Early Risk Factors and Developmental Pathways to Chronic High Inhibition and Social Anxiety Disorder in Adolescence

The data were drawn from a larger longitudinal study, the Wisconsin Study of Families and Work (30). A sample of 570 pregnant women over age 18 who were living with the baby's biological father were recruited from prenatal clinics in two cities during the second trimester of pregnancy. Follow-up assessments were conducted when the children were in infancy (at 1, 4, and 12 months), at preschool age (ages 3.5 and 4.5 years), and at grades 1, 3, 5, 7 and 9; teachers participated beginning at grade 1. At grade 1, 468 (82%) of the families remained in the study. However, because of budgetary constraints, the full grade 1 assessment (which required home visits) was conducted only on one-half of the first cohort (N=100) and the complete second cohort (N=197) of families living within a 4-hour driving distance of the project offices. Of these 297 families and teachers who participated in the full grade 1 assessment, the present study included the 238 for whom complete multi-informant inhibition data through grade 9 were available (missing data were primarily due to home schooling and school or teacher refusal). The study was approved by the Institutional Review Board of the University of Wisconsin-Madison. Informed consent was obtained from adults at each assessment; beginning at grade 5, children provided informed assent.

The sample of 238 children consisted of 111 boys and 127 girls, 11% of them ethnic minorities. At recruitment (1990–1991), the mothers ranged from 20 to 41 years of age (median=30 years); 96% were married. One percent of the mothers had less than a high school degree, 40% were high school graduates, 38% were college graduates, and 21% had education beyond college; the fathers' educational level had a similar distribution. Annual family income ranged from $7,500 to over $200,000 (median=$49,000, comparable to the average family income in Wisconsin) (31). There were no demographic differences between the 238 participating families and the remainder of the original 570 families, with two exceptions: participating mothers were 1 year older on average than nonparticipating mothers (mean=29.97 years [SD=4.20] compared with mean=28.94 years [SD=4.43]; t=–2.80, df=568, p=0.005), and participating families had a higher annual income on average than nonparticipating families (mean=$52,232 [SD=$22,797] compared with mean=$48,106 [SD=$23,853]; F=4.27, df=1, 564, p=0.039).

Maternal stress was assessed in the infancy and preschool periods with maternal reports of five domains. 1) Maternal depression was assessed by the Center for Epidemiologic Studies Depression Scale (CES-D) (32); alpha coefficients exceeded 0.85 for all assessments. 2) Family expressed anger was assessed with the average of three items tapping marital conflict and scores for anger expression and the negative subscale of the Family Expressiveness Questionnaire (33–35); all alpha values were >0.70. 3) Parenting stress was calculated from scores on the competence and child-reinforces-parent subscales of the Parenting Stress Index and the average of three negative items (e.g., "I often feel angry with my child") from the Child-Rearing Practices Report (34, 36, 37); all alpha values were >0.65. 4) Role overload was calculated from the role restriction subscale of the Parenting Stress Index, and the average of five items from the role overload scale (e.g., feeling "pulled apart by conflicting obligations") (34, 36, 37); all alpha values were >0.75. 5) Financial stress was calculated as the average of four items (e.g., "difficulty making monthly payments"); all alpha values were >0.70. Separately for the infancy and preschool periods, composites were constructed using principal components analysis, where the first component accounted for at least 50% of the variance and all factor loadings exceeded 0.50. The two resulting composites were highly correlated (r=0.73) and thus were averaged.

Children's behavioral inhibition at age 4.5 years was assessed from videotaped observations of a 2-hour home visit conducted by two research assistants. The visit included entrance to the home, saliva collection, cognitive testing, and a series of emotion-eliciting tasks (38), all used to measure the child's reactions to unfamiliar people, objects, and test procedures. After the visit, the research assistants independently reviewed the videotape once for the entire assessment and rated the children on four 5-point items assessing global shyness, initiative with tasks, exploration of unfamiliar objects, and social engagement with the research assistants (all kappa values ≥0.80). Each item was averaged across the two research assistants, and the four averaged items were combined into a single composite score using principal components analysis (65% of the variance was accounted for; all factor loadings >0.71).

Children's salivary cortisol levels at age 4.5 years were assessed with saliva samples (passive drool) collected by parents on three consecutive afternoons. Eighty-six percent of the samples were collected in the target period between 3:00 p.m. and 7:00 p.m., and the remainder were collected earlier in the afternoon. Cortisol was assessed with the Pantex RIA (Santa Monica, Calif.) modified for saliva. The detection limit of the assay (ED₈₀) was 0.03 mg/dl. The mean interassay variation was 7.4%, and the mean intra-assay variation was 3.8%. Based on previous analyses (21, 28) showing that only antibiotics and cold or allergy medications affected cortisol values in this sample, medication use was coded 1 if the child was currently on any of these medications and 0 otherwise. To control for time of collection and medication use, the log-transformed cortisol values were residualized for hour of collection and medications.

Children's school-age inhibition was assessed from interviews with mothers, teachers, and children during the spring of grades 1, 3, 5, 7, and 9. Adults completed the MacArthur Health and Behavior Questionnaire (39) social inhibition scale (five items, such as "shy asking peers to play," "shy with unfamiliar adults"); all alpha values were >0.74. At grade 1, children were asked a parallel set of items from the Berkeley Puppet Interview social scales (40) (α=0.67). At subsequent grades, the Berkeley Puppet Interview was modified to an age-appropriate questionnaire format; all alpha values were >0.71. Multi-informant scores were computed at each assessment using a principal components analysis-based approach (41) in which the first component measures the core characteristic of interest (here, inhibition), defined as what the three reports shared in common (i.e., all positive loadings >0.50), and the second and third components distinguish the sources of error that result from reporters' differing perspectives (i.e., child self-view versus adult view) and contexts (i.e., home versus school). To construct a longitudinal inhibition score reflecting both level and chronicity, the first component scores of the principal components analysis were used to categorize children at each assessment as being in the lower 25% (score=1), the middle 50% (score=2), or the upper 25% (score=3) of inhibition; these scores were summed over the five assessments, resulting in a longitudinal score ranging from 5 (in the lower 25% at all times) to 15 (in the upper 25% at all times).

Lifetime DSM-IV diagnoses were obtained with the Schedule for Affective Disorders and Schizophrenia for School-Age Children–Present and Lifetime Version (K-SADS-PL) (42) for the subset of the children who participated in an intensive laboratory assessment at grade 9 (N=60; 25%). Recruitment was based on living in close proximity to the laboratory and reporting no exclusionary criteria for functional MRI (e.g., metal orthodontics); there was no significant difference on the longitudinal inhibition score between these children and the remaining 178 children (mean=10.30 [SD=2.95] compared with mean=9.87 [SD=2.63]). A board-certified child psychiatrist (M.J.S.) with expertise in the K-SADS-PL trained and supervised a child psychiatrist and a clinical child psychologist in its administration. Children were interviewed first and mothers second, with additional interviewing conducted when their responses related to diagnostic criteria disagreed. All K-SADS-PL interviews were reviewed by M.J.S. and the administering clinician, and disagreements were resolved by consensus.

Data analyses addressed two questions. First, Spearman correlations identified the early risk factors for chronic high inhibition, and a hierarchical multiple regression analysis defined the longitudinal linkages among the risk factors. Child gender and exposure to early maternal stress were included in the first step of the regression analysis. Next, to identify possible mediators and moderators, childhood behavioral inhibition and afternoon cortisol levels at age 4.5 years were entered, together with the four pairwise interactions of child gender and exposure to maternal stress with each of the two 4.5-year variables; no significant interactions were found. A dichotomous variable was also included to assess any effect of psychotropic medication use at grade 9 (N=14; 6%); no significant effect was found. Second, among the subset of children with lifetime psychiatric data, a chi-square analysis ascertained differences in DSM-IV diagnoses according to level and chronicity of school-age inhibition; t tests examined differences in risk factors and school-age inhibition for children with and without a lifetime history of social anxiety disorder.

The distribution of the longitudinal inhibition score is shown in Figure 1 (mean=9.98, SD=2.71). For the chi-square analysis only, these scores were collapsed to form five groups, where chronic inhibition was defined as evidencing the same level (low, middle, high) for at least four of the five assessments and never evidencing a nonadjacent level. By this definition, 55% of the children evidenced chronic inhibition, including all children in the chronic high (scores of 15 or 14) and chronic low (scores of 5 or 6) groups and three-fourths of those in the middle (scores of 9, 10, or 11) group. None of the children in the low-middle (scores of 7 or 8) or middle-high (scores of 12 or 13) groups evidenced chronicity.

Identification of Early Risk Factors and Pathways to Chronic High Inhibition

All four early risk factors were significantly correlated with the longitudinal inhibition score (Table 1; see Table S1 in the data supplement that accompanies the online edition of this article for correlations with inhibition at each grade). As hypothesized, chronic high inhibition was associated with being female, exposure to higher levels of early maternal stress, and, at age 4.5 years, higher levels of behavioral inhibition and afternoon cortisol. Each pair of risk factors within a time period (i.e., child gender and exposure to maternal stress beginning at birth, and behavioral inhibition and cortisol levels at age 4.5 years) were uncorrelated; however, there were significant correlations between the earlier risk factors and those at age 4.5 years, possibly reflecting mediational processes.

The hierarchical multiple regression analysis identified two mediational chains (Figure 2). The first chain began with the association of child gender with longitudinal inhibition (β=1.34, p<0.001), which at a trend level (Sobel's test, p=0.069) was partially mediated by early behavioral inhibition and cortisol levels at age 4.5 years (the effect was reduced but remained significant; β=1.13, p=0.001). The second chain began with exposure to early maternal stress (β=0.64, p=0.001), which was partially mediated (β=0.61, p=0.001; Sobel's test, p=0.048) by children's cortisol levels at age 4.5 years.

Among the 60 children for whom data were available on lifetime DSM-IV diagnoses, significant differences were observed in diagnostic status among the five longitudinal inhibition chronicity groups (χ²=31.36, df=16, 60, p=0.01) (Table 2). Whereas 80% of the children in the chronic high group received at least one lifetime psychiatric diagnosis, all in the internalizing (anxiety or mood) domain, the percentage decreased across the groups to only 20% in the chronic low group, all in the externalizing domain. Notably, 50% of the children in the chronic high group received a lifetime diagnosis of social anxiety disorder, compared with 27% in the middle-high group, 5% in the middle group, and none in the low-middle and chronic low groups. Among the three higher chronicity groups, there were few differences in other anxiety disorders (predominantly specific phobias) or mood disorders.

Because of the overlapping time intervals over which inhibition (and risk factor) ratings and lifetime diagnoses of social anxiety disorder were obtained, additional analyses were conducted to substantiate the prospective nature of this association. Comparisons of adolescents with and without a lifetime diagnosis of social anxiety disorder showed that there were no significant group differences for the early risk factors or grade 1 inhibition, but adolescents diagnosed with social anxiety disorder had significantly higher inhibition scores beginning at grade 3 (Table 3). This pattern corresponds closely with the interview descriptions of the adolescents diagnosed with social anxiety disorder as being "always" shy since very early childhood, with the defining clinical features of significant distress and impairment manifesting later, in elementary or middle school, when the demands of social relationships, including performance situations, increased.