The Physiological Regulation of Emotion During Social Interactions: Vagal Flexibility Moderates the Effects of a Military Parenting Intervention on Father Involvement in A Randomized Trial

Abstract

To make prevention programs more effective and understand “what works for whom”, evidence regarding what individual characteristics predict intervention responsiveness is needed. Previous studies have evaluated a military parent training program known as After Deployment Adaptive Parenting Tools/ADAPT, yet less is understood about the program’s varying effects for fathers. We tested the physiological regulation of emotion during social interactions as a moderator predicting fathers’ responsiveness in a randomized trial of ADAPT, in which emotion regulation was operationally measured through vagal flexibility (VF; dynamic changes in cardiac vagal tone). Families with a child aged between 4–13 years for whom physiological data were gathered (n = 145) were randomly assigned to ADAPT (14-week face-to-face group intervention) or a control group (services as usual). Fathers in these families were National Guard/Reserve members who had been deployed to war in Iraq and/or Afghanistan and recently returned. Prior to the intervention, cardiac data was collected in-home throughout a set of family interaction tasks and VF was operationalized as the changes in high frequency (HF) power of heart rate variability (HRV) from a reading task to a problem-solving task. Parenting behaviors were observed and coded based on theory-driven indicators pre-intervention and at 1-year follow-up. Results of structural equation modeling showed that VF significantly moderated fathers’ intervention responsiveness, such that fathers with higher vs. lower VF exhibited more effective parenting at 1-year follow-up if they were randomized into ADAPT vs. the control group. This study is the first to demonstrate that parasympathetic vagal functioning may be a biomarker to predict response to a military parenting intervention to enhance parenting in combat deployed fathers. The implications for precision-based prevention are discussed.

Since 9/11 2001, the wars known as Operations Enduring Freedom, Iraqi Freedom (OEF/OIF), and New Dawn have resulted in prolonged and repeated deployments of large numbers of U.S. military personnel –mostly men (84.1% Active Duty and 80.7% Guard and Reserve members) - to Afghanistan and/or Iraq (Department of Defense, 2016). Deployment places unique challenges and stressors on service members, their partners, and children. Military children are as resilient as civilian children, yet a growing body of literature has documented that school-aged children of deployed parents reported elevated rates of mental health problems in comparison with non-deployed military children or national norms (Card et al., 2011; Flake, Davis, Johnson, & Middleton, 2009; Gorman, Eide, & Hisle-Gorman, 2010; Lester et al., 2010; Millegan, Engel, Liu, & Dinneen, 2013; Wadsworth, Bailey, & Coppola, 2017).

Over the past few decades, a significant number of civilian parenting interventions have been developed, and a substantial body of evidence has demonstrated their effectiveness in strengthening parenting practices and thus reducing children’s mental health problems (Sandler, Schoenfelder, Wolchik, & MacKinnon, 2011). One such program is the Parent Management Training –Oregon Model (PMTO), based on Social Interaction Learning (SIL) theory (Patterson, 1982; Forgatch & Gewirtz, 2017). The model suggests that stressful contexts breed coercive parent-child interactions which, in turn, are associated with disruptive child behaviors, whereas positive effective parenting practices are associated with improvements in child behavior. Thus, parenting is a crucial mediator for the impact of family stressors on children’s wellbeing.

Given the large number of military families exposed to the stressor of wartime deployment, an adaptation of PMTO was developed for military families, known as After Deployment Adaptive Parenting Tools/ADAPT. In addition to PMTO skills, ADAPT provided training in parental emotion socialization, in recognition of the emotional challenges associated with deployment and its consequences (e.g., family members’ anxiety during the deployment). We previously documented the program’s efficacy in improving observed parenting behaviors and child adjustment (Gewirtz, DeGarmo, & Zamir, 2018). However, not all parents gain benefits from programs in the same way. Insufficient data exists regarding what participant characteristics may be associated with greater or lesser responsiveness to the ADAPT program.

To improve effect sizes of interventions, researchers need to identify pre-existing individual characteristics that can predict varying responsiveness for precision-based prevention (Gardner, Hutchings, Bywater, & Whitaker, 2010). It is important to investigate biomarkers as mechanisms that may underlie the behavioral outcomes of interest, as assessments that use observable behavior, surveys, and background characteristics have been unable to substantially improve effect sizes of intervention programs so far (Fishbein & Dariotis, 2017). Chesmore, Piehler, & Gewirtz (2018) found that posttraumatic stress disorder (PTSD) status significantly predicted whether fathers benefited from the ADAPT program. While this finding was informative, a PTSD diagnosis is based on self-reported behavioral symptoms and thus lacks specificity in terms of pathophysiology. Classifying individuals based on their diagnosis does not give us the precision that we would expect for personalized interventions, because individuals with the same diagnoses exhibit heterogeneous pathophysiology (Insel, 2014).

Research is needed to identify malleable variables to inform precision-based interventions (Biglan, 2004). In the context of parenting interventions which aim to reduce children’s mental health problems, evidence on malleable predictors for parents’ responsivity to the intervention can be used to inform practice by identifying those who may need a greater dose or additional interventions to increase their likelihood of benefit.

To extend previous studies, we sought to test a pre-intervention biomarker, the physiological regulation of emotion, or vagal flexibility (VF), as a moderator predicting changes in parenting among a sample of combat deployed fathers. Our focus on VF is in part due to the accumulated evidence supporting the relations between cardiac vagal tone and health outcomes among combat deployed male service members (Minassian et al., 2014; Pyne et al., 2016) and civilian samples (Balzarotti, Biassoni, Colombo, & Ciceri, 2017). In particular, our earlier work using cross-sectional data found that high VF was protective for deployed fathers with poor emotion regulation to maintain positive engagement behaviors during their interactions with children (Zhang, Hoch, Gewirtz, Barnes, & Snyder, 2020). Thus, VF may be an important biomarker of effective emotion regulation and social engagement in parent-child contexts, making it critical for parenting. Secondarily, derived cardiac measures such as VF offer the ability to collect data non-intrusively, in settings outside the laboratory, and at relatively low cost. For a physiological measure to “scale up” effectively for use in intervention tailoring, these are important characteristics. Finally, there is emerging evidence showing the malleability of this variable (Krygier et al., 2013; Tan, Farmer, Sutherland, & Gevirtz, 2011).

The vagus nerve is the major tract of the parasympathetic nervous system connecting to the heart and many other organs, and the concept of cardiac vagal tone has attracted increasing scholarly attention in the past two decades. Porges (2007) suggests that the mammalian vagus nerve serves as a “brake” to slow heart rate and helps to regulate biobehavioral responses under environmental challenges. Vagal control over heart rate can be measured by extracting the high frequency (HF) component of beat to beat cardiac data streams (heart rate variability/HRV). Baseline or resting HF-HRV reflects a general level of parasympathetic functioning. Studies have shown that low baseline HF-HRV is associated with mental health problems such as depression (Rottenberg, 2007), anxiety (Chalmers, Quintana, Abbott, & Kemp, 2014), as well as combat-related PTSD (Minassian et al., 2014). High baseline HF-HRV has typically been associated with adaptive parasympathetic functioning and health outcomes (Balzarotti et al., 2017), though there are mixed findings. For example, Butler et al. (2006) found that individuals with higher baseline HF-HRV experienced and expressed more negative emotion after viewing an upsetting film than those with lower baseline HF-HRV. One plausible explanation for these mixed findings is that high baseline HF-HRV may not necessarily index adaptive emotion regulation when emotion regulation is considered to be a dynamic process that often occurs in social contexts, which cannot be precisely captured by baseline or resting cardiac vagal tone.

From this perspective, VF may be more advantageous than baseline HF-HRV for capturing the dynamic aspect of emotion regulation. VF reflects phasic HF-HRV and reflects the capacity to release the vagal “brake,” accelerate heart rate, and increase metabolism, as ways to engage in a stressful environment without compromising control. VF is typically calculated as the change in HF- HRV from baseline to a stressful situation. VF provides information about emotion regulation under stressful situations because it connotes individuals’ capacities to dynamically adjust their physiological arousal levels in line with environmental demands. As Porges (2007) posited in polyvagal theory, rapid suppression of cardiac vagal tone under stressful situations facilitates emotional processes and allows the body to cope with challenges, which reflect engagement and approach behaviors. Empirical studies have supported the benefits of high VF. For example, studies found that VF predicted recovery from depression in a clinical sample (Rottenberg et al., 2005) as well as individual sensitivity to emotional information in social contexts among a community sample (Muhtadie, Koslov, Akinola, & Mendes, 2015).

Existing studies with civilian populations have linked parental VF to parenting, although many of those studies focused on mothers. For example, stressed mothers exhibited more negative intrusiveness with their children if their VF was lower (Mills-Koonce et al., 2009). Moreover, Connell, Dawson, Danzo, & Mckillop (2017) showed that high VF was associated with fewer expressions of parental anger during parent-child interactions. We know little about fathers’ parenting in relation to cardiac vagal tone or VF.

The current study

We tested VF as a moderator of the ADAPT program among combat deployed fathers. Given the relationships between VF and emotion regulation/social communication, we hypothesized that fathers with higher (vs. lower) VF would be better able to master parenting tools, and thus they would show greater improvements in program-targeted parenting practices at 1-year follow-up if they were randomized to the ADAPT program. To be consistent with previous studies (Chesmore et al., 2018; Gewirtz et al., 2018), we chose a theory-based latent construct of parenting, which is the key intervention target in the ADAPT program drawn from PMTO: skill encouragement, discipline, monitoring, problem solving, and positive involvement.

Method

Participants

The sample was 145 post-deployed military fathers, their partners/spouses, and a target child, who participated in a randomized controlled trial/RCT of the ADAPT program. This group constituted all the families from the full sample of 336 families (282 deployed fathers) in the RCT. Because HRV was not a key study variable in the larger RCT project, there were a relatively large proportion of participants whose HRV data were not available for analysis in the current study (see Online Supplementary Material Figure 1 for details). Specifically, 137 fathers were excluded because they did not have HRV data or their HRV data did not meet our criteria for analysis. Online Supplementary Material Figure 1 provides a CONSORT chart with more details. Independent-sample t-tests and chi-square tests were performed to compare if there were key differences on demographic variables between fathers who were included in the analyses and those who were not. No significant individual or familial differences were found except for a significant difference in child sex, χ² = 5.85, p < .05; fathers included in this study were more likely to have a target boy vs. girl.

Procedure

Recruitment.

Families were eligible to participate if at least one parent had been deployed to Afghanistan and/or Iraq since 2001 and at least one child in the home was 4–13 years old. Parents were recruited using strategies including presentations at NG/R family pre-deployment and reintegration events, a targeted mailing from the local Veterans Affairs Medical Center, etc. Interested families completed an online survey to screen for eligibility. If eligible, parents completed a questionnaire battery online and a home-based assessment before the randomization.

Study design.

Families were randomly assigned into ADAPT program (60%) or a control group (i.e., online resources and tip sheets) (40%). The unequal allocation ratio is due to anticipated drop-outs in the intervention group. Group differences on baseline demographics and scores on key variables were examined (Supplementary Material Table 1, available online), showing that they were comparable on all variables except for marital status, F (1, 142) = 8.11, p< .01; more fathers in the control group reported being married than in the intervention group. This difference may have occurred by chance due to sampling variability. The intervention program has been described previously (Gewirtz et al., 2018). Briefly, the program was a 14-week multi-family group-based program to prevent child adjustment problems in post-deployed military families by enhancing effective parenting. Weekly 2-hour meetings were held in convenient community locations, typically on weekday evenings; dinner and childcare were provided. The program used active teaching techniques (e.g., role play, discussion) to promote effective parenting.

In-home physiological data collection and family interaction tasks.

In-home assessments were administered by 2–3 trained technicians. Family members were guided to put on a Polar heart rate monitor (Model: RS800CX) on their chest for inter-beat intervals (IBIs) recording. The devices and data collection methods have been previously validated for collection of HRV (Quintana et al., 2012). Family members were videotaped completing a set of standardized and validated family interaction tasks (Forgatch, Knutson, & Mayne, 1992). The HF-HRV data collected is from fathers; in two-parent families, mothers also participated in some of these tasks. Tasks were: 1) a reading task by father-(mother)-child; 2) a problem solving task by father-child; 3) a father-mother-child problem solving task (if applicable); 4) a father-(mother)-child toy play task; 5) a father-mother problem solving task (if applicable); 6) a father-(mother)-child puzzle task; 7) a deployment discussion task by father-child; 8) a father-(mother)-child monitoring discussion task; and 9) a father-(mother)-child fun activity planning task. Each task lasted 4–5 minutes. At the end, family members were instructed to take off their heart rate monitors and were debriefed. The order of the tasks was counter-balanced between fathers and mothers: half of fathers completed the father-child problem solving task immediately after the reading task; the other half completed a different task in between the reading and father-child problem solving tasks. In the current sample, 57.2% of the fathers completed the problem solving task immediately after the reading task, and the rest completed a different task between the reading and problem solving tasks.

Physiological data processing.

Coders watched videos and entered time-stamps using BORIS software (Friard, Gamba, & Fitzjohn, 2016), which allows time locked coding of events from video streams. Task onset and offset times were matched to watch times for the same events using scripts written for the project in the R statistical software (R Core Team, 2018). Inter-beat interval/IBI data were cleaned and processed using the RHRV package (Martínez et al., 2017). Data processed by RHRV visually reviewed by the researchers for error rates and uncorrected errors. To check the reliability of the built-in error correction and derivation of HF power in RHRV, heart rate data from the reading and father-child problem-solving tasks per father were also processed in Kubios –a widely used software for HRV analysis (Tarvainen, Niskanen, Lipponen, Ranta-aho, & Karjalainen, 2014). The HF-HRV values extracted from RHRV and Kubios were strongly correlated (rs > 0.8), suggesting that RHRV data correction and extraction is reliable and valid.

Measures

Possible confounding variables for HRV data were time of assessment, medication, and alcohol use. Demographics, length of deployments, and PTSD symptoms were covariates.

Assessment time.

In-home assessments were conducted at different times in a day. The times of day when physiological assessment started were extracted from the heart rate monitor.

Medication status.

Fathers were asked to report what medications they were using, and medication types/names were evaluated to determine their possible effects on vagal tone based on the literature and consultations with a pharmacist. Answers were categorized as: 0 = no effects or no medication use; 1 = low priority medication; and 2 = high priority medication (e.g., beta blockers). Fathers who had a value of 2 were excluded from the study (n = 2). Most fathers scored 0 (n = 125) and a small number of fathers scored 1 (n = 20).

Alcohol use.

The Alcohol Use Disorder Identification Test (AUDIT: Babor, Higgins-Biddle, Saunders, & Monteiro, 2001) was used. The scale has 10 items and a categorical variable was calculated (0 = no risk; 1 = low risk; 2 = risk; 3 = high risk). Three individuals scored 2 and 1 individual scored 3, thus a dummy variable was created with 0 indicating no-risk (n = 115) and 1 indicating at-risk (n = 29).

Vagal flexibility (VF).

HF powers from the high frequency band (0.15 –1.20 Hz) were extracted for each 30-sec epoch of the reading and the father-child problem solving tasks. Low frequency (LF) power was extracted to allow the calculation of normalized units (HFnu = HF powers divided by the total of HF and LF powers; possible range: 0 to 100). A median of the 30-sec HFnu epoch values was computed per task. Finally, VF was computed by subtracting HFnu during problem solving task from HFnu during reading task: greater VF indicates more withdrawal of cardiac vagal tone, which can range between −100 to 100.

Observed parenting.

Trained coders, blind to intervention condition, scored all family interaction tasks with the exceptions of the reading task, using the Coder Impressions System (Forgatch et al., 1992). Inter-rater reliability was assessed with intraclass correlation coefficients (ICCs). Scores were obtained on five indicators –problem-solving, skill encouragement, monitoring, positive involvement and discipline. Problem-solving assessed the quality of parent-child problem solving, putting the solution to use, and extent of resolution (9 items; α = .87–.89; ICC = .88–.94). Skill encouragement focused on parents’ capacities to promote child skills through positive reinforcement and scaffolding (8 items; α = .76–.83; ICC = .72–.76). Monitoring evaluated parents’ supervision of and knowledge about the child’s activities especially in situations when parents are not physically present (4 items; α = .60–.71; ICC = .64–74). Positive involvement evaluated parental warmth, affection, sensitivity, and empathy toward the child (10 items; α = .75–.76; ICC = .76–.84). Discipline concerned parents’ use of overly strict, coercive, authoritarian, and inconsistent parenting strategies (8 items; α = .75; ICC = .58–.78), but to be consistent, this variable was reverse-coded so that higher scores mean more effective parenting. Of note, among the five parenting factors, positive involvement is about parent-child relationship enhancement whereas the other four factors focus on behavioral management. Items for the first three factors were rated on a 5-point scale ranging from 1 (untrue) to 5 (very true). Items for the last two factors were rated on a 6-point scale ranging from 1 (never) to 6 (always).

Demographic variables were father education, marital status (0 = married; 1 = not married), child age (in years), child sex (0 = boy; 1 = girl), and number of children at home. Education was coded as 1= Some high school or less, 2 = GED, 3 = High school diploma, 4 = Some college, 5 = Associate’s degree, 6 = Four-year college degree, 7 = Master’s degree, and 8 = Doctoral or professional degree.

Length of deployment was reported by fathers a scale of 1 –7 (1 = 6-months or less; 2 = 7~12 months; 3 = 13~18 months; … 6 = 31~36 months; and 7 = 37 months or more).

PTSD symptoms.

The military version of the Posttraumatic Stress Disorder Checklist (PCL; Weathers, Litz, Herman, Huska, & Keane, 1993) was used to assess posttraumatic symptoms related to military experiences. Respondents were asked to rate 17 items on a 5-point scale (1 = not at all; 5 = extremely), with each item indicating the severity of symptoms, for instance, “repeated, disturbing dreams of a stressful military experience”, “trouble falling or staying asleep,” and “feeling emotionally numb or being unable to have loving feelings for those close to you”. A composite score was created to indicate PTSD symptom severity. In the current sample, 16.0% met the clinical criteria for PTSD, based on the symptom-clustering method developed by Hoge et al. (2004).

Analyses

Data analyses followed several steps. First, HF-HRV measures were evaluated and examined for correlations with potential confounders. Second, descriptive statistics and intercorrelations were computed for key study variables. Third, a measurement model was estimated for the parenting constructs. Fourth, the intent-to-treat (ITT) intervention effect on parenting outcome as well as the moderation effect by VF were estimated in structural models, while controlling for baseline parenting, parent education, marital status, length of deployment, PTSD symptoms, child age, child sex, and number of children. Model fit indices were evaluated using recommended criteria (McDonald & Ho, 2002): a χ² /df ratio below 2.00, a comparative fit index (CFI) above .95, a standardized root mean square residual (SRMR) below .08, and a root-mean-square error of approximation (RMSEA) below .06. All structural equation modeling analyses were computed in Mplus 8.1 (Muthén & Muthén, 1998–2017).

Missing data.

Observed parenting pre-intervention data yielded 4.80~6.20% missingness on all indicators except for monitoring (11%). This is likely because the monitoring task was one of the last family interaction tasks and child fatigue led some families to end the assessment early. Observed parenting 1-year follow-up data had 22.8–24.1% missingness on all indicators due to attrition. All demographic or deployment-related variables had less than 2% missing values. The VF variable had complete data (as a result of sample manipulation, see Online Supplementary Material Figure 1). Independent sample t-tests showed significant differences between fathers who had missing values pre-intervention (n = 16) and those who did not (n = 128): fathers who missed parenting measures pre-intervention reported lower education, t (142) = 2.18, p < .03, younger child age, t (143) = 2.19, p < .05, and less length of deployment, t (143) = 2.28, p < .05. To handle this, full information maximum likelihood (FIML) was used.

Results

Sample characteristics.

Fathers were 37.46 years old on average (SD = 6.75; Range: 24 –58). Their socioeconomic profiles were mostly middle-class: 26.6% reported annual household income above $100,000 while more than half (64.3%) reported annual household income between $40,000 –$99,999. Half (51.4%) of the fathers had completed at least a 4-year college degree while 40.3% had attended some college or received an Associate’s degree. Most fathers were White (89.7%) and married or partnered (88.3%) for 9.57 years on average (SD = 5.28; Range: 1 –27). The mean of number of children at home was 2.45 (SD: 1.00; Range: 1 –6); about half of the target children were boys (53.8%), and children were 8.66 years old on average (SD = 2.74; Range: 4.06 –13.61). Most fathers had been deployed 1–3 times (39.3%, 37.2%, 17.2%, once, twice, and three times, respectively). One third (35.2%) of the sample reported a cumulative length of deployment as less than 12 months, another one third (31.7%) reported 13–24 months, 22.8% reported 25–36 months, and 10.3% reported more than 37 months.

Preliminary results.

HF-HRV during reading and HF-HRV during problem solving were moderately correlated (r = .43, p < .001). As predicted, on average, HF-HRV decreased from reading (M = 32.00, SD = 15.34) to problem solving (M = 26.75, SD = 11.31), t(144) = 4.33, p < .001, suggesting vagal suppression. VF was a normally distributed variable in the sample with a considerable range (M = 5.24, SD = 14.58, Skewness = 0.22, Kurtosis = 0.30, Range = −37.32 ~ 47.02). No significant VF difference was detected between those who completed the problem-solving task immediately after the reading task and those who completed another task in between the two (p > .05). No evidence was found for VF differences between alcohol use at-risk and no-risk groups (p > .05), or medication-use and no-medication-use groups (p > .05). Home assessment time and VF were not significantly correlated (r = 0.10, p > .05). While fathers with PTSD appeared to have lower VF (n = 23, M = 2.31, SD = 15.70) than those without PTSD (n = 121, M = 5.56, SD = 14.18), the difference was not statistically significant, t (142) = 0.99, p > .05. And the association between VF and PTSD symptoms was weak and not statistically significant (r = −.15, p > .05). VF was significantly correlated with length of deployments (r = −.20, p < .05), meaning that the longer the fathers had been deployed, the lower VF. Observed parenting data were normally distributed except for discipline; scores were negatively skewed (Skewness was 1.7 at baseline and 5.2 at 1-year follow-up), indicating that most parents did not show much harsh punishment and coercive parenting. No pre-existing differences were found between the intervention and control groups on key variables except that the control group’s skill encouragement was higher than the intervention group, t (135) = 2.36, p < .05. The descriptive statistics and intercorrelation coefficients among key variables are presented in Supplementary Material Table 2 (available online).

Confirmatory factor analysis (CFA).

A theory-driven measurement model was estimated, consisting of two latent constructs, pre-intervention parenting and parenting at 1-year follow-up, each indicated by five observed variables (positive involvement, problem-solving, discipline, skill encouragement, and monitoring). The same indicators were allowed to covary across time (e.g., discipline pre-intervention correlated with discipline 1-year). The resultant model demonstrated a good fit to the data: χ² (30) = 36.62, p = .19, χ² /df <2.00, CFI = 0.98, RMSEA = .04, SRMR = .06. At pre-intervention and follow-up respectively, standardized factor loadings were 1.00 and 0.95 for positive involvement, 0.54 and 0.53 for problem solving, 0.30 and 0.40 for discipline, 0.67 and 0.54 for skill encouragement, 0.35 and 0.19 for monitoring. Thus, positive involvement was a dominant factor for the latent parenting construct whereas monitoring and discipline contributed less to the latent parenting construct than other factors in the current sample. Therefore, as post hoc analyses, we analyzed five parenting indicators separately as individual outcomes in addition to the latent construct of parenting. Testing of measurement invariance for the parenting construct at two time points showed that configural and metric invariance were met.

ITT analysis.

A structural model estimating ITT intervention effect without moderation was computed: parenting at 1-year follow-up was predicted by group assignment while controlling for baseline parenting and covariates (education, marital status, child age, child sex, number of children, length of deployment, and PTSD symptoms). The model showed an acceptable fit with a CFI slightly lower than the recommended threshold: χ² (93) = 115.06, p = .06, χ² /df < 2.00, CFI = 0.93, RMSEA = .04, SRMR = .06. No ITT effect was detected for fathers’ observed parenting at 1-year follow-up, B = 0.055, S.E. = 0.079, β = 0.065, p > .05. Results from post hoc analyses showed no ITT effects for skill encouragement, discipline, monitoring, problem solving, or positive involvement, ps > .05.

Moderation analysis.

The moderation model was identical to the ITT model described above except for two additional predictors for parenting outcome: mean-centered VF and the interaction between group assignment and mean-centered VF. This model (Figure 1) showed a good fit to the data: χ² (110) = 127.62, p > .05, χ² /df < 2.00, CFI = 0.95, RMSEA = .03, SRMR = .06.. Group assignment and VF did not significantly predict parenting at 1-year follow-up, ps > .05, but VF significantly moderated the intervention effect on parenting at 1-year follow-up while controlling for baseline parenting and covariates, B = 0.013, S.E. = 0.005, p < .05. Plotting the region of significance revealed that fathers with higher levels of VF showed higher levels of observed parenting at 1-year follow-up, relative to intervention fathers with lower levels of VF, if they were randomized into the intervention vs. control. The region of significance (Online Supplementary Material Figure 2) on the right side included individuals with VF levels greater than 0.78 SD above the sample mean (22.76% of the sample), meaning that these fathers who were assigned to ADAPT showed better parenting at 1-year follow-up, after adjusting for baseline parenting and covariates, in comparison to those who were assigned to the control group. The region of significance on the left side included fathers with extremely low VF (2.33 SD below the sample mean) and these fathers exhibited lower levels of parenting at 1-year follow-up, accounting for baseline parenting and covariates, if they were randomized into the ADAPT program vs. the control group. However, only 1.40% of the sample (2 out of 145) had such low VF levels. Between the two regions of significance are the fathers who were randomized into the intervention group did not demonstrate significant differences in parenting at 1-year follow-up than the controls. The model explained 44.7% of the variance in parenting at 1-year (R²= .447).

Figure 1.

Structural equation model for testing intervention effects moderated by vagal flexibility. Paths are standardized coefficients. * p<.05, ** p<.01, ***p<.001. The same indicators are specified to covary across time points (not shown). ITT = intent-to-treat intervention. PINV = positive involvement. PS = problem solving. DIS = harsh discipline (reversed). ENC = skill encouragement. MON = monitoring. Significant paths are shown as solid lines. Nonsignificant paths are shown as dotted lines. Model fit: χ² (110) = 127.62, p > .05, χ² /df < 2.00, CFI = 0.95, RMSEA = .03, SRMR = .06. All baseline control variables were specified to covary with each other.

Results of post hoc analyses testing individual parenting outcome showed that VF significantly moderated intervention effects on positive involvement (B = 0.015, S.E. = 0.005, β = .364, p < .01). The region of significance for this interaction effect showed that fathers with higher VF (1.06 SD above the mean, 15.2% of the sample) showed higher levels of positive involvement at 1-year if they were randomized into the intervention vs. control condition, while fathers with lower VF (1.10 SD below the mean, 11.7% of the sample) showed lower levels of positive involvement at 1-year if they were randomized into the intervention vs. control.

Discussion

Identifying potentially malleable biomarkers is a critical path towards precision medicine because it goes beyond diagnostic psychopathology which is based on behavioral symptoms and does not inform pathophysiology or treatment responses (Insel, 2014). Indeed, to understand normal and abnormal behaviors, we need a better understanding about mechanisms of pathophysiology to inform more precise, targeted interventions that may benefit individuals with different characteristics who may have different needs. This study is the first to report deployed fathers’ parasympathetic regulation of emotion during social interactions as a predictor of their responsivity to a parent training program designed for military families. Our key finding is that fathers with higher levels of VF pre-intervention randomized to the ADAPT condition showed significantly improved observed parenting at 1-year relative to controls.

Cardiac vagal tone, in general, and VF, in particular, index social engagement behaviors (Porgest, 2007). In the Research Domain Criteria project (RDoC), which is being developed to deconstruct diagnostic classifications and identify underlying mechanisms for mental disorders, one of the domains is systems for social processes (Cuthbert & Insel, 2013). It emphasizes social communication, affiliation, and attachment. Specifically, it concerns one’s capacities of facial, vocal, gestural, and postural processing. The vagus nerve supports these processes as it not only connects to organs such as the heart, but also it has integrated with the nuclei that regulate the muscles of the face, neck and head. Thus, higher VF may support better social communication by effectively regulating the body’s metabolism and organs/muscles involved in vocalizations and body language, contributing to individuals’ engagement behaviors in the environment (Porges, 2011, p. 140). This is consistent with a prior study showing that individuals with higher VF showed greater social sensitivity than those with lower VF (Muhtadie et al., 2015).

In addition, VF may index emotion regulation (Porges, 2007; Balzarotti et al., 2017). Considering its particular role in social communication, VF may have greater implications for interpersonal (vs. intrapersonal) emotion regulation. Emotion regulation has been theoretically linked to parenting (Crandall, Deater-Deckard, & Riley, 2015). Previously in a cross-sectional study, we found that combat deployed fathers’ VF curtailed the negative influence of experiential avoidance on observed parenting behaviors such as observed positive engagement (Zhang et al., 2019). Experiential avoidance is an indicator of poor emotion regulation at an intrapersonal level and is associated with poor parenting behaviors (Brockman et al., 2016). Yet, this association was not evident in the subgroup of fathers with higher VF. Fathers with high VF may be more capable of regulating negative emotional arousal efficiently and effectively in social environments (e.g., parent-child interaction), to a degree that they demonstrate vagal withdrawal for increased engagement and communication behaviors.

Taken the above literature into account, our findings revealed that adaptive parasympathetic regulation may enable combat deployed fathers to show enhanced parenting behaviors –i.e. to benefit from a parent training program. Fathers with higher VF may be more prepared to learn and consolidate what they learned in the intervention than those with lower VF. We also found, in post hoc analyses, that only positive involvement was significantly improved in fathers with high VF as a result of the intervention. Among the five parenting factors, positive involvement is arguably the only factor that solely focuses on affective aspects of the parent-child relationship while the other factors pertain primarily to behavioral parenting skills (i.e. child management). It is possible that VF is most saliently associated with this affective aspect of parenting. Considering that our father sample was deployed military service members reintegrating into families and communities, and their spouses had been taking up most of the child management responsibilities while they were absent, fathers’ improved positive involvement with their child is quite meaningful. A warm, nurturant, and involved father is key to a positive father-child relationship, which is protective for both the fathers and the children.

While only two fathers who were randomized into the intervention (vs. control) showed poorer parenting at 1-year is a negligible number, the results from post hoc analyses warrant discussion. In the full sample, most fathers’ cardiac vagal tone decreased from baseline to problem solving, which is consistent with the literature that HRV decreases during mental demand (Shahrestani, Stewart, Quintana, Hickie, & Guastella, 2015). Such decreases are considered adaptive (Balzarotti et al., 2017). Low VF means increased cardiac vagal tone from baseline to problem solving, and post hoc analyses showed that fathers with low VF showed lower levels of positive involvement at 1-year, adjusting for baseline positive involvement and covariates, if they were randomized into the ADAPT vs. control condition. These fathers might have attempted effortful physiological regulation of their negative emotional arousal during the problem-solving task with their child. This may reflect a deficit in the ability to withdrawal vagal tone during attention-demanding situations (Porges, 2011). These fathers may show depression, anxiety, PTSD, or comorbid problems that may or may not constitute a diagnosis. In fact, a prior study found that from baseline to a stressful task, healthy individuals exhibited decreased cardiac vagal tone, but depressed individuals exhibited increased cardiac vagal tone (Rottenberg, Clift, Bolden, & Salomon, 2007). An abnormal pattern of vagal reactivity may index emotion regulation problems at a physiological level. Taken together with prior findings that fathers with clinical levels of PTSD symptoms showed poorer parenting if they were randomized into the ADAPT (Chesmore et al., 2018), we may need to develop a more tailored intervention for these high-risk fathers.

We consider VF to be an underlying mechanism of psychopathology but not necessarily indexing specific psychopathology such as PTSD. A meta-analysis assessed HRV studies in the context of social interactions and found that compared to healthy individuals, those with psychopathology did not show as much flexibility of cardiac vagal tone (Shahrestani et al., 2015). Yet, in this meta-analysis, none of the studies assessed PTSD. Few studies have examined the link between cardiac vagal tone and PTSD in military samples (Minassian et al., 2014); those that have done so have primarily been lab-based studies lacking a social context for measuring cardiac vagal tone. In our study, we controlled PTSD symptoms in the model, and found that the correlation between VF and PTSD symptoms was weak and did not yield statistical significance.

We measured baseline cardiac vagal tone during a reading task (rather than at “true” resting vagal tone) to capture the parent-child context. In our study, baseline cardiac vagal tone was moderately associated with VF. Prior findings showed that higher baseline cardiac vagal tone was associated with less perceived stress, anxiety, and depression, but VF was only significantly correlated with loneliness –a variable that has implications for social relationships (Muhtadie et al., 2015). VF may be a distinct measure of emotion regulation from baseline cardiac vagal tone, but more research is needed to clarify their difference and commonality.

A few studies have tested child cardiac vagal tone as a moderator of child outcomes in interventions (Bagner et al., 2012; Glenn et al., 2018). These interventions directly targeted at risk children or youth. While the ADAPT program did not target children directly, children’s cardiac vagal tone might also be a moderator for their health outcomes. For example, children’s vagal suppression moderated the relationships between paternal socialization and children’s anxious difficulties (Hastings et al., 2007).

Several limitations and future directions should be considered. First, the sample size may have limited the observable range of VF and the statistical power to detect effects in the structural equation model, which might explain the finding that most fathers with about mean levels of VF were not impacted by the intervention. In particular, the weak magnitude of factor loadings (e.g., < .50) in the CFA model and the amount of missing data (22.8–24.1% at 1-year follow-up) both necessitated larger sample sizes for adequate statistical power. Also, the observed correlation between cardiac vagal tone and parenting in the current study was weak, thus more evidence is needed to draw conclusions regarding the relationship between parental cardiac vagal tone and parenting behaviors in other parent populations in both civilian and military populations, in particular, those who are at risk for mental health problems such as PTSD. Second, VF was operationally measured as vagal activity changes across two tasks that were differently manipulated. In contrast, most studies in the literature directly assessed baseline resting cardiac vagal tone using a resting task, which makes it difficult to compare findings across studies but makes this study more novel. In addition, physiological assessments were completed in families’ homes instead of a laboratory, and VF was measured during family interaction tasks. Such a design has both strengths and limitations (e.g., other family members’ physiological reactivity was not controlled for in the analyses). Third, because the sample in the current study had all been NG/R members who were exposed to combat trauma and returned from war, our findings may not be generalized to other family contexts or to mothers.

Our study suggests that high-risk fathers need interventions to strengthen emotion regulation in general, and VF in particular, in order to benefit from ADAPT. This has several implications for prevention-based interventions. First, future researchers may utilize adaptive intervention designs to examine whether it would be fruitful to provide additional intervention before the delivery of the ADAPT program for high-risk fathers, which would require program-level personalization. While the ADAPT program has already incorporated components to enhance emotion regulation in parents through brief mindfulness exercises, the dosage is limited and thus a more intense emotion regulation training intervention may be necessary. Second, it could be that the existing dosage of mindfulness training within ADAPT is sufficient, but greater participant engagement is needed. If this is the case, strategies to promote engagement could be designed at participant-, provider-, and even community-levels (Supplee, Parekh, & Johnson, 2018). For example, participants’ motivation and goals to practice mindfulness and engage in adaptive emotion regulation could be addressed. Given the insufficient number of studies on this topic, more research is needed to inform best practices.