Within-person patterns of psychobiological stress response correspondence: Links to preadolescent internalizing problems and coping behaviors

Abstract

Though close correspondence across the affective experience and physiologic arousal levels of the stress response (i.e., psychobiological stress response correspondence) is thought to support efficacious stressor management (i.e., effortful coping), little empirical evidence has demonstrated such correspondence. Identifying children for whom a lack of such correspondence is characteristic may help inform efforts to detect and prevent risk for internalizing psychopathology. Using a community sample of preadolescents (N=151, M_age=10.33 years, Min_age=8.92, Max_age=12.00, 51.7% male), this person-centered study examined internalizing problem and coping linked variability in psychobiological stress response correspondence. Children were exposed to the Trier Social Stress Test (TSST). Self-reported negative affect (NA) ratings and salivary cortisol (SC) levels were assessed at six time points during the experimental protocol. Multitrajectory modeling revealed four subgroups. Relative to In-Touch (i.e., Moderate NA–Moderate SC; n=65), Unmindful (i.e., Moderate NA–Low SC; n=49) were more likely to present with parent-reported but not self-reported internalizing problems; Vigilant (i.e., High NA–Low SC; n=13) were more likely to present with self- and parent-reported internalizing problems, more likely to report experience of involuntary stress responses (e.g., rumination), less likely to use engagement coping (e.g., problem-solving), and more likely to use wishful thinking coping (e.g., “I wish problems would just go away.”); Denial (i.e., Low NA–High SC; n=24) self-reported similarly low internalizing problems, though they were also more likely to report reliance on denial coping (e.g., “I pretend problems don’t exist.”). Findings illustrate meaningful heterogeneity in preadolescent psychobiological correspondence with implications for multimodal assessment and outcome monitoring in coping-based preventative interventions.

Introduction

Lack of correspondence across the affective experience (e.g., self-report of negative affect) and physiological arousal (e.g., salivary cortisol) levels of analysis in response to a stressor, referred to as psychobiological stress response correspondence, has empirically been more the rule than the exception (Campbell & Ehlert, 2012). Due to reliance on variable-centered approaches (e.g., bivariate correlations), few studies have identified, for example, subgroups for whom higher (e.g., moderate affective experience–moderate physiological arousal) or lower (e.g., low affective experience–high physiological arousal) psychobiological stress response correspondence may be characteristic. Understanding within-person patterns of such correspondence may help inform efforts to detect children at risk for internalizing problems (e.g., anxiety, depression) that traditional assessment modalities (e.g., self-report) may overlook. Such research may also help clarify whether specific strategies for managing stressors (i.e., coping skills) contribute to such risk and potentially interfere with detection (Kuhlman & Mousavi, 2020). Integrating data across the affective experience and physiological arousal levels of analysis, this study examined coping-based contributions to psychobiological stress response correspondence and related risk for internalizing problems in a community sample of preadolescent boys and girls.

Functions of Psychobiological Stress Response Correspondence and Related Risk for Internalizing Problems

From a functional perspective, psychobiological correspondence is thought to help an individual effectively respond to a stressful situation (Shirtcliff et al., 2014). That is, upon detection of a stressor, affective experience orients a person to the source of stress and signals the need to take action while physiological arousal mobilizes resources needed to do so. At the neurobiological level, prefrontal cortical regions integrate sensory information to support an individual’s appraisal of the source of stress and assessment of relevant coping options. Projections from these regions to subcortical limbic areas signals the hypothalamic-pituitary-adrenal axis (HPA) to increase arousal and provide energy (i.e., increased glucose mobility) that helps individuals effectively manage the source of stress (Herman et al., 2005). Thus, greater psychobiological correspondence may index more coherent stress response function that is capable of supporting more efficacious stressor management (i.e., effortful coping).

Conversely, lower psychobiological correspondence may interfere with attempts to efficaciously meet the demands of a stressor in ways that signal risk for internalizing problems. A mismatch between affective experience and physiological arousal may reflect difficulties with awareness of internal states and appropriate orienting towards or away from sources of stress. In the case of low affective experience accompanied by high physiological arousal, these difficulties may be associated with effortful disengagement (e.g., denial) from the source of stress; e.g., negative affect suppression and risk-conferring physiological rebound (e.g., hyperarousal; Wenzlaff & Wegner, 2000). This claim is consonant with the verbal-autonomic response dissociation literature, which demonstrates that repression of self-reported distress in the face stressors can in part explain avoidance–illness linkages (Schwerdtfeger & Kohlmann, 2004). Alternatively, in the case of high affective experience accompanied by low physiological arousal, such difficulties may be associated with involuntary stress responses (e.g., hyper-vigilance) to the source of stress; e.g., negative affect related threat monitoring and risk-conferring physiological habituation (e.g., hypoarousal; Anderson & Hope, 2009; Pine et al., 2009). In each case, these difficulties may contribute to a compromised ability to marshal requisite executive resources and, thus, utilize engagement coping (e.g., problem solving) in service of identifying and navigating stressors.

Preadolescence presents as a ripe time for the study of psychobiological stress response correspondence. Preadolescence is marked by the development of executive capacities that support children’s ability to attend to, reflect on, and express internal states, as well as utilize complex coping (Zimmer-Gembeck & Skinner, 2016). Extensive psychobiological stress response system reorganization also takes place during this developmental stage (Spear, 2009). These cognitive and biological changes occur against the backdrop of preadolescents’ exposure to a host novel acute stressors in their day to day lives (e.g., academic, interpersonal), those which they are increasingly expected to navigate in autonomous and independent fashion. When poorly managed, these stressors are known to confer risk for poor mental health outcomes into adulthood (e.g., Compas et al., 2017). Thus, understanding how preadolescent normative (i.e., high) or atypical (i.e., low) patterns of correspondence either support or constrain children’s ability to cope with such stressors may help update existing stress response models of psychobiological risk for the development of internalizing psychopathology (Spear, 2009; Zimmer-Gembeck & Skinner, 2016).

Person-Centered Approaches to Psychobiological Stress Response Correspondence

Given the previous accounts, both normative and atypical patterns of psychobiological stress response correspondence should be assumed. However, Campbell and Ehlert (2012) noted that only 25% of laboratory studies that examined such correspondence using the Trier Social Stress Test (TSST; Seddon et al., 2020) found significant affective experience-physiological arousal associations, noting in particular this lack of association in studies examining subjective negative affect and salivary cortisol stress response indices, specifically. Thus, whether self-reports of affective distress and markers of neuroendocrine function, in particular, correspond in the face of socio-evaluative stress is largely unknown. Still further, no study has attempted to establish such relations in children.

Recognition that such lack of correspondence may stem from reliance on modest variable-centered methods as well as unexamined individual differences therein has led to the call for more nuanced modeling approaches and consideration of psychosocial concomitants (Campbell & Ehlert, 2012). Recent person-centered studies suggest that lack of correspondence at the sample average level may be explained by the presence of subgroups with unique within-person, affective experience-physiological arousal correspondence patterns linked to symptomatic functioning and coping skill utilization. For example, Villada et al. (2016) used cluster analyses of 35 young adults’ average negative affect and salivary cortisol levels in response to the TSST to identify two subgroups, with the low correspondence subgroup (e.g., high negative affect–low salivary cortisol) reporting greater internalizing problems and reliance on avoidant coping. Thus, lack of correspondence may not be methodological noise, but perhaps a source of theoretically and clinically informative heterogeneity to be illustrated with methods well-equipped to model such variability. Such has been proposed in both theoretical and empirical accounts of repressive coping (Schwerdtfeger & Kohlmann, 2004). However, no study has examined patterns of verbal –physiologic dissociation using neuroendocrine indices of arousal in youth.

Recent Advances in Person-Centered Methods

The emerging person-centered psychobiological stress response correspondence literature could be advanced in several ways. Larger studies that focus on children and examine correspondence with neuroendocrine markers, specifically, are needed. Such studies may benefit from using paradigms known to reliably elicit affective (e.g., negative affect) and HPA (e.g., salivary cortisol) responses in youth samples (e.g., TSST; Seddon et al., 2020). Also, person-centered studies often examine either basal levels (e.g., pre-TSST levels) or summative stress response scores (e.g., area-under-the-curve increase; AUCi) in isolation. Although useful, these approaches do not take advantage of stress response data collected at multiple time points during laboratory-based stressor protocols. More nuanced understanding about specific facets of psychobiological correspondence that might contribute to children’s coping and symptomatology (e.g., basal level correspondence, reactivity correspondence) may be obtained by utilizing methods that illustrate how their stress responses in the face of acute stress unfold over time across multiple levels of analysis.

Multitrajectory modeling (MTM; Nagin et al., 2018) is one such method that may address some of these limitations. MTM can be used to identify subgroups of individuals in a population who share similar joint trajectories (e.g., affective response, physiological response) of a phenomena of interest (e.g., psychobiological stress response correspondence). MTM may help address the limitations of prior work noted above, insofar as it can capture correspondence via full stress response function (as opposed to summative scores) and model coordinated stress response function across affective experience and physiological arousal indices simultaneously (as opposed to in isolation). Following identification, theoretically-informed aspects of psychosocial functioning can be examined as correlates of subgroup membership. In this study, and because MTM permits analysis of distinct stress response phases (i.e., baseline, pre-TSST; reactivity, post-TSST), it may also provide informative detail about how phase-specific correspondence relates to children’s internalizing problems and coping skill utilization (Campbell & Ehlert, 2012).

MTM is also useful when research suggests that the expectation of population heterogeneity (i.e., that individuals in a sample may exhibit multiple unique patterns of change) might be stronger than that of population homogeneity (i.e., that individuals in a sample all exhibit more or less the same single pattern of change) in a construct of interest. For the current study, prior developmental theory and empirical evidence provides such guidance, each having posited and illustrated qualitatively-distinct, healthy and dysregulated change patterns in children’s acute stress responses to the TSST (Gunnar et al., 2009; Van Ryzin et al., 2009). However, as this work has tended to focus on various psychobiological indices in isolation (e.g., cortisol only) and given the paucity of empirical, multi-level correspondence studies focusing on youth, we can only speculate based on this knowledge that heterogeneity in psychobiological stress response correspondence might be an informative characteristic of the preadolescent period.

The Current Study: Exploratory Aims and Hypotheses

To address these gaps, we exposed a community sample of 151 preadolescents to the TSST (for details, see Wadsworth et al., 2018). Over the course of the experimental protocol, we collected data at each of seven time points on children’s affective experience (i.e., self-reported negative affect; NA) and physiological arousal (i.e., salivary cortisol; SC) levels. We then utilized MTM (Nagin et al., 2018) to identify subgroups of children based on the extent to which they exhibited similar joint NA and SC stress response trajectories.

Aim 1: We sought to identify subgroups of children with distinct within-person patterns of psychobiological stress response correspondence. H₁: Based on the extant developmental (Gunnar et al., 2009), correspondence (Campbell & Ehlert, 2012), and person-centered (Villada et al., 2016) literature, we expected to identify at least one subgroup reflecting higher correspondence (e.g., Moderate NA–Moderate SC) and two subgroups reflecting lower correspondence (e.g., Low NA–High SC, High NA–Low SC). Aim 2: We sought to examine children’s internalizing problems, effortful coping, and involuntary stress responses as correlates of subgroup membership. H₂: Based on evidence of links between lack of correspondence and internalizing difficulties (Anderson & Hope, 2009; Pine et al., 2009), we expected the likelihood of membership in subgroups reflecting lower (relative to higher) correspondence to increase with greater levels of internalizing problems, disengagement coping (e.g., effortful avoidance for the Low NA–High SC group), and involuntary stress responses (e.g., threat monitoring or hypervigilance for the High NA–Low SC group), as well as decrease with greater levels of engagement coping (e.g., both Low NA–High SC and High NA–Low SC groups). Given the lack of uniformity in the way three constituent types of disengagement coping (i.e., avoidance, denial, and wishful thinking) relate to children’s emotional clarity (Flynn & Rudolph, 2007), we explored each as independent correlates of subgroup membership.

Method

Participants

Study flyers were distributed to 4^th and 5^th grade children (N=151, M_age =10.33 years, Min_age=8.92, Max_age=12.00, 51.7% male) attending suburban and semirural elementary schools in the Northeastern US. Interested parents (85.4% mothers) completed online questionnaires and consent. Families were then contacted to schedule an in-person visit. Median annual household income was $70,500 (Min=$13,638.84; Max=$500,000). Most participants identified as White (93.4% of parents, 90.7% of children).

Procedures

As outlined in Wadsworth et al. (2018), children were asked to refrain from brushing their teeth, consuming a large meal, or dairy/sugary/acidic foods within an hour of arrival. To reduce diurnal variation effects, visits were scheduled in the afternoon within 30 min of each other: 3:30 pm, 4:00 pm, 4:30 pm. At the start of the visit, parents and children were briefly reminded of the study protocol provided to them during their online enrollment and consent. Parents then provided written consent and children provided written assent. Children then provided a saliva sample and affect rating (T1). An experimenter then administered questionnaires to the child (40 min). Immediately after, a second sample/rating was collected (T2). Children then completed the TSST, where they were instructed to prepare (5 min) and verbally deliver (5 min) a speech and perform mental subtraction (5 min) in front of a “panel of experts” (i.e., confederates unknown to participants and masked to coping condition). Immediately after, a third sample/rating was collected (T3). Children were then escorted (15 s) to one of two rooms based on a randomly-assigned coping condition (10 min): distraction (n=78; room with musical instruments, art supplies, toys; children invited to engage with the materials), avoidance (n=73; room free of distraction materials; children prompted to try not to think about their performance). This aspect of the study design was included to address hypotheses not relevant to the current investigation. Study variables did not vary across conditions (all p>.05). Children were randomized to condition within gender. Study staff were not privy to the results of this process. Immediately after, a fourth sample/rating was collected (T4). Children then completed a manipulation check interview with the experimenter (10 min). Then, a fifth sample/rating was collected (T5). Next, the child and experimenter followed along with a progressive muscle relaxation audio tape (10 min). PMR helped a) minimize potential risks to psychological well-being as per our Institutional Review Board (IRB) and b) ensure the TSST did not have linger effects on children. Prior studies have used similar post-TSST methods (e.g., 35-min calming movie, Shull et al., 2016; 45-min relaxing reading, Zoccola et al., 2014). A final sixth sample/rating was collected (T6). Children were then debriefed and families received $50 for participating. All study procedures were approved by the Pennsylvania State University IRB.

Measures

Salivary cortisol (SC).

Six saliva samples were collected via passive drool (Davis et al., 2002), stored at −20° C in an ultra-low temperature freezer, and transported on dry ice to the Biomarker Core Lab at Penn State University. Cortisol levels were determined using a commercial expanded-range high-sensitivity enzyme immunosorbent assay kit. Cortisol extraction was run in duplicate and batched in the same order as random assignment. The intra-assay and inter-assay coefficients of variation were 2.0% and 10.2%, respectively.

Negative affect (NA).

Immediately following each saliva sample, children used a visual analogue scale to rate the extent to which they presently identified with six feeling items: Nervous, Sad, Angry, Happy, Surprised, and Excited. For each feeling, children were presented with Likert type response options ranging from Not at All (1) to Very (5). An NA composite was computed by aggregating children’s Nervous, Sad, and Angry ratings (α=.86). Similar self-generated Likert type and visual analog scales have been used to assess negative affect in response to the TSST (for review, see Campbell & Ehlert, 2012).

Internalizing problems.

The Behavior Assessment System for Children, 2^nd Edition (BASC-2; Reynolds & Kaphaus, 2004) was used to assess child internalizing problems. Children completed the Self Report of Personality–Child (SRP–C, ages 8–11) and parents completed the Parent Rating Scales–Child (PRS–C, ages 6–11). One child (12.00 years) was outside the BASC-2 age range. Removing this child from analyses did not alter study conclusions. Thus, the child was retained in final analyses. The SRP and PRS are reliable and valid measures of problem behavior with good internal consistency (SRP Internalizing Problems, α=.95; PRS Internalizing Problems, α=.92). Broadband internalizing problems T-scores with combined gender norms were used to ensure scale integrity.

Responses to stress.

Children completed the Responses to Stress Questionnaire – Peer Domain (RSQ; Connor-Smith et al., 2000), a 57 item measure containing 3 coping and 2 involuntary stress response factors. Primary Control (α=.79) items assess problem solving, emotion regulation, and emotion expression. Secondary Control (α=.75) items assess positive thinking, distraction, cognitive restructuring, and acceptance. Disengagement (α=.75) items assess avoidance, denial, and wishful thinking. Subscales for avoidance (α=.62), denial (α=.65), and wishful thinking (α=.58) were also explored. Involuntary Engagement (α=.88) items assess rumination, intrusive thoughts, emotional and physiological arousal. Involuntary Disengagement (α=.86) items assess emotional numbing, cognitive interference, inaction, and escape. Primary and Secondary Control factors (r=.63) and Involuntary Engagement and Disengagement factors (r=.88) were strongly correlated. Thus, Engagement Coping and Involuntary Stress Response composites were generated by aggregating Primary and Secondary Control Coping (Sontag & Graber, 2010) and Involuntary Engagement and Disengagement Stress Response factor scores (Epstein-Ngo et al., 2013).

Covariates.

Covariates include gender (boys=0, girls=1), age, pubertal status, medication use, and coping condition (in MTM analyses). Pubertal status was assessed with the Pubertal Development Scale (PDS; Petersen et al., 1988). Parents rated (0=has not yet begun, 3=seems completed) height, body hair growth, menarche (0=no, 1=yes) and breast growth for girls (α=.71), and voice changes, height, body and facial hair growth for boys (α=.34). This low alpha for boys relative to that observed of girls may be expected given that boys enter puberty later than girls. Also noteworthy, relative to Petersen et al. (1988), boys’ pubertal status values were low: voice deepening (M=0.12, SD=0.40), growth spurt (M=1.12, SD=0.84), facial hair (M=0.08, SD=0.27), and body hair (M=0.38, SD=0.61). Removing scale items and log transforming PDS item values did not improve α levels. Medications were rated (e.g., 0=not plausible, 2=very plausible) across eleven pathways by which they may have influenced cortisol (Granger et al., 2009). The summed score across medication ratings was used. While an examination of coping condition effects on recovery phase psychobiological correspondence was entertained, such investigation was deemed beyond the scope of the paper and best addressed in a separate manuscript. Nevertheless, coping condition was retained as a time-varying covariate in all MTM analyses.

Data Preparation

Data reduction and preprocessing.

Children whose age was atypical for fourth and fifth grade (n=1) and who did not participate in the experiment (n=1) were excluded from analyses (final N=151). A fourth-root transformation was used to successfully normalize stress response data (Miller & Plessow, 2013). As recommended (Connor-Smith et al., 2000) RSQ ratio scores were created by dividing the engagement coping, disengagement coping, and involuntary stress response factors by the sum of all RSQ factor scores, reflecting the predominance of a specific way of responding to stress. Avoidance, denial, and wishful thinking subscale ratio scores were created in similar fashion for exploratory purposes.

Overview of Analyses

Aim 1:

Multi-trajectory modeling (MTM; Nagin et al., 2018) was used to identify within-person patterns of psychobiological stress response correspondence based on the extent to which children exhibited similar NA and SC trajectories (i.e., baselines, reactivity patterns). The PROC TRAJ procedure (SAS 9.4) with the MULTGROUPS option employed was utilized, with Full-Information-Maximum likelihood (FIML) handling of missing data; Little’s (1988) MCAR test, Χ ² (277)=247.912, p>.250. At each step of model specification (e.g., one-group, two-group), non-significant higher order polynomial functions (e.g., quartic, cubic) for each indicator were eliminated. The log Bayes factor approximation [2log_e(B₁₀)] was utilized at each step as a fit index (e.g., [2log_e(B₁₀)]>10 supports more complex solution; (Nagin et al., 2018). Given our sample size (N=151) and recommendations (N>100; Nagin, 2005), we limited model specification to four groups. In other words, while we expected MTM to identify at least three subgroups, we allowed theory and model fit indices to guide us to the most informative and parsimonious solution. Following specification, we evaluated MTM adequacy via average posterior probability (AvePP_j> 0.70), odds of correct classification (OCC_j> 5.00), and the ratio of the probability of subgroup assignment to the proportion of adolescents assigned to subgroups ([Prob_j/Prop_j]≈1) (Nagin et al., 2018). After adequacy evaluation, Wald tests were used to distinguish the subgroups, delineating how intercept and rates of change for NA and SC trajectories were comparatively “higher” or “lower” (e.g., these terms are used exclusively to describe NA and SC baseline levels) or “more pronounced” or “less pronounced” (e.g., these terms are used exclusively to describe the magnitude of NA or SC response patterns) across the subgroups.

Aim 2:

Multinomial logistic regressions (with listwise deletion to handle missing correlate data; 0.9%) were used to examine subgroup membership and correlate associations using PROC CATMOD. An initial model examined study covariate to subgroup membership associations. Subsequent multinomial logistic regressions tested one of eight focal correlates, with study covariates and correlates in each model entered together in a single step.

Results

Results are organized by our study aims. Descriptive and bivariate statistics are shown in Table 1. NA and SC values for children who were and were not taking medication did not significantly differ (all p>.250). SC levels over the course of the experimental procedure were strongly correlated, as were NA levels. Of the 36 possible NA – SC bivariate associations, only four were significant: T3 NA – T2 SC (r=−.19), T4 NA – T2 SC (r=−.27), T5 NA – T2 SC (r=−.19), T4 NA – T3 SC (r=−.18).

Table 1.

Descriptives and Partial Correlations for Child Internalizing Problems, Coping, and Subjective and Objective Stress Response Indices

		1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.	13.	14.	15.	16.	17.	18.
1.	IP-C	—
2.	IP-P	.28*	—
3.	EC-C	−.71*	−.16	—
4.	DC-C	.31*	−.09	−.52*	—
5.	ISR-C	.69*	.22	−.94*	.18*	—
6.	Conditiona	.04	.02	−.04	−.05	.06	—
7.	SC – T1	−.16	−.16	.07	.20*	−.17	−.04	—
8.	SC – T2	−.24*	−.20*	.22*	.05	−.28*	−.02	.68*	—
9.	SC – T3	−.16	−.21*	.15	−.02	−.17	.10	.50*	.83*	—
10.	SC – T4	.04	−.10	.03	−.03	−.03	.14	.14	.35*	.63*	—
11.	SC – T5	.01	−.11	.01	−.03	−.01	.13	.10	.29*	.57*	.94*	—
12.	SC – T6	−.07	−.15	.05	−.03	−.05	.14	.14	.34*	.62*	.89*	.96*	—
13.	NA – T1	.15	.14	−.14	−.02	.17	.07	.04	.04	.03	−.04	−.07	−.08	—
14.	NA – T2	.22*	.20*	−.25*	.11	.25*	.06	−.05	−.10	−.07	−.07	−.07	−.10	.48*	—
15.	NA – T3	.27*	.31*	−.24*	.12	.22*	.01	−.14	−.19*	−.10	.07	.01	−.08	.22*	.39*	—
16.	NA – T4	.31*	.24*	−.23*	.03	.25*	−.02	−.17	−.27*	−.18*	.04	.02	−.01	.42*	.50*	.43*	—
17.	NA – T5	.36*	.18*	−.31*	.16	.29*	−.02	−.12	−.19*	−.12	.06	.02	−.01	.34*	.47*	.32*	.74*	—
18.	NA – T6	.39*	.16	−.27*	.12	.25*	.05	−.06	−.09	−.07	.08	.05	−.01	.28*	.51*	.30*	.51*	.60*	—
M		45.56	53.04	0.46	0.15	0.39	0.52	0.08	0.08	0.09	0.14	0.14	0.11	3.00	2.63	4.00	3.05	2.64	2.20
SD		7.16	11.71	0.07	0.03	0.06	0.50	0.06	0.07	0.07	0.12	0.17	0.15	1.02	0.82	1.68	1.28	1.07	0.61
Min		36.00	31.00	0.27	0.09	0.24	0.00	0.01	0.01	0.01	0.01	0.01	0.01	1.00	2.00	1.00	2.00	2.00	2.00
Max		81.00	92.00	0.65	0.24	0.57	1.00	0.36	0.44	0.50	1.02	1.60	1.40	8.00	5.00	9.00	8.00	8.00	5.00

Note. N=151;

^a=Spearman’s Rho; IP=Internalizing Problems; C=Child Report; P=Parent Report; EC=Engagement Coping; DC=Disengagement Coping; ISR=Involuntary Stress Responses; SC=Salivary Cortisol; NA=Negative Affect; T=Time. Condition coded 0 for avoidance and 1 for distraction.

^*p<.05.

The subgroup labeling conventions that follow were informed by an overall synthesis of the pattern of person-entered results for each subgroup. That is, a combination of both a) trajectory distinction analyses (see Aim 1 Results) and b) multinomial logistic regression analyses that helped to differentiate the subgroups (see Aim 2 Results) guided naming of the subgroups. While we introduce the subgroup labels here in the Results section, this overall synthesis and thus justification of the labels for each subgroup follows in the Discussion.

Aim 1: Are there Subgroups of Preadolescents with Qualitatively Distinct Patterns of Psychobiological Stress Response Correspondence?

MTM parameter estimates and adequacy indices are shown in Table 2, alongside the results of our trajectory distinction analyses (i.e., differing superscripts) that helped label the subgroups. MTM revealed four subgroups (Figure 1). The log Bayes factor comparing the two- to one-group [2log_e(B₁₀)≈261.14], the three- to two-group [2log_e(B₁₀)≈155.76], and the four- to three-group [2log_e(B₁₀)≈43.72] solutions provided strong evidence for the four-group solution (Nagin et al., 2018), with MTM adequacy indices suggesting the final model fit the data well. The In-Touch group (Figure 1A) was consistent with our expected Moderate NA–Moderate SC group. The In-Touch group included 65 children (43.0%) and was characterized by higher psychobiological correspondence. Specifically, In-Touch children displayed moderate NA and SC baseline levels as well as moderately pronounced NA and SC reactivity patterning. Because the In-Touch group was largest and potentially reflected more healthy psychobiological stress response correspondence (Campbell & Ehlert, 2012), it was used as reference in trajectory distinction analyses¹. Three subgroups reflecting lower psychobiological stress response correspondence emerged. The Unmindful group (Figure 1B) displayed an unexpected pattern of psychobiological correspondence. The Unmindful group included 49 children (32.5%) and was characterized by moderate NA and SC baseline levels, but moderately pronounced NA reactivity patterning paired with less pronounced SC reactivity patterning. The Vigilant group (Figure 1C) was consistent with our expected High NA–Low SC group. The Vigilant group included 13 children (8.6%) and was characterized by high NA baseline levels but low SC baseline levels, as well as more pronounced NA reactivity patterning but moderately pronounced SC reactivity patterning. The Denial group (Figure 1D) was consistent with our expected Low NA–High SC group. The Denial group included 24 children (15.9%) and was characterized by moderate NA baseline levels but high SC baseline levels, as well as less pronounced NA reactivity patterning but also more pronounced SC reactivity patterning.

Table 2.

Parameter Estimates (Standard Errors) and Model Adequacy Indices for Final Multitrajectory Modeling Four-Group Solution

	Negative Affect (NA)	Salivary Cortisol (SC)	AvePP j	OCC j	Prob j	Prop j	Ratio
In-Touch (n=65)			.931	40.545	.427	.430	0.993
Intercept	3.172* (0.167) A	0.532* (0.017) A
Linear	−0.748* (0.114)	−0.001* (0.002)
Quadratic	0.037* (0.006)	0.001* (0.001)
Cubic	−0.001* (0.001)	−0.001* (0.001) a
Quartic	0.001* (0.001) a	-
Unmindful (n=49)			.957	67.238	.332	.325	1.022
Intercept	2.684* (0.159) A	0.478* (0.012) A
Linear	−0.459* (0.112)	−0.004* (0.001)
Quadratic	0.001* (0.006)	0.001* (0.001)
Cubic	−0.001* (0.001)	−0.001* (0.001) b
Quartic	0.001* (0.001) a	-
Vigilant (n=13)			.972	104.424	.086	.086	1.000
Intercept	4.677* (0.071) B	0.448* (0.017) B
Linear	−0.205* (0.053)	−0.013* (0.003)
Quadratic	0.008* (0.003)	0.001* (0.001)
Cubic	−0.001* (0.001) b	−0.001* (0.001) a
Denial (n=24)			.935	43.029	.155	.159	0.978
Intercept	2.902* (0.064) A	0.600* (0.020) C
Linear	−0.048* (0.046)	−0.009* (0.002)
Quadratic	0.007* (0.002)	0.001* (0.001)
Cubic	−0.001* (0.001) c	−0.001* (0.001) c

Note. AvePP_j=Average posterior probability; OCC_j=Odds of correct classification; Prob_j=Probability of group assignment; Prop_j=Proportion of children assigned to each group; Ratio=Ratio of Prob_j to Prop_j. Upper- and lower-case superscripts denote significant differences in intercept and polynomial estimates, respectively, within the same stress response index.

^*p<.05.

Figure 1.

Negative affect and salivary cortisol response trajectories to the Trier Social Stress Test (TSST) for the final four-group solution. Reverse fourth root transformed values presented for visual clarity and ease of interpretation.

Aim 2: Are Preadolescent Internalizing Problems and Coping Meaningfully Associated with Subgroup Membership?

Multinomial logistic regression estimates are shown in Tables 3 and 4. Our covariate model was not significant, X²(12)=19.650, p=.074. Gender, X²(3)=5.357, p=.15, age, X²(3)=4.576, p=.21, puberty, X²(3)=3.026, p>.25), and medication use, X²(3)=0.846, p>.25, were not associated with subgroup membership. Our self-reported, X²(15)=37.881, p<.001, and parent-reported, X²(15)=33.453, p=.004, internalizing problem models were significant. Vigilant children (relative to In-Touch) were more likely to present with both self-reported (B=0.221, SE=0.059, p<.001) and parent-reported (B=0.077, SE=0.029, p=.007) internalizing problems. Unmindful children (relative to In-Touch) were more likely to present with parent-reported (B=0.045, SE=0.019, p=.018), but not self-reported (B=0.036, SE=0.033, p>.25) internalizing problems. Membership in the Denial subgroup (relative to In-Touch) was not related to self- or parent-reported internalizing problems.

Table 3.

Parameter Estimates (Standard Errors) for Multinomial Logistic Regressions Predicting MTM Subgroup Membership

Model	Internalizing Problems Child Report			Internalizing Problems Parent Report			Engagement Coping Child Report			Involuntary Stress Responses Child Report
Comparison Group	Denial	Unmind-ful	Vigilant	Denial	Unmind-ful	Vigilant	Denial	Unmind-ful	Vigilant	Denial	Unmind-ful	Vigilant
Gender	0.241 (0.664)	−0.662 (0.459)	−2.164* (1.058)	0.117 (0.678)	−0.591 (0.464)	−1.649 (0.948)	0.280 (0.665)	−0.606 (0.456)	−1.857 (0.958)	0.263 (0.668)	−0.611 (0.455)	−1.832 (0.969)
Age	0.002 (0.001)	0.001 (0.001)	−0.002 (0.002)	0.002 (0.001)	0.001 (0.001)	−0.001 (0.002)	0.002 (0.001)	0.001 (0.001)	−0.002 (0.002)	0.002 (0.001)	0.001 (0.001)	−0.002 (0.002)
Puberty	0.868 (0.614)	0.139 (0.458)	1.476 (0.866)	0.850 (0.612)	0.001 (0.472)	0.632 (0.745)	0.882 (0.617)	0.119 (0.458)	1.027 (0.767)	0.863 (0.614)	0.114 (0.457)	1.035 (0.720)
Medication	0.161 (0.217)	0.109 (0.173)	0.131 (0.355)	0.169 (0.217)	0.127 (0.176)	0.067 (0.329)	0.177 (0.221)	0.123 (0.176)	−0.024 (0.335)	0.195 (0.223)	0.121 (0.177)	−0.062 (0.337)
Behavior	0.015 (0.045)	0.036 (0.033)	0.221* (0.059)	−0.021 (0.030)	0.045* (0.019)	0.077* (0.029)	0.948 (3.813)	−0.082 (2.818)	−16.599* (6.358)	−2.935 (4.351)	0.270 (3.278)	21.299* (7.590)
X2 (df)	37.881* (15)			33.453* (15)			29.015* (15)			31.785* (15)
Nagelkerke’s R2	.260			.235			.205			.223

Note. Child gender coded 0 for boys and 1 for girls. The In-Touch subgroup served as the reference group.

^*p<.05.

Table 4.

Parameter Estimates (Standard Errors) for Multinomial Logistic Regressions Predicting MTM Subgroup Membership

Model	Avoidance Coping Child Report			Denial Coping Child Report			Wishful Thinking Coping Child Report
Comparison Group	Denial	Unmindful	Vigilant	Denial	Unmindful	Vigilant	Denial	Unmindful	Vigilant
Gender	0.287 (0.668)	−0.564 (0.458)	−1.387 (0.929)	0.336 (0.673)	−0.581 (0.454)	−1.582 (0.923)	0.327 (0.664)	−0.606 (0.458)	−1.806 (0.950)
Age	0.002 (0.001)	0.001 (0.001)	−0.001 (0.002)	0.003* (0.001)	0.001 (0.001)	−0.001 (0.002)	0.002 (0.001)	0.001 (0.001)	−0.001 (0.002)
Puberty	0.877 (0.625)	0.153 (0.464)	0.931 (0.728)	0.887 (0.629)	0.166 (0.471)	0.768 (0.736)	0.864* (0.625)	0.135 (0.465)	0.875 (0.714)
Medication	0.168 (0.219)	0.125 (0.174)	0.065 (0.322)	0.232 (0.229)	0.133 (0.176)	0.031 (0.318)	0.154 (0.222)	0.126 (0.175)	0.127 (0.328)
Behavior	2.139 (19.931)	−7.966 (15.538)	−43.535 (30.574)	64.559* (25.632)	5.929 (17.939)	−22.302 (31.123)	−9.527 (18.955)	2.051 (13.760)	49.554* (23.153)
X2 (df)	22.123 (15)			28.533* (15)			25.639* (15)
Nagelkerke’s R2	.160			.202			.184

Note. Child gender coded 0 for boys and 1 for girls. The In-Touch subgroup served as the reference group.

^*p<.05.

Our engagement coping, X²(15)=29.015, p=.016, and involuntary stress response, X²(15)=31.785, p=.007, models were significant. Vigilant children (relative to In-Touch) were less likely to self-report use of engagement coping (B=−16.599, SE=6.358, p=.009) and more likely to present with self-reported involuntary stress responses (B=21.299, SE=7.590, p=.005). Membership in the Unmindful or Denial groups (relative to In-Touch) was not associated with involuntary stress response or engagement coping.

Our disengagement coping model was not significant, X²(15)=20.651, p=.148. As per Flynn and Rudolph (2007), three subtypes of disengagement coping were explored as correlates of subgroup membership. Our denial coping model was significant, X²(15)=28.533, p=.018. Specifically, denial coping increased the likelihood of membership in Denial relative to In-Touch (B=64.559, SE=25.632, p=.012). Our wishful thinking coping model was significant, X²(15)=25.639, p=.042. Specifically, wishful thinking coping increased the likelihood of membership in Vigilant relative to In-Touch (B=49.554, SE=23.153, p=.032). Our avoidance coping model was nonsignificant, X²(15)=22.123, p=.105, and avoidance coping was not associated with subgroup membership X²(3)=2.473, p>.25.

Summary of the Findings

Multitrajectory modeling (MTM) of children’s negative affect (NA) and salivary cortisol (SC) responses to the Trier Social Stress Test (TSST) revealed four subgroups: In-Touch (i.e., higher psychobiological stress correspondence, Moderate NA–Moderate SC; n=65), Unmindful (i.e., lower psychobiological stress correspondence, Moderate NA–Low SC; n=49), Vigilant (i.e., lower psychobiological stress correspondence, High NA–Low SC; n=13), Denial (i.e., lower psychobiological stress correspondence, Low NA–High SC; n=24).

Relative to In-Touch: 1) Unmindful were more likely to present with parent-reported but not self-reported internalizing problems, 2) Vigilant were more likely to present with self- and parent-reported internalizing problems, more likely to report involuntary stress responses, less likely to use engagement coping, and more likely to use wishful thinking coping, 3) Denial self-reported similar levels of internalizing problems, involuntary stress responses, and engagement coping, but also were more likely to report reliance on denial coping.

Discussion

The current study identified unique patterns of psychobiological stress response correspondence that were associated with preadolescent internalizing problems and coping in theoretically meaningful and clinically informative ways. Multi-trajectory modeling (MTM; Nagin et al., 2018) of negative affect and salivary cortisol response levels revealed four distinct subgroups, one reflecting higher (e.g., In-Touch) and three reflecting lower (e.g., Vigilant, Unmindful, Denial) psychobiological correspondence. Identification of these subgroups helps to clarify the relative paucity of significant bivariate associations between negative affect and salivary cortisol levels in both the current study and extant adult literature (Campbell & Ehlert, 2012). Also, the presence of these subgroups suggests that preadolescence may be a period defined by significant heterogeneity in psychobiological correspondence. In the discussion that follows, we synthesize this matrix of person-centered findings for each group and elaborate on implications for incorporation of multi-informant (e.g., self-report, parent-report) and multilevel assessment (e.g., affect, physiology) in preventative interventions targeting children’s risk for internalizing problems.

In-Touch children were less likely to report internalizing problems, avoidant coping skills, and involuntary stress responses. On the contrary, these children tended to rely on engagement coping for managing stress in their day-to-day lives. This constellation of findings is in line with the functionalist perspective (Shirtcliff et al., 2014) as it pertains to healthy psychobiological stress response correspondence (Campbell & Elhert, 2012). For these children, moderate increases in negative affect may help orient them to a novel source of stress (e.g., stressor, cognitive and emotional distress) while attuned corresponding increases in physiological arousal help mobilize resources for evaluating coping options (e.g., problem-solving, distraction) and executing the most suitable strategy. Attunement between affective experience and physiologic arousal may help In-Touch children identify novel sources of stress and marshal resources to efficaciously apply the most appropriate skill. These claims are consonant with both theory and evidence that efficacious coping contributes to children’s healthy adjustment (Spear, 2009; Zimmer-Gembeck & Skinner, 2016).

Conversely, Vigilant children were more likely to present with internalizing problems, avoidant coping, and involuntary stress responses. The matrix of associations for children in this group is consistent with patterns that have been observed of socially anxious youth in particular (Pine et al., 2009). Specifically, joint high negative affect and low physiological arousal has been thought to reflect difficulty with appropriate orienting (e.g., diverting) of attention in the face of stress. For these children, this difficulty may manifest as sustained attention on possible threats in social environments (i.e., hypervigilance), the physiological result of which is habituated basal arousal (Anderson & Hope, 2009). If these previously documented patterns map onto those observed in Vigilant, this might also clarify our Vigilant reactivity findings. That is, socially anxious youth have been shown to display less pronounced physiological response when an actual threat occurs, because the threat matches expectation, requires little further appraisal, and necessitates fewer physiological resources for assessment of coping options (Hoffman et al., 1995). That reliance on wishful thinking in the peer domain (e.g., “I wish problems with peers would just go away.”) and less use of engagement coping (e.g., “I try to think of different ways to fix the problem I have with my peers.”) also characterized Vigilant children is further consonant with these claims, given that the former involves anxiously anticipating threats (Englemann et al., 2019) and passively waiting for them to resolve as opposed to active efforts to change the stressor (Connor-Smith et al., 2000). Longitudinal research is needed to ascertain if wishful thinking confers risk for this unique pattern of psychobiological correspondence and related internalizing difficulties.

Parent-report (e.g., internalizing problems), but not self-report (e.g., internalizing problems, involuntary stress responses, coping), measures detected maladjustment in Unmindful children. Discrepancies between parent- and self-reports of internalizing problems are not uncommon as children age into adolescence, become more independent in their experience with and management of stressors, and also become more adept reporters of their internal states (Jensen et al., 1999; Kolko & Kazdin, 1993). However, when parents report higher internalizing problems than their children, this discrepancy has often been attributed to a child’s lack of insight into and understanding of their internal states and related difficulties communicating internal experiences (Becker et al., 2016). Thus, both lower psychobiological correspondence as well as discrepant parent-child reports observed of these children in particular perhaps provides converging evidence that the Unmindful profile reflects difficulties with lack of awareness of internal states (Campbell & Ehlert, 2012). That self-reports of coping failed to detect maladjustment in these children is also consonant with this observation, insofar as insight into one’s internal states is likely requisite for accurate reporting of how frequent one experiences involuntary stress responses but also uses effortful coping to manage them. If so, one postulation may be that more moderate negative affect, lower internalizing problem, and less frequent involuntary stress response levels reported by Unmindful children are, in actuality, more elevated and predominant. Indeed, clinicians often rely on parent reports for treatment planning and clinical decision making when parent (high)-child (low) discrepancies arise in assessments of preadolescents (Becker et al., 2016). Future research incorporating additional informants (e.g., teacher-reported internalizing problems, experimenter-reported or observed negative affect) is needed to test these claims.

Both self- and parent-report measures failed to detect internalizing problems in Denial children. However, these children also reported that they tend to rely on denial as a primary means of coping (i.e., “I try to believe that problems with peers don’t exist.”, “In front of others, I act like problems with my peers never happened.”). Thus, one possibility may be that the Denial profile reflects a penchant for downplaying negative affect experiences and disclosing lower levels of internal distress to others (i.e., parents, experimenters), particularly when faced with peer-related challenges (e.g., speaking to a new class). Of concern, these children concomitantly exhibited HPA activation patterns similar to those with known links to internalizing psychopathology (e.g. hyperreactivity; Lopez-Duran et al., 2009), patterns that were not better explained by developmental factors (i.e., more pronounced cortisol reactivity associated with age or pubertal status; Gunnar et al., 2009). Longitudinal research is needed to determine whether the Denial profile reflects a form of psychobiological risk for internalizing problems that conventional assessments may fail to notice, with denial-related correspondence deficits reflecting minimization of the existence of problems or distress while, over time, biological risk accumulates “underneath the skin.” This assertion is supported by evidence in adult literature, whereby verbal-autonomic response dissociation in the face stressors explains avoidance–illness linkages (Schwerdtfeger & Kohlmann, 2004).

Strengths, Limitations, and Future Directions

Does preadolescent affective experience correspond with physiological arousal in response to an acute stressor (Campbell & Ehlert, 2012)? Our findings suggest both “yes” (e.g., In-Touch) and “no” (e.g., Vigilant, Unmindful, and Denial). As such, identification and intervention efforts may benefit from multi-informant and multilevel approaches. That is, relying solely on children’s self-report may overlook biological dysregulation that would otherwise signal need for intervention. For children both open to reporting and with greater awareness of their experience of high levels of negative affect, such maladjustment may be more readily detected via traditional self-report measures. However, for children who may tend to report experiencing low negative affect but concomitantly exhibit dysregulated stress physiology (e.g., hyperarousal, Lopez-Duran et al., 2009; hypoarousal; Anderson & Hope, 2009; Pine et al., 2009), such risk may be difficult to detect. Children’s tendency to cope by downplaying their experience of distress (e.g., Denial; Hampel & Petermann, 2005) their lack of awareness of internal states and distress (e.g., Unmindful; Flynn & Rudolph, 2014) may each factor into this difficulty. As knowledge about mechanisms (e.g., HPA) implicated in stress-illness linkages improves, multimodal assessments may help to identify those children who at the surface may appear healthy but “underneath the skin” are developing a biological disposition that increases risk for psychopathology (Kuhlman & Mousavi, 2020).

The current study is not without limitations. First, though our samples size fell within recommended guidelines (e.g., N>100; Nagin 2005), it was still small for a person-centered design. Also, while there was substantial variability with respect to socioeconomic status (e.g., annual household income), the sample was relatively homogenous with respect to race. These factors limit the generalizability of the findings. Second, though common with person-centered studies of multilevel stress response function (e.g., Koss, Cummings, Davies, & Cicchetti, 2017), unequal sample sizes across our identified subgroups may have reduced power to detect correlate effects. Thus, cautious interpretation and replication of our findings is warranted. Third, the results are limited to a sample of predominantly White, suburban and semirural families. Complementary research with ethnically diverse samples exposed to greater environmental risk (e.g., poverty) is needed. Fourth, our study focused solely on self- and parent-reported child internalizing problems, as well as self-reported child responses to stress. Inclusion of parent-report of child coping may have further distinguished the Unmindful group. Future studies might also benefit from inclusion of teacher-reports of child internalizing problems and coping. Fifth, our focus on a single time point limits inference about whether our profiles function as risk factors for internalizing problems. Future longitudinal studies may reveal that Denial risk for internalizing problems emerges over time alongside the accumulation of avoidance-driven HPA dysregulation. Sixth, our study did not assess stress expressions (e.g., facial expressions, body posture). Person-centered study across experience, expression, and physiological arousal levels of analysis may further inform identification efforts (i.e., observations of children’s affect in response to stressors).