Preventing Onset of Anxiety Disorders in Offspring of Anxious Parents: A Randomized Controlled Trial of a Family-Based Intervention

A randomized controlled trial was conducted with families in which at least one parent met criteria for a current anxiety disorder according to DSM-IV-TR criteria and at least one child did not have an anxiety disorder. In each family, only one parent and one child were enrolled in the study, although all family members were invited to participate in the intervention. The study was funded by the National Institute of Mental Health and was conducted in an outpatient research clinic at the Johns Hopkins University School of Medicine. The trial was approved by the Johns Hopkins University institutional review board; parents provided written informed consent and children provided assent after receiving a complete description of the study.

Participants were 136 volunteer families. Children were between 6 and 13 years old, did not meet criteria for an anxiety disorder, were not currently receiving treatment for anxiety, and did not have any medical or psychiatric conditions contraindicating the study interventions (e.g., suicidality), based on clinical interview. At least one biological parent in each family met criteria for a current primary diagnosis of an anxiety disorder (Table 1), and none had any medical or comorbid psychiatric conditions contraindicating study participation (e.g., substance use disorder), based on clinical interview. There were no restrictions regarding family composition. Comorbid psychiatric disorders were allowed but had to have lower severity or impairment levels than the anxiety disorder. The presence of psychiatric disorders and associated severity or impairment were determined by the parent and child versions of the Anxiety Disorders Interview Schedule for DSM-IV (14, 15).

Parental and child psychiatric disorders were assessed by trained independent evaluators. Onset of a child anxiety disorder was the primary outcome measure. The Anxiety Disorders Interview Schedule is the gold-standard assessment tool for determining anxiety disorders and is well validated (16, 17). The instrument yields both a diagnosis (present/absent) and a clinical severity rating that ranges from 0 to 8; scores of 4 or higher are indicative of a clinical disorder. Clinical severity ratings represent the degree of severity and impairment or interference in functioning. The independent evaluators were supervised by a senior child psychiatrist (M.A.R.) who reviewed all diagnoses and clinical severity ratings and remained blind to intervention condition throughout the study. The sum of the child Anxiety Disorders Interview Schedule clinical severity ratings across all anxiety disorders, as determined by the evaluator after interviewing the parent and child separately, was used as a measure of anxiety symptom severity/impairment. Interrater agreement on a randomly selected 25% of Anxiety Disorders Interview Schedule administrations was 97% for parents and 97% for children. Diagnostic agreement was defined as matching on the presence/absence of a disorder and a clinical severity rating within 1 point.

Demographic information, obtained from the participating anxious parent, included parent age, gender, race/ethnicity, education, income, and marital status, as well as child age and gender. Parental anxiety, examined as a moderator, was assessed using the trait version of the State-Trait Anxiety Inventory (baseline internal consistency, 0.92) (18). The overall severity levels of the parents’ psychopathology and anxious modeling (examined as potential mediators) were assessed with the global severity index of the Brief Symptom Inventory (baseline internal consistency, 0.95) (19) and the anxious modeling subscale of the Learning History Questionnaire–Revised (baseline internal consistency, 0.83) (20), respectively. Child maladaptive cognitions, examined as a mediator, were assessed using the social subscale of the Children’s Negative Cognitive Error Questionnaire (baseline internal consistency, 0.80) (21). Finally, mental health service utilization for anxiety (dichotomized as yes/no) was obtained from the anxious parent by monthly telephone check-ins and/or in-person interviews after randomization over the course of the study.

The Coping and Promoting Strength intervention consisted of eight weekly 60-minute sessions and three optional monthly booster sessions. (The Coping and Promoting Strength program manual and the information-monitoring condition handout are available from the first author on request.) Each family met individually with a trained therapist. The intervention (13) targets theory-driven modifiable child and parent risk factors. Child factors included elevated anxiety symptoms, social avoidance/withdrawal, maladaptive cognitions, and deficits in coping and problem-solving skills; parent factors included anxiety-enhancing parenting behaviors (e.g., modeling of anxiety, overcontrol/overprotection). Families were taught how to identify the signs of anxiety and how to reduce anxiety (psychoeducation), cognitive restructuring, in vivo desensitization, problem solving, and parenting strategies (e.g., contingency management and increasing child independence and autonomy). The first two sessions were with parents alone; the remaining sessions included all interested family members.

Participants in the information-monitoring condition were given a 36-page pamphlet containing information about anxiety disorders and associated treatments (22). The pamphlet did not include detailed information about the anxiety-reduction strategies included in the Coping and Promoting Strength program. The information-monitoring condition was selected as a control to mimic usual care; high-risk offspring of anxious parents are unlikely to receive any preventive intervention as part of usual care other than general prevention approaches such as an educational pamphlet.

Participant families were recruited through advertisements in local newspapers, mailings to local physicians and psychiatrists, community flyers, and radio advertisements. Interested families called study staff and completed a telephone screening in which key inclusion criteria such as psychiatric disorder and age were assessed. Families who met initial inclusion criteria were invited for an in-person baseline assessment. All participating families were expected to complete assessments, administered by interviewers blind to intervention condition and reviewed by a senior child psychiatrist, at end of the intervention (or 8 weeks after randomization) and at follow-ups 6 and 12 months after the 8 weeks. (The flow of participant families through the study is presented in a CONSORT diagram in the data supplement that accompanies the online edition of this article.) Eligible families were randomly assigned in a 1:1 ratio to an intervention condition (using random numbers derived from randomization.com). An independent evaluator contacted families monthly by telephone to assess anxiety symptoms and mental health service utilization.

Initial analyses examined baseline equivalence of the intervention and control groups, using chi-square tests for categorical variables and t tests for continuous variables. Attrition analyses (23) were conducted to examine whether attrition rates differed across conditions. Outlier analyses were conducted to determine whether outliers substantially biased the results (24). Intention-to-treat analyses were performed in Mplus, version 7.11 (25), using full-information maximum-likelihood estimation to handle missing data. Intervention effects were examined using three analytical approaches. Baseline child anxiety symptom scores were included as a covariate in these analyses. We first compared the intervention and control groups at the postintervention assessment and each of the follow-up assessments, using analysis of covariance (ANCOVA) for continuous outcomes and logistic regression for binary outcomes. The second approach used survival analysis to examine the intervention effect on the incidence of anxiety disorders over 1 year. The third approach compared the trajectory of anxiety symptoms over time for the two groups, using multigroup piecewise growth curve modeling to examine whether the program effect after the intervention persisted, faded, or increased over time. Alpha was set to 0.05 (two-tailed). Several moderators of intervention effects were examined, including child age, gender, and baseline anxiety symptoms; parent gender; parent-child gender match; parent baseline anxiety severity; and psychiatric comorbidities. One moderator was examined at a time by adding group-by-moderator interaction into the model. If none of the moderation effects were significant, we reassessed the model without the interaction. Theory-driven parent and child mediators of intervention effects on anxiety symptoms and diagnosis at 12-month follow-up were also examined, including reduction of parental anxiety and global psychopathology (measured by the Brief Symptom Inventory), parental anxious modeling, and child cognitive errors at the postintervention and 6-month follow-up assessments (26).

Table 1 summarizes participants’ baseline clinical and demographic characteristics. The two groups did not differ significantly on child demographic variables or on levels of child or parent anxiety. Mean parental age was higher in the information-monitoring group. There were also group differences in the attrition rate, with significantly more children in the intervention condition failing to complete the 1-year assessment within the allotted window (p=0.03; see the CONSORT diagram in the online data supplement). Attrition analysis comparing families that remained in the study and families that dropped out indicated that parents who dropped out were significantly younger on average (mean age, 41.3 years compared with 37.2 years; p=0.002). Attrition analyses for group-by-attrition interaction were not performed because of the small attrition rate (N=4) in the information-monitoring group. One outlier, in the information-monitoring group, was identified on the clinical severity rating of the Anxiety Disorders Interview Schedule at the 6-month follow-up (outlier score, 35, compared with scores ≤18 for all others). For analyses involved with this variable, the results were consistent between the models with and without the outlier. We present the more conservative results without the outlier in the model.

Among the families in the Coping and Promoting Strength intervention group, the average number of sessions attended was 6.63 out of eight (range, 0–8) and 1.2 out of three booster sessions (range, 0–3); the total mean number of sessions attended was 9.01 out of 11 (SD=2.63, range, 0–11). The participating anxious parent attended all sessions and was accompanied by the other parent in a mean of 4.07 of the eight sessions (range, 0–8). Intervention adherence was assessed on 25% of each family’s recorded sessions by independent evaluators. Each evaluator rated adherence to specific session objectives (e.g., explaining the intervention model of anxiety reduction, teaching relaxation or cognitive restructuring techniques). The average adherence ratings per family ranged from 86.36% to 100%, and the mean adherence rating across all sessions was 97.58% (SD= 3.51), reflecting high levels of clinician adherence. Total number of sessions attended was unrelated to child outcomes.

At the postintervention and follow-up assessments, the independent evaluators were asked to guess group assignment. Kappa values for agreement between actual assignment and guessed assignment were 0.32, −0.08, and 0.03 at the three assessments, respectively. The independent evaluators seemed to be more aware of the group condition at the postintervention assessment than at the 6-month and 12-month follow-up assessments.

Table 2 lists the numbers of children in each group who developed an anxiety disorder over the course of the 1-year study. Cumulatively, three children in the intervention group (5.26%) and 19 children in the control group (30.65%) developed an anxiety disorder (odds ratio=8.54, 95% CI=2.27, 32.06; number needed to treat=3.9). The results of the survival analysis indicated that the rate of onset of anxiety disorders for children in the control group was 6.6 times higher than that for children in intervention group during the 1-year period (hazard ratio=6.60, 95% CI=2.00, 21.82; p=0.002). Figure 1 illustrates the hazard functions for the two groups.

Table 3 shows the results of ANCOVAs for comparisons between the intervention and control groups on severity of anxiety symptoms at each of the postrandomization assessments. On average, youths who received the Coping and Promoting Strength intervention had significantly lower symptom scores at the postintervention assessment, as well as the 6- and 12-month follow-up assessments, compared with the control group.

Figure 2 illustrates the piecewise growth models for the intervention and control groups. Results of the multigroup comparisons of the growth factors on anxiety symptoms (i.e., the intercept, changes at the postintervention assessment, and linear growth after the postintervention assessment) showed that although both groups improved from baseline to end of intervention, the scores were reduced significantly more in the intervention group (B=−5.23, SE=0.82, p<0.01) compared with the control group (B=−2.21, SE=0.63, p<0.01) (difference=−3.02, 95% CI=−5.04, −1.00]). From the postintervention assessment to the 12-month follow-up, additional significant reductions occurred for the intervention group (B=−0.43, SE=0.13, p<0.01) but not for the control group.

The intervention effects were moderated by baseline child anxiety symptoms but not by any of the other moderators examined (child age and gender; parent gender, anxiety, comorbidity; parent-child gender match). Probing of the intervention effect at one standard deviation above and below the mean of baseline child anxiety symptoms (27) indicated that the benefit of the intervention was stronger for children with higher baseline anxiety symptoms than for those with lower baseline anxiety symptoms (see Table 3).

Mediation analyses revealed that significant reductions in both parental modeling of anxiety and parental global distress (but not anxiety alone) at the postintervention and 6-month follow-up assessments significantly mediated the intervention effects on outcomes for severity of child anxiety symptoms at the 1-year follow-up (mediation effect—where ab refers to the product of the unstandardized betas, a=regression path for intervention → mediator, and b=regression path for mediator → anxiety—through postintervention assessment anxiety modeling: ab=0.392, 95% CI=0.113, 0.642; through 6-month assessment anxiety modeling: ab=0.423, 95% CI=0.106, 0.725; through postintervention assessment parental distress modeling: ab=0.233, 95% CI=0.014, 0.488; through 6-month assessment parental distress modeling: ab=0.337, 90% CI=0.018, 0.713). Changes in child maladaptive cognitions did not mediate the intervention effects on outcomes.

Parent reports of the use of mental health services for anxiety (yes/no) indicated that 21.5% of the control families, compared with 13.2% of the intervention families, reported receiving mental health services for their child’s anxiety over the course of the study (not a statistically significant difference).

The study had several limitations. Our sample was made up of volunteers from predominantly Caucasian, two-parent families with relatively high incomes, and the findings may not generalize to nonvolunteers or to a more racially or economically diverse population. It is also possible that the parents in this sample were less severely ill than many parents with anxiety disorders, as they had few comorbid diagnoses and had greater resources to support their engagement in prevention. A majority of the parents had a primary diagnosis of generalized anxiety disorder, which may reflect the excessive worry about their child’s well-being that is characteristic of this disorder (in contrast to social anxiety and avoidance, which may have reduced the likelihood of these parents enrolling). The children in this study may also not be representative of the population of high-risk offspring both because youths were excluded if they already had an anxiety disorder or any other psychiatric condition that required immediate treatment.

Another limitation is that the information-monitoring condition did not control for therapist time and attention. Thus, whether this intervention is superior to another active prevention program is unknown. Moreover, we did not conduct an evaluation of whether specific intervention components (e.g., relaxation, problem solving) were uniquely responsible for the beneficial outcomes. The completion rate for study evaluations was higher in the control group than in the intervention group, a finding that may be attributed to the offer of receiving the intervention after the 1-year assessment. We did not examine an exhaustive list of potential moderators and mediators, the sample size for specific moderation effects was small, and analyses were likely to be underpowered to detect these effects. Finally, the length of the follow-up was limited to 12 months; a longer follow-up is needed to further assess mental health outcomes and the balance of costs and benefits.