 sample. The distribution of scores for the sample as a whole approximated normality (skew = −.11; kurtosis = −.36).
Distribution of TF-20 sum scores in the overall sample (Visit 1).
Factor structure of TF-20 items
We conducted an ESEM corrected for non-independence of twins using TF-20 data from the visit 1 assessment. Based on an initial exploratory factor analysis revealing the presence of one large factor and two much smaller factors underlying the TF-20 item set, we specified a bifactor ESEM with geomin rotation in which items loaded onto a general factor (corresponding to dispositional threat sensitivity; Kramer et al., Reference Kramer, Patrick, Hettema, Moore, Sawyers and Yancey2020) and two orthogonal subfactors. The model fit well (CFI = .962, TLI = .946, RMSEA = .035 [90% CI: .028–.041]), with all but one item loading above .3 on the general factor. Table 1 lists the wording of the items and their loadings on the general factor of this model.
TF-20 item wording and factor loadings for exploratory structural equation model
Note. (R) indicates item is reverse coded. Item responses range from 0–3. SE = standard error. All loadings are significant at p < .001.
Test-retest reliability and internal consistency
The correlation between TF-20 total scores across visits 1 and 2 was r = .83 (95% CI: .79–.87), indicating good test-retest reliability for the TF-20 total score.Footnote 4 Internal consistencies for the TF-20 total at each visit were similarly high: α = .86 (95% CI: .85–.88) for visit 1, and .90 (95% CI: .89–.92) for visit 2.Footnote 5
Convergent & discriminant validity
Table 2 lists between-subject correlations between TF-20 total scores and the various continuous-score criterion measures at the initial (T1) assessment point (N = 692). Providing evidence for convergent validity, the TF-20 sum score was strongly and significantly correlated with child-report scales that tap fear symptomatology and associated processes: FSSC-R (.63), BIS (.56), SCARED (.52), and CASI (.44). Indicative of discriminant validity, correlations with child-report scales assessing dysphoric/depressive symptoms (SMFQ, ARI) were significantly lower (rs = .27 and .23, respectively) than those for fear symptom scales (i.e., Steiger’s Z-test for dependent correlations was significant at p < .001 for all comparisons).
Correlations between self-report TF-20 and other self/parent-report measures
Note. Not all scales were administered to both parents and children, as indicated by empty cells. FSSC-R = Revised Fear Survey Schedule for Children; SCARED = Screen for Child Anxiety Related Disorders; BIS/BAS = Behavior Inhibition System/Behavioral Activation System scales; CASI = Child Anxiety Sensitivity Index; ARI = Affective Reactivity Index; SMFQ = Short Mood and Feelings Questionnaire; JEPQ = Junior Eysenck Personality Questionnaire; BIQ = Behavioral Inhibition Questionnaire; ICU = Inventory of Callous-Unemotional Traits; CBCL = Child Behavior Checklist.
The TF-20’s correlations with parent-report versions of the BIS and SCARED (i.e., BIQ and parent-SCARED) were also significant, but expectably smaller (rs = .23 and .27, respectively) given that TF-20 scores were available only in child self-report form. Corresponding correlations with parent-report versions of the SMFQ and ARI were weak and nonsignificant (rs = .07 and −.01). The TF-20’s correlation with another measure of dysphoric symptomatology collected from parents only, the Withdrawn/Depressed scale of the CBCL, was .13, p < .005.
Among clinical diagnoses (Table 3), TF-20 scores were positively correlated with the following current or past diagnoses: specific phobia, social phobia, separation anxiety, and GAD (rs = .16–.32, all ps < .05). The TF-20 was not associated significantly with diagnoses of major depression or post-traumatic stress disorder, or with any externalizing disorder diagnosis.
Polyserial correlations between self-report TF-20 and disorder diagnoses
Note. N = total number exhibiting disorder in past (left of slash) or currently (right of slash). NA indicates correlation was not computed due to small number of cases (< 5).
Additional support for convergent validity was provided by results of the CFA we conducted using the FSSC-R, SCARED, and BIS as indicators of a phobic fear symptom dimension and the SMFQ, ARI, and JEPQ-N as indicators of a dysphoria/distress symptom dimension. The 2-factor model specifying these two symptom dimensions as correlated latent factors provided acceptable fit to the data (CFI = .980, TLI = .931, RMSEA = .079 [90% CI: .058–.101]). Regression analyses revealed a very strong positive relationship for the fear factor of this model with TF-20 scores (B = .96, p < .001) when accounting for its overlap (covariation) with the dysphoria factor, in contrast with a modest negative relationship for the dysphoria factor when accounting for its overlap with the fear factor (B = −.33, p < .001). Additionally, when added to the model as an indicator of the latent fear factor, the TF-20 sum-score measure evidenced a very strong loading on this factor (.65). The factor loadings for this CFA model with and without the TF-20 are listed in Supplemental Table B.
As shown in Table 2, discriminant validity was evidenced by small negative correlations for the TF-20 with measures of externalizing symptoms (i.e., parent-rated CBCL rule-breaking and aggressive behavior scales, rs = −.11 and −.09, respectively, and externalizing composite score r = −.10) and associated traits of impulsive reward seeking (i.e., child-report BAS scale, r = –.17) and callous-unemotionality (i.e., parent-rated ICU, r = −.12).Footnote 6
Twin biometric modeling
To disaggregate genetic and environmental influences on the TF-20, we specified full ACE univariate and multivariate twin models. Individual parameters in the univariate model were constrained to zero to test their significance, and AIC values were then compared (Table 4). The best-fitting model was model II (AE model), as indicated by the lowest AIC and a nonsignificant decrement in fit relative to model I (p = 1.0). Additive genetic influences accounted for approximately 43% (95% CI = 33–53%) of variance in TF-20 scores, with the remaining 57% of variance (95% CI = 46–66%) attributable to unique environmental factors.
Twin model fit statistics for TF-20
Note. −2LL = negative two log likelihood; DF = degrees of freedom; AIC = Akaike information criterion; ACE = full model with genetic (A), common environmental (C), and nonshared environmental (E) domains included; AE = submodel with only genetic and unique environment included; CE = submodel with only common and unique environment included; E = submodel with only unique environment included. Bolding denotes the best-fitting model (i.e., AE).
We investigated genetic and environmental overlap between the TF-20 and other internalizing measures that showed at least moderate phenotypic overlap. This was accomplished with two models: a multivariate correlated-factors model that examined the role of genetic versus environmental influences in observed relations among the four fear/anxiety measures (TF-20, FSSC-R, SCARED, BIS), and a correlated common pathway model that estimated the overlap in these sources of influence between TF-20 and latent factors corresponding to phobic fear and dysphoric distress symptomatology. A full ACE model was initially fit to the four fear-related measures; however, negative variances were estimated for the common environment (C) factors for TF-20, FSSC-R, and BIS. Based on the best-fitting univariate models of TF-20, FSSC-R, and BIS, we constrained C to be zero in the multivariate correlated factors model (estimated parameters = 32, −2LL = 19,008.25, DF = 2642, AIC = 13,724.249). Proportions of variance in each measure accounted for by genetic and environmental factors are presented in Table 5, and genetic/environmental correlations between the different measures are shown in Table 6. The etiological correlation estimates between scores for the TF-20 and the other internalizing measures (r g s = .71, .56, and .54 for FCCS-R, SCARED, and BIS, respectively) indicate a moderate degree of genetic overlap between the TF-20 and these fear/anxiety measures.
Multivariate correlated factors model: proportions of variance in each scale measure accounted for by genetic and environmental factors
Multivariate correlated factors model: genetic and environmental correlations. Between fear-related scale measures
Lastly, we fit a correlated common pathway twin model specifying latent factors of fear and dysphoric symptomatology based on previous work by our group (Hettema et al., Reference Hettema, Bourdon, Sawyers, Verhulst, Brotman, Leibenluft, Pine and Roberson-Nay2020) (estimated parameters = 45, −2LL = 28,186.14, DF = 4467, AIC = 19,252.14). The proportions of variance and genetic/environmental correlations are displayed in Figure 2. Overall, strong overlap was found between the TF-20 and the latent fear factor in terms of both genetic (.68) and environmental correlations (.69). While the TF-20 showed an environmentally based association with the dysphoria factor (.62), there was essentially no genetic overlap (r = .01).
Common pathway model examining the etiological overlap between the TF-20 and latent factors defined by measures of fear and dysphoria. Single-headed arrows denote proportions of variance due to genetic (A) or environmental (E) factors. Double-headed arrows denote correlations between those sources of variance. Subscripts F and D on common A and E factors (top level) denote fear and dysphoria, respectively. Subscripts Si (i = 1–6) on A and E factors (bottom level) denote sources of variance specific to each of the six measured phenotypes: SCARED = Screen for Child Anxiety Related Disorders; FSSC-R = Revised Fear Survey Schedule for Children; BIS = Bbehavior Inhibition System; ARI = Affective Reactivity Index; SMFQ = Short Mood and Feelings Questionnaire; JEPQ-N = Junior Eysenck Personality Questionnaire – Neuroticism scale.
Discussion
The current study used data from a large community sample of child-aged twins to evaluate the psychometric properties of an established measure of dispositional threat sensitivity, the 20-item Trait Fear scale (TF-20; Kramer et al., Reference Kramer, Patrick, Hettema, Moore, Sawyers and Yancey2020), in youthful participants and examine its phenotypic relations with fear versus dysphoric-distress disorder symptoms and diagnoses and evaluate the role of overlapping genetic and environmental influences in these observed relations.
The TF-20 evidenced excellent psychometric properties in our 9 to 14 year old twins. Consistent with findings for young adults (Kramer et al., Reference Kramer, Patrick, Hettema, Moore, Sawyers and Yancey2020), sum scores for the TF-20 were close to normally distributed, and its items as a whole loaded quite uniformly onto a general factor interpretable as dispositional threat sensitivity. The normality of the score distribution accords with the concept of threat sensitivity as a continuous dimensional trait, ranging from low to high – with moderate levels occurring most commonly in the population (Kramer et al., Reference Kramer, Patrick, Krueger and Gasperi2012; Patrick et al., Reference Patrick, Durbin and Moser2012). High reliability was evident for overall TF-20 scores, in terms of both test-retest reliability from study visit 1 to visit 2 (M retest interval ∼ 26 days) and internal consistencies for scores from the two assessment sessions.
As a measure constructed to specifically index variations in dispositional threat sensitivity, the TF-20 displayed expected patterns of convergent and discriminant associations with psychopathology-related criterion variables assessed via parent-report as well as child self-report. The TF-20 showed a particularly strong positive relationship with the FSSC-R, a child-adapted version of an inventory that assesses fear in relation to a range of specific situations and stimuli, the Fear Survey Schedule (FSS). This strong association was expected, given that the adult version of the FSS served as an indicator of the general factor in Kramer et al.’s (Reference Kramer, Patrick, Krueger and Gasperi2012) fear/fearlessness model, which the TF-20 was designed to index. The TF-20 also showed moderately high correlations with the BIS, a self-report measure of sensitivity to aversive cues, and the SCARED, a youth-report measure of phobic fear and anxiety symptoms. Importantly, and in line with prior research (Kramer et al., Reference Kramer, Patrick, Hettema, Moore, Sawyers and Yancey2020; Nelson et al., Reference Nelson, Strickland, Krueger, Arbisi and Patrick2016), the TF-20’s observed correlations with these fear-symptom measures (M = .57) were more than twice those for child-report measures of dysphoric/distress symptomatology (i.e., SMFQ and ARI, M = .25).
The TF-20 also showed a robust association with anxiety sensitivity as measured by the CASI, a construct reflecting the tendency to fear bodily sensations that has been established as a risk factor for panic disorder (Silverman et al., Reference Silverman, Fleisig, Rabian and Peterson1991). Notably, the TF’s association with the Panic symptom scale of the SCARED was lower, and weaker than its correlation with all other subscales of the SCARED, including Generalized Anxiety. A potential explanation for the TF’s comparatively modest association with the SCARED Panic scale, given our sample’s very low rate of diagnosable panic disorder (see Table 3), is that general threat sensitivity relates more to latent risk for this disorder at younger ages, prior to when panic symptoms emerge. Regarding the TF’s stronger than expected association with the SCARED’s Generalized Anxiety scale, anxiety of this type in younger aged individuals tends to covary as much or more with fear symptomatology as depressive symptomatology (Lee et al., Reference Lee, Aggen, Carney, Hahn, Moroney, Machlin, Brotman, Towbin, Leibenluft, Pine, Roberson-Nay and Hettema2017; Waszczuk et al., Reference Waszczuk, Zavos, Gregory and Eley2014), whereas in adults it connects more to dysphoric-distress disorders than phobic fear conditions (e.g., Krueger, Reference Krueger1999; Watson, Reference Watson2005). The difference may lie in the fact that in younger aged individuals, the cardinal feature of generalized anxiety, worry, entails concerns about specific types of things—in particular, concerns about social standing and performance—rather than broad, non-circumscribed concerns. This more context-related form of worry is in fact reflected in a number of items comprising the Generalized Anxiety subscale of the SCARED (e.g., “I worry about other people liking me.”; “I worry about being as good as other kids.”; “I worry about things working out for me.”; “I worry about how well I do things.”)
The correlations of the TF-20 with parent-report measures of phobic fear (SCARED total and subscales, BIQ) and dysphoric/distress symptoms (SMFQ, ARI, CBCL Withdrawn/ Depressed scale) were uniformly lower than those for counterpart child-report measures, as expected given report-modality mismatch. However, paralleling findings for the child-report measures, rs with dysphoric symptoms were less than half those with fear symptoms. Further indicative of associational specificity (discriminant validity), the TF-20’s correlations with scale measures of externalizing symptomatology (attentional, aggressive, and rule-breaking), available only in parent-report (CBCL) form, were nonsignificantly negative.
Turning to child-report measures of personality, TF-20 scores showed a positive association with broad negative affectivity as assessed by the JEPQ Neuroticism scale, and opposing negative relations with the Extraversion scale of the JEPQ, encompassing traits of social gregariousness, activity, and excitement seeking, and Carver and White’s (Reference Carver and White1994) BAS, which indexes impulsive reward seeking. The TF-20 also showed a modest negative correlation with the parent-rated ICU scale, which assesses observable propensities toward callousness and emotional insensitivity.
Additionally, the TF-20 evidenced convergent and discriminant validity in relation to psychiatric disorder diagnoses. It related significantly to the three fear disorders that occurred with sufficient prevalence to permit analysis (specific phobia, social phobia, separation anxiety) as well as to the fear disorder composite. Of note, the TF’s correlations with fear disorder diagnoses were lower than with self-reported fear symptoms, likely due to modality mismatch (i.e., parent interview vs. child questionnaire) and to expectably lower rates of diagnosable expression of risk for psychopathology at younger versus older ages. The TF-20 also correlated significantly with generalized anxiety which, as noted earlier, aligns more with the fear disorder spectrum than the dysphoric disorder spectrum in younger-aged samples. Predictive specificity of the TF-20 in relation to fear disorder diagnoses was evidenced by a lack of significant associations with major depression, post-traumatic stress disorder, or the dysphoric disorder composite, and negligible relations with externalizing disorder diagnoses (oppositional defiant, attention deficit hyperactivity, externalizing composite).
Univariate twin modeling revealed that additive genetic influences accounted for a substantial portion (43%) of the variance in TF-20 scores in the current youth sample, with the remainder (57%) attributable to environmental factors not shared between twins. These estimates are quite similar to those reported by Kramer et al. (Reference Kramer, Patrick, Krueger and Gasperi2012) for the general factor of their fear/fearlessness model in a much larger adult twin sample (N = 2,511). Of note, measures of various normal-range personality traits also show similar proportions of genetic versus environmental determinants in younger- versus older-aged individuals. For example, the estimated heritability of Eysenck’s broad trait of neuroticism is 0.4–0.5 (van den Berg et al., Reference van den, Séphanie, de Moor, McGue, Pettersson, Terracciano, Verweij, Amin, Derringer, Esko, van Grootheest, Hansell, Huffman, Konte, Lahti, Luciano, Matteson, Viktorin, Wouda and Boomsma2014; Vukasović and Bratko, Reference Vukasović and Bratko2015). Also notable, similar to normative personality traits, common (shared family) environment did not contribute significantly to individual differences in dispositional fear/fearlessness in youth. By contrast, common environment has been found to account for 10–20% of variance in many forms of childhood psychopathology (Burt, Reference Burt2009).
A multivariate correlated-factors twin model was used to clarify the basis of the TF-20’s observed (phenotypic) associations with other fear-related constructs assessed via child-report. Genetic correlations of .54 to .71 were evident between the TF-20 and the three other fear-specific measures (FSSC-R, SCARED, BIS), indicating a moderate contribution of additive genetic influences to the observed covariance among these different fear measures. The remainder of their covariance was attributable to overlap in environmental influences among them.
In addition, we used a correlated common pathway twin model to further clarify the role of genetic versus environmental influences in the TF-20’s observed relations with latent fear and dysphoria factors of internalizing psychopathology defined using pertinent child-report measures of each. Consistent with hypothesis, we found substantial overlap between genetic influences accounting for variance in TF-20 scores and those accounting for variance specific to the latent fear factor, as evidenced by a strong genetic association between the two. By contrast, there was negligible overlap between the variance in TF-20 scores attributable to genes and genetic influences contributing specifically to the latent dysphoria factor. Of note, the variance in TF-20 scores attributable to environmental influences showed very similar associations with the environment-related variance in latent fear and latent dysphoria. Taken together, these findings suggest that, whereas most of the phenotypic relationship between TF-20 scores and dysphoric disorders reflects overlapping environmental influences, genetic influences unique to the construct assessed by the TF-20 intersect selectively with heritable risk for fear-related disorders.
Implications, limitations, and future directions
Considered in the light of previous research on the threat sensitivity dimension indexed by the TF-20, current study results have important implications for conceptualizing and assessing dispositional risk for fear-related pathologies. Corroborating results from prior work with adults and children (Nelson et al., Reference Nelson, Strickland, Krueger, Arbisi and Patrick2016; Palumbo et al., Reference Palumbo, Patrick and Latzman2021), current study analyses revealed preferential relations for dispositional threat sensitivity with phobic fear disorders relative to dysphoric distress disorders. Extending twin-modeling findings for adults (Kramer et al., Reference Kramer, Patrick, Krueger and Gasperi2012), our study also demonstrated moderate-level heritability for scores on this trait dimension in children aged 9-14. Additionally, the current study demonstrated specificity in the contribution of common genetic influences to threat sensitivity’s relationship with phobic fear symptoms: whereas TF-20 scores showed a moderately high genetic correlation with a latent fear-symptom factor, its genetic correlation with a counterpart dysphoric-symptom factor was negligible.
These findings indicate that the TF-20 assesses an underlying dispositional risk factor for fear-related pathologies. What is the nature of this underlying dispositional risk factor? In line with prior published work (Patrick et al., Reference Patrick, Durbin and Moser2012), we posit that the TF-20 – and other scale measures designed to harmonize with it (for a review, see Patrick, Reference Patrick2022) – index biobehavioral threat sensitivity in the modality of self-report. As a biobehavioral construct, threat sensitivity can also be indexed using physiological response (e.g., Kramer et al., Reference Kramer, Patrick, Krueger and Gasperi2012; Yancey et al., Reference Yancey, Venables and Patrick2016) or task behavioral measures (e.g., Yancey et al., Reference Yancey, Bowyer, Foell, Boot and Patrick2019). Indeed, Venables et al. (Reference Venables, Hicks, Yancey, Kramer, Nelson, Strickland, Krueger, Iacono and Patrick2017) found that when dispositional threat sensitivity was operationalized using aversive-reactivity measures together with scale-assessed trait fear, it evidenced an even stronger genetic correlation (.80) with phobic fear symptomatology than we found for the TF-20 alone in the current study (.68). The implication is that assessments of genetic risk for fear-related pathologies could be optimized by combining indicators of threat sensitivity from different response modalities. Additionally, measures of threat sensitivity from non-report modalities could be used to assess fear-disorder liability in nonverbal individuals (e.g., very young children).
Some limitations of the current study warrant mention. One is that the data were cross-sectional in nature. Although specificity was evident in threat sensitivity’s phenotypic and genotypic associations with phobic fear relative to dysphoric distress symptomatology, longitudinal research demonstrating prospective prediction of phobic fear disorders from earlier-assessed threat sensitivity is needed to firmly establish it as a dispositional liability factor. Further research is also needed to examine threat sensitivity’s association with GAD symptoms at younger versus older ages, in order to clarify whether this disorder aligns more with the phobic fear or the dysphoric distress spectrum. Another study limitation is that the youthful age of participants limited the opportunity for dispositional risk to be expressed in the form of psychopathology symptoms. In this regard, we posit that TF-20 scores related more to CASI-assessed panic disorder risk in our study than to SCARED-assessed panic symptoms because most participants were below the age of panic disorder expression, and that their youthful age may also have constrained the TF-20’s relations with full fear-disorder diagnoses. Longitudinal research can provide a means to test these hypotheses. A further limitation is that the current study included only report-based measures (i.e., child questionnaires, parent ratings, and interview-based assessments). Given the potential value of physiological and behavioral response measures for assessing risk in individuals of various ages, further research using measures of these types is needed. One potential avenue for work of this kind is to utilize data from already existing multi-modal assessment studies in which effective proxy measures of trait fear have been developed (e.g., Bertoldi et al., Reference Bertoldi, Evans, Oskarsson, Joyner, Tuvblad, Baker, Raine, Schwartz and Patrick2022; Brislin & Patrick, Reference Brislin and Patrick2019; Palumbo et al., Reference Palumbo, Patrick and Latzman2021).
Notwithstanding these limitations, the current study importantly extends previous research on dispositional risk for internalizing psychopathology. It adds to growing evidence for a dispositional construct of threat sensitivity, quantifiable through measures in different response modalities, as a specific liability for disorders marked by excessive, irrational fear. Our findings encourage further research utilizing longitudinal collection of multi-modal data from general community, twin, and clinical samples to further understand the nature of this dispositional liability and experiential factors that moderate outcomes associated with it.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0954579425000380.
Acknowledgments
We are grateful to Keanan J. Joyner of the University of California at Berkeley for his assistance with modeling analyses.
Funding statement
The work reported in this article was supported by grants from the National Institute of Mental Health (C.S.: R01MH098055, T32MH020030; JMH and RRN: R01MH098055; DSP: NIMH-IRP-ziamh002781) and the US Army Research Institute for the Behavioral and Social Science (C.P.: W911NF-14-1-0018 & W81XWH-10-2-0181). The content of this paper is solely the responsibility of the authors and does not necessarily represent the official views of the U.S. Government, Department of Defense, Department of the Army, Department of Veterans Affairs, or U.S. Recruiting Command. Funding sources had no role in the study design in the collection, analysis and interpretation of data, in the writing of the report, or in the decision to submit the article for publication.
Competing interests
None.
Open access
