Promoting Empathy and Affiliation in Relationships (PEAR) study: protocol for a longitudinal study investigating the development of early childhood callous-unemotional traits

Abstract

Introduction

Children with callous-unemotional (CU) traits are at high lifetime risk of antisocial behaviour. Low affiliation (ie, social bonding difficulties) and fearlessness (ie, low threat sensitivity) are proposed risk factors for CU traits. Parenting practices (eg, harshness and low warmth) also predict risk for CU traits. However, few studies in early childhood have identified attentional or physiological markers of low affiliation and fearlessness. Moreover, no studies have tested whether parenting practices are underpinned by low affiliation or fearlessness shared by parents, which could further shape parent–child interactions and exacerbate risk for CU traits. Addressing these questions will inform knowledge of how CU traits develop and isolate novel parent and child targets for future specialised treatments for CU traits.

Methods and analysis

The Promoting Empathy and Affiliation in Relationships (PEAR) study aims to establish risk factors for CU traits in children aged 3–6 years. The PEAR study will recruit 500 parent–child dyads from two metropolitan areas of the USA. Parents and children will complete questionnaires, computer tasks and observational assessments, alongside collection of eye-tracking and physiological data, when children are aged 3–4 (time 1) and 5–6 (time 2) years. The moderating roles of child sex, race and ethnicity, family and neighbourhood disadvantage, and parental psychopathology will also be assessed. Study aims will be addressed using structural equation modelling, which will allow for flexible characterisation of low affiliation, fearlessness and parenting practices as risk factors for CU traits across multiple domains.

Ethics and dissemination

Ethical approval was granted by Boston University (#6158E) and the University of Pennsylvania (#850638). Results will be disseminated through conferences and open-access publications. All study and task materials will be made freely available on lab websites and through the Open Science Framework (OSF).

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The Promoting Empathy and Affiliation in Relationships (PEAR) study is designed to assess the development of callous-unemotional (CU) traits across early childhood.

• The PEAR study includes a range of assessment methods (eg, observation, computer tasks, questionnaire, eye-tracking) to assess core constructs across multiple domains (eg, behaviour, attention, physiology).

• Data collection is longitudinal and allows for the investigation of biobehavioural (parent and child low affiliation and low threat sensitivity) and contextual (ie, parental harshness, low warmth and low emotion scaffolding) risk factors for CU traits, focusing on their interplay over time.

• Although the origins of CU traits can be traced to infancy, the PEAR study begins recruitment at age 3, which balances a focus on early development with feasibly having young children complete computer and observational tasks.

• The study is restricted to parent–child dyads residing in the metropolitan areas of Philadelphia or Boston.

Introduction

Disruptive behaviour is a core feature of oppositional defiant disorder (ODD) and conduct disorder, which are among the most common psychiatric conditions of childhood.^{1 2} Disruptive behaviour disorders (DBDs) cause stress to parents and teachers and vast economic costs via health, legal and school expenditures.^{3 4} Around 10%–50% of children with DBD have callous-unemotional (CU) traits (DBD+CU), defined by callousness, uncaring and remorselessness.^{5 6} CU traits predict risk of violence, psychopathy and arrest, even accounting for DBD severity.^7–9 DBD+CU is more heritable than DBD without CU traits (DBD-only)¹⁰ and associated with distinct neural and behavioural correlates.^{11 12} However, few longitudinal studies have investigated risk factors for CU traits in young children, which limits our ability to develop targeted treatments for DBD+CU beginning early in life. The Promoting Empathy and Affiliation in Relationships (PEAR) study aims to advance knowledge about risk factors for CU traits, with the goal of informing more effective treatments for DBD+CU.

Theoretical framework

The PEAR study draws on seminal theoretical frameworks within developmental psychopathology that leverage knowledge of typical and disrupted trajectories of child development.^13–15 Likewise, the premise for the PEAR study is that the development of CU traits can be understood in the context of multiple biological, psychological and social factors that interact over time. CU traits are similarly proposed to arise from probabilistic interactions of these factors, which gradually consolidate into a distinct, recognisable and diagnosable syndrome.¹⁶ The PEAR study follows this logic by adopting a process-oriented framework representing CU traits as the outcome of a dynamic and multidomain system (figure 1).¹⁵

Figure 1.

A dynamic systems and process-oriented developmental model depicting the development of callous-unemotional (CU) traits in the context of multiple biological, psychological and social factors interacting over time. Note: The Promoting Empathy and Affiliation in Relationships (PEAR) study is guided by longstanding developmental science and process-oriented frameworks (eg, probabilistic epigenetics; developmental psychopathology) that specify complex behavioural outcomes in the context of multiple biological, psychological and social factors interacting over time.^{13 14} In the PEAR study, CU traits are proposed to develop downstream of individual differences in threat sensitivity and affiliation, which are studied dynamically across the interacting systems of physiology, attention, behaviour and parenting. Conceptual figure inspired by Gottlieb.¹⁵

The PEAR study is also guided by the Sensitivity to Threat and Affiliative Reward (STAR) model, which proposes that low affiliation and fearlessness are inherited biobehavioural dimensions that increase risk for CU traits¹⁷ (figure 2). Affiliation is characterised as the motivation for and enjoyment of social closeness.^{18 19} This definition draws on studies that have investigated the biological basis of social bonding^{19 20} and both the interpersonal^{21 22} and neurobehavioural^{23 24} features underlying adult psychopathy. Low affiliation increases risk for CU traits by disrupting children’s initiation and enjoyment of social closeness with others.^{17 24} Likewise, drawing on the adult psychopathy literature^25–27 and developmental models of conscience and moral learning,^{28 29} fearlessness refers to reduced sensitivity to social and non-social threat cues. Fearlessness increases risk for CU traits by disrupting children’s ability to learn about or adaptively respond to negative environmental input that would otherwise provoke behaviour change (eg, others’ distress, punishment).^{30 31}

Figure 2.

Sensitivity to Threat and Affiliative Reward (STAR) model, which conceptualises callous-unemotional (CU) traits as being underpinned by low threat sensitivity/fearlessness and low affiliation. Note: Figure adapted from Waller and Wagner.¹⁷ The PEAR study focuses on assessing risk for CU traits, as represented in the lower left quadrant of the model (ie, low threat sensitivity/fearlessness and low affiliation). However, the STAR model also makes testable predictions about the other quadrants, including the combination of low threat sensitivity and high affiliation combining to produce a phenotype resembling, in its most adaptive form, boldness and extraversion, but at its most maladaptive, harmful levels of social dominance. Likewise, at high levels of fear and low affiliative reward, the model hypothesises a phenotype characterised by social inhibition, feelings of inadequacy and a hypersensitivity to negative evaluation. Finally, at high levels of fear and high levels of affiliation reward, the model hypothesises an interpersonal profile characterised by extreme and pathological dependence, separation distress and need for social relationships whose loss is fear-provoking. PEAR, Promoting Empathy and Affiliation in Relationships.

Multidomain assessment of threat sensitivity and affiliation

Initial support for the STAR model comes from studies documenting links between CU traits and low affiliation and fearlessness using questionnaires^{32 33} or observational tasks.^34–39 CU traits have also been linked to low affiliation and fearlessness using computer tasks, including difficulties recognising fearful, angry, or sad facial or bodily expressions of emotion^40–44 or responding to positive emotions and laughter.^{45 46} Functional MRI studies of older children give insight into the biobehavioural mechanisms underlying CU traits, with tasks tapping into neural processes relevant to the STAR dimensions. For example, CU traits have been linked to reduced amygdala reactivity to fearful faces^{47 48} (ie, presumed to reflect low threat sensitivity and/or affiliation), reduced insula activation to others’ pain⁴⁹ (low threat sensitivity and/or affiliation) and reduced insula activation to laughter (low affiliation).⁵⁰ Finally, blunted physiological arousal to cues of threat or affiliation have been linked to DBD+CU in older children,^51–53 including by studies examining startle responses^{54 55} (ie, low threat sensitivity) and respiratory sinus arrhythmia, which indexes connections between the frontal cortex, amygdala, nucleus solitary tract and inputs to the sinoatrial node as children respond to social stimuli^{56 57} (ie, low affiliation).

This evidence provides initial support for the STAR model, but studies are needed to address several limitations. First, few studies have focused on early childhood, a period when individual differences in the defining features of CU traits first emerge (ie, low empathy and guilt)^58–60 and when interventions to mitigate risk for DBD+CU may have the greatest potential for effectiveness.^{61 62} Second, while some studies suggest that CU traits arise from lower basal physiological functioning and arousal,⁶³ the evidence is inconsistent, potentially reflecting differences in sample age, sample type or assessment context.^64–67 Studies have also largely focused on single regulatory systems, whereas CU traits likely reflect disrupted coordination across systems (eg, sympathetic, parasympathetic, hypothalamus–pituitary–adrenal),^{68 69} which has yet to be investigated within an integrated framework in early childhood. Third, unlike other Research Domain Criteria domains,⁷⁰ few computer tasks exist to assess individual differences in affiliation and fearlessness in young children (eg, adapted for touch screen), which reduces the potential for dissemination in large-scale studies or treatment settings. Fourth, prior work examining attentional biases associated with CU traits has yet to combine eye-tracking with physiological data collection, an approach that could clarify interactions between the biobehavioural mechanisms underpinning CU traits (figure 1). For example, low physiological arousal may be evident even if children attend to relevant emotional stimuli, or instead, could reflect a failure to orient to relevant cues or disengage from non-relevant cues when faced with competing stimuli.⁷¹ Finally, no prior studies have tested specificity in the prediction of CU traits, which undermines knowledge of the unique biobehavioural markers of CU traits and our ability to establish whether sensitivity to threat or affiliation are transdimensional risk factors for other psychiatric disorders (eg, autism spectrum disorder, ODD, social anxiety) or the other quadrants of the STAR model¹⁷ (figure 2).

Parenting influences

We also need fine-grained knowledge about how parenting influences the development of CU traits in early childhood. Parenting exerts an environmental influence on CU traits.^{72 73} In particular, low parental warmth and greater parental harshness predict increases in CU traits across childhood.^73–79 Low parental warmth undermines affiliative parent–child interactions and restricts opportunities for children to experience and develop schemas for empathic and caring behaviour,^{35 80} while parental harshness desensitises children to threat and models aggression as an acceptable interpersonal strategy.⁸¹ For DBD broadly, the most effective parenting interventions involve decreasing harshness and increasing positive reinforcement.^82–84 However, meta-analytical work demonstrates that even after receiving treatments that include a parent training component, DBD+CU children exhibit greater DBD symptom severity than DBD-only children.⁸⁵

To improve treatment outcomes, we need adapted treatments or personalised modules that address the unique socioaffiliative difficulties associated with DBD+CU, including low affiliation or fearlessness in children. However, DBD+CU children may also share such characteristics with their parents, which could shape parent–child interactions in ways that further exacerbate risk for CU traits.^{62 77} For example, associations between parenting and CU traits could reflect passive gene–environment correlations (eg, parents low on warmth and children with CU traits share inherited low affiliation) or evocative gene–environment correlations (eg, fearless children evoke harshness from parents with similar traits).⁸⁶ However, no studies have examined whether parenting predicts CU traits over and above fearlessness or low affiliation measured independently in parents and/or children. In addition, no studies have examined whether parent and child fearlessness and low affiliation interact with parenting to predict CU traits. Moreover, while adaptive physiological regulation in parents has been linked to more effective parenting behaviours,⁸⁷ no studies have investigated the attentional or physiological processes related to fearlessness or affiliation in parents, which could shape their propensity to respond with harshness or warmth, thus influencing the development of CU traits in children. New additions to treatment modules could result from a multimethod investigation of affiliation and fearlessness in parents, including helping parents to better attend to or recognise emotion cues in children (eg, attentional measures) or understand their own responses to emotion cues (eg, physiological measures).

In addition, studies need to examine parental emotion scaffolding, characterised by teaching and supporting children’s emotional understanding and learning, which shapes emotional resonance, regulation and expression.^88–91 Prior studies have linked CU traits to disrupted parental emotion scaffolding, including lower parental acceptance of emotion⁹² and restricted expression of mental state or emotion language.^{93 94} These findings are consistent with developmental studies demonstrating that improvements in children’s emotion understanding predict increases in prosocial and empathic behaviour.^88–90 Alongside evidence that DBD+CU children show difficulties recognising and responding to emotions,^{44 95 96} this research suggests that parental emotion scaffolding represents a critical parenting component to characterise, with promise as a potential treatment target. Thus, studies are needed that explore the main and interactive effects of parental harshness, warmth and emotion scaffolding in relation to CU traits during early childhood. These findings can inform developmental models and guide the creation of more effective treatments for DBD+CU, including teaching parents new techniques (eg, emotion scaffolding skills).

PEAR study aims

The PEAR study is a longitudinal study that will advance knowledge of developmental pathways to CU traits across early childhood (figure 3). Aim 1 of the PEAR study will investigate how low threat sensitivity and low affiliation relate to increases in CU traits across early childhood. Aim 2 will characterise parents’ low affiliation and low threat sensitivity and examine links with parental warmth, emotional scaffolding and harshness. Aim 3 will test interactive associations between parent and child low affiliation and low threat sensitivity, parenting practices and CU traits over time (figure 3). Across aims, the use of multimethod assessments of behaviour, physiology and attention, as well as observational, task and report-based measures, allows for comprehensive, multidomain phenotyping of the STAR constructs, supporting the future development of precision treatments for DBD+CU.

Figure 3.

The Promoting Empathy and Affiliation in Relationships (PEAR) study will examine the interaction between low affiliation and low threat sensitivity in parents and children, as well as with parenting practices, to understand risk for callous-unemotional (CU) traits across early childhood. Note: Aim 1 will investigate how children’s low affiliation and low threat sensitivity at time 1 are related to increases in CU traits across early childhood from time 1 to time 2. Aim 2 will investigate parents’ low affiliation and low threat sensitivity and examine cross-sectional and longitudinal links with parental warmth, emotional scaffolding and harshness at time 1 and time 2, respectively. Aim 3 will test interactive associations between parent and child low affiliation and fearlessness, parenting practices and CU traits over time. All three aims prioritise multimethod assessments of behaviour, physiology and attention, as well as observed, task and report-based measures, which will allow for comprehensive, multidomain phenotyping of the core constructs (see figure 1 and table 2). We will include additional variables in the models to adjust for demographic confounds and/or to address specificity in the prediction of CU traits relative to other dimensions of psychopathology in early childhood (see table 2 and online supplemental materials).

Methods

Study design

The PEAR study involves a longitudinal multisite design at two sites in the US with planned recruitment of 500 parent–child dyads (ie, 250 at each site). Data will be collected during lab visits at time 1 (ages 3–4) and time 2 (ages 5–6). Planned time 1 recruitment is for 550 parent–child dyads, which allows for an estimated 10% attrition rate at time 2 to increase the likelihood of obtaining a final sample of 500 dyads with data at both time points.

Study setting and procedures

Study data will be collected during 2.5–3 hours lab visits at the University of Pennsylvania or Boston University. Visits are divided into data collection blocks: (1) parent completes questionnaires, (2) child and parent complete computer and/or eye-tracking tasks and (3) parent and child complete observational tasks together/alone. Both data collection sites are equipped with identical equipment: (1) Multiple pan-tilt-zoom cameras and microphones integrated with Noldus Observer Software to facilitate coding of observable parent and child behaviour; (2) Biopac MP160 data acquisition and analysis systems with AcqKnowledge V.5 software, allowing for collection of physiological data from parents and children and synchronisation with video recordings; (3) Wireless BioNomadix modules to continuously collect parents’ and children’s electrocardiographic and respiratory data; (4) Wireless BioNomadix module to collect event-related electrodermal data (ie, during seated computer tasks); (5) SR Research Eyelink 1000 to capture eye-tracking data during seated computer tasks and (6) Pupil Invisible mobile eye-tracking glasses from Pupil Labs to collect mobile (ie, ambulatory) eye-tracking data during parent–child interaction tasks. All computer tasks have been built in SR Research Experiment Builder or Psychopy and are administered via high-refresh-rate touchscreen display or computer monitor. Following the visit, all participating families are compensated and provided with mental health resources. Licensed clinicians are available at both sites to provide additional support as needed.

Eligibility criteria

Eligible child participants are 3–4 years at time 1 and living with at least one biological parent who consents to participate, with English spoken <50% of the time at home. We select 50% of the sample as ‘high risk’ based on parental endorsement of five items to assess CU traits: ‘no guilt after misbehaviour’, ‘punishment does not change behaviour’, ‘unresponsive to affection’, ‘shows little affection’ and ‘too little fear’.⁹⁷ Based on prior studies in early childhood,^{97 98} Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM–5; American Psychiatric Association, 2013) criteria for the limited prosocial emotions (CU traits) specifier,^{99 100} and other large-scale studies that have recruited with an oversampling for externalising problems,^{101 102} we designate children as ‘high risk’ if parents endorse two or more of the five items. We also recruit an estimated 2:1 ratio of male to female children to account for higher prevalence of rates of DBD among boys¹⁰³ (table 1).

Table 1.

Summary of planned enrolment estimates based on risk status (high vs low risk for CU traits), child sex and site

	University of Pennsylvania		Boston University		Planned subtotals across sites by sex and age
	Low risk	High risk	Low risk	High risk
Male (age 3 at time 1; age 5 at time 2)	42	42	41	41	166
Female (age 3 at time 1; age 5 at time 2)	20	20	21	21	82
Male (age 4 at time 1; age 6 at time 2)	42	42	42	42	168
Female (age 4 at time 1; age 6 at time 2)	21	21	21	21	84
Planned subtotal within site	125	125	125	125	Anticipated total sample with data at both time points across sites, n=500
Planned total within site	250		250

Estimates reflect planned recruitment of 250 children at both sites (ie, combined 500 children) who will be assessed twice in the lab: time 1 (aged 3–4; years 1–2) and time 2 (aged 5–6; years 3–4). During data collection, each site will assess four families per week (≈16 per month). Note that our actual planned time one recruitment will produce a total sample of 550 (ie, add 10% to each number above at time 1), which allows for an estimated 10% attrition rate at time two and increases the likelihood of a final sample of 500 dyads with data at both time points.

CU, callous-unemotional.

Recruitment

Participants are recruited via established methods. First, our prior collaborative longitudinal work^104–106 and literature reviews^{107 108} show that social media is highly effective for recruitment. We commissioned a professionally designed and family-friendly logo for the PEAR study to support recruitment efforts. The paid functions within social media advertising provide inbuilt filtering features, which can be implemented to flexibly target under-represented groups throughout recruitment. Second, institutionally maintained databases identify individuals who have previously agreed to be contacted about research participation. These databases are generated through departmental support, collaboration across labs, outreach efforts in the community and phone calls or mailings from birth records. Finally, recruitment efforts are bolstered by the location of both labs in major metropolitan areas with vibrant paediatric research communities, including the Children’s Hospital of Philadelphia and the Child and Adolescent Fear and Anxiety Treatment Center at the Center for Anxiety and Related Disorders at Boston University, which have extensive protocols for connecting families with research. When families hear about the study, they are directed to the study website and screened for inclusion using a brief survey (www.thepearstudy.com). The website includes information describing the study and compensation in plain and engaging language. Participants are compensated US$150 at time 1 and US$170 at time 2, with additional incentives to maximise participation, including babysitting for younger siblings, snacks and transportation support as needed.

Measures

Measures are gold-standard assessments of core constructs or newly developed instruments, which were validated through in-person or online pilot studies. We assess affiliation and fearlessness in parents and children across multiple domains, including behavioural, attentional and physiological responses. Parental harshness, warmth and emotion scaffolding are measured using parent report and observer ratings, with an emphasis on adapting coding schemes for observed measures to be culturally sensitive to the intersection of race and culture with parenting strategies and child behaviour.^109–111 Table 2, online supplemental materials and the PEAR study preregistration on the Open Science Framework (https://osf.io/b2rg5/)¹¹² summarise the assessment framework, including full description of methods and measures.

Table 2.

Overview of study measures, including construct, target, method (including multidomain assessment) and variable type in planned analyses

Construct	Target	Method	Multidomain		Study variable
			+ physiology	+ eye-tracking
Psychopathology
CU traits	Child	PQ: ICU,42 CPTI,140 SDQ141			Dependent variable (aims 1, 2, 3)
ODD/CD symptoms	Child	PQ: CPTI,140 CBCL142			Control variables to establish specificity (aims 1, 2, 3)
ADHD symptoms	Child	PQ: CBCL,142 SDQ141
ASD traits	Child+parent	PQ: AQ143
Anxiety/depression	Child+parent	PQ: child, CBCL142; parent, PHQ9,144 GAD7145
Psychopathic traits	Parent	PQ: SRP146 147
STAR model dimensions
Low affiliation	Child+parent	PQ: SRS2,148 STARS149 CT: Emotion Recognition,150–153 Emotion Induction,154 155Social Preference,156 157CAMP OR: Child, SCALA	✓	✓	Independent variables (aim 1, 2)+moderator (aim 3)
Low threat sensitivity	Child+parent	Pq: BIQ158 STARS149 Ct: Visual Search,159 160Emotion Recognition,150–153Emotion Induction154 155 Or: Child, Stranger Approach161–164	✓	✓
Parenting practices
Harshness, warmth, + emotion scaffolding	Parent	PQ: PBACE,165 CEPAQ,166PS167 OR: Storybook, Magnet and Conversation Task101 168	✓	✓	Independent variable (aims 1+2)+moderator (aim 3)
Additional demographic, contextual or individual-level characteristics
Sex, age, race/ethnicity, education	Child+parent	PQ: demographic interview			Control variable to establish specificity (aims 1, 2, 3)
Parent characteristics	Parent	PQ: SUS,169 MSSI,170 PSI,171 QSSI,172 Mini-IPIP,173 DERS-5,174 ACES,175 CTS2-SF176 CT: EF touch, Theory of Mind177 Reward Learning178 179	✓	✓
Child characteristics	Child	PQ: Pediatric-ACES,180 CSPS,181 CCTI,182 CARES,183 BISQ-R184 CT: EF Touch,185 186 Stars in Jars,187 Theory of Mind,177Picture Vocabulary Test188	✓	✓
Neighbourhood+family disadvantage	Child+parent	PQ: demographic Interview, Neighbourhood Risk,189 190 CHAOS,191 CEFIS192 OR: geocoding ZIP code193

See online supplemental materials for detailed description of every measure, including observational tasks, computer tasks and questionnaires. Although some children will turn six by our time two assessment, we will continue to administer the CBCL 1.5–5.5 version to ensure continuity in measurement and since five items established as indexing CU traits in early childhood are not all retained in the CBCL 6–18 version.

ACES, Adverse Childhood Experiences Scale; AQ, Autistic Spectrum Quotient; ASD, Autism Spectrum Disorder; BCSP, Brief Child Sleep Questionnaire; BIQ, Behavioural Inhibition Questionnaire; CAMP, Child Affiliative Motivations and Preferences Task; CARES, Components of Affiliative Reward Experiences Scale; CBCL, Child Behaviour Checklist; CCTI, Colorado Child Temperament Inventory; CD, conduct disorder; CECPQ, Comprehensive Early Childhood Parenting Questionnaire; CEFIS, Coronavirus Disease Exposure Family Impact Scale; CHAOS, Confusion, Hubbub and Order Scale; CPTI, Child Problematic Traits Inventory; CSPS, Child Social Preference Scale; CT, computer task; CTS, Conflict Tactic Scale; CU, callous-unemotional; DERS, Difficulties in Emotion Regulation Scale; EF, Executive Functioning; GAD, Generalised Anxiety Disorder; ICU, Inventory of Callous-Unemotional Traits; Mini IPIP, Mini International Personality Item Pool; MSSI, Maternal Social Support Index; ODD, Oppositional Defiant Disorder; OPS, O’Leary Parenting Scale; OR, observer rating; PBACE, Parents’ Beliefs About Children’s Emotions; PHQ2, Patient Health Questionnaire; PQ, Parent-reported Questionnaire; PSI, Parental Stress Index; QSSI, Quality of Social Support; SCALA, System for Coding Affiliation in Lab Assessments; SRP, Self-Report Psychopathy Scale; SRS2, Social Responsiveness Scale; STARS, Sensitivity to Threat and Affiliative Reward Scale; SUS, Substance Use Screener.

Data management

The Biostatistics and Epidemiology Data Analytics Center at Boston University School of Public Health manages study data using Research Electronic Data Capture data collection tools.^{113 114} Study data also reside inside a secure, centralised and HIPAA-compliant environment. The data are stored in a restricted folder on a secure server to which only authorised PEAR study members have access. The folder is electronically encrypted, with access requiring a Virtual Private Network and two-factor authentication. All actions in the database are logged for data auditing and traceability.

Analytical plan

We will use structural equation modelling (SEM) with robust full information maximum likelihood (FIML) or weighted least square means and variance (WLSMV) estimation¹¹⁵ to address study aims. We will use hierarchical factor models that parse method and construct variance to develop measurement models for core constructs, with alternative data reduction approaches (eg, multiple indicator latent factors, bifactor models) as needed to guide creation of latent variables when we have multiple measures/methods for constructs (eg, fearlessness, affiliation, parenting). Primary analyses will allow us to retain multiple-indicator latent factors, although strategies for estimating factor scores will be implemented if full measurement models appear intractable. We will probe conditional or moderated associations following recommended approaches.^{116 117} We will model non-independence within dyads using multilevel SEM, which allows decomposition of between-dyad and within-dyad influences. False discovery rate corrections will be used to address multiple comparisons.¹¹⁸

Our main hypotheses centre on direct associations between parent and child fearlessness and low affiliation, parenting practices and child CU traits. Aim 1 will be tested by regressing CU traits onto latent fearlessness and affiliation factors. An interaction term between fearlessness and affiliation will be added to test whether the combination of fearlessness and affiliation explains additional variance in CU traits. Aim 1 will be tested cross-sectionally (ie, within time 1 or 2) and longitudinally by regressing CU traits at time 2 onto predictors at time 1, accounting for autoregressive effects. Aim 2 will be addressed by regressing multimethod factors of parenting (warmth, harshness and emotion scaffolding) onto parents’ affiliation and fearlessness. Aim 2 will also be tested cross-sectionally and longitudinally. A series of path models will be used to address aim 3. First, to explore child–parent evocative effects, we will test associations between child fearlessness at time 1 and parental harshness at time 2 and between child affiliation at time 1 and parental warmth and emotion scaffolding at time 2 within a correlated dependent variables model, accounting for autoregressive relations. Second, to examine parent–child effects controlling for passive gene–environment correlations, we will test the main effects of parental harshness, warmth and emotion scaffolding at time 1 in the prediction of child CU traits at time 2, accounting for child and parent fearlessness and low affiliation. Third, to examine potential dyadic interactive effects, we will separately test various two-way interactions between parent and child fearlessness, low affiliation and parenting in the prediction of CU traits at time 2, accounting for autoregressive effects.

Power calculation

Statistical power was determined using Monte Carlo simulation studies, which specified multilevel simultaneous equation frameworks that accommodate covarying outcomes. All simulation studies included a population generating model of N=500, assumed a type 1 error rate of 0.05, and involved 5000 replications.¹¹⁹ Following established recommendations,¹¹⁹ each simulation specified small and medium direct effects as R²=0.02 and 0.13, respectively.¹²⁰ Each Monte Carlo study specified main effects and interactions on one outcome or multiple correlated outcomes. Results from simulations using bootstrapped standard errors to determine statistical significance at a 0.05 level (N=500) indicated power of 0.99 to detect medium-sized main effects (eg, affiliation and fearlessness to CU traits), and power of 0.93 to detect medium-sized interactive effects (ie, moderation). Results indicated a power estimate of 0.89 to detect joint contributions of parenting practices, parent and child affiliation and fearlessness, and child CU traits from time 1 to time 2. Across all models, power ≥0.80 was retained to detect small-sized to medium-sized effects (R²≈0.02–0.10). With a sample size of 500, we retain a power of 0.80 to detect bivariate correlations |r|≥0.125, which corresponds to a small-sized to medium-sized effect.¹²⁰

Management of bias

Various strategies are used to minimise methodological bias. First, to reduce attrition of participants, established retention strategies for longitudinal studies will be applied,¹²¹ including ensuring flexible and advanced scheduling, recruiting research staff who are diverse in race and ethnicity, sending visit reminders, providing positive inducements (eg, babysitting), sending birthday cards and newsletters, and offering flexible solutions to support transportation (eg, Uber). Second, to reduce confounder bias, socioeconomic and demographic factors will be statistically adjusted in analyses. Contingent relationships between study variables based on these factors (ie, moderation) will also be tested in exploratory models. Third, missing data patterns will be handled using FIML or WLSMV estimation as relevant, both of which represent best practices for accommodating missing or unbalanced data.¹²² Finally, procedural equivalence is maximised through identical equipment, jointly developed study procedures, and weekly meetings between staff at both sites. In addition, we ran joint training sessions across sites, study coordinators conduct weekly reviews of videos of cross-site visits, and principal investigators engage in reciprocal site visits. During analyses, multigroup modelling will establish measurement invariance for study variables across sites.¹²³

Patient and public involvement

There was no patient or public involvement in the design of the PEAR study.

Ethics and dissemination

The PEAR study operates under a single Institutional Review Board that oversees data collection and modifications (Boston University, #6158E (IRB of record); University of Pennsylvania, #850638). We obtained consent from parents using electronic signatures. Minimal risk/distress to participants is anticipated, but contact numbers for counselling services are provided to families following completion of study visits. Findings will be disseminated through peer-reviewed journals (open access where feasible), conferences, professional associations and public mental health services that treat DBD. Findings will be presented in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology statement.¹²⁴ Finally, the study website and social media will also be used to disseminate results once recruitment ends.

Discussion

DBDs cause harm to children’s well-being, suffering to families and communities, and vast economic costs to society. CU traits designate children at high risk for developing DBD and who end treatment for DBD with greater symptom severity.⁸⁵ We need studies that begin early in life to identify modifiable risk factors associated with the development of CU traits. The PEAR study adopts a prospective longitudinal design that will advance knowledge about the development of CU traits using a multimethod approach that combines assessment of behaviour, attention and physiology.

The PEAR study has several limitations. First, while we focus on CU traits, other risk factors for DBD include disinhibition^{125 126} and executive function difficulties.^{127 128} We include measures to assess these constructs (table 2). However, our study focuses on threat sensitivity, affiliative processes and parenting specifically in relation to the development of CU traits. Second, we focus on parenting practices because parents represent the most proximal environmental influence on children, particularly in early childhood.¹²⁹ However, more distal environmental factors also impact risk for psychopathology, including disorganisation, instability in the home and neighbourhood disadvantage.^130–132 Brief measures of these factors are included. However, it is outside the scope of the study to assess these constructs with the same depth as the STAR dimensions. Third, our data collection targets early childhood, when individual differences in the defining features of CU traits are reliably measurable (ie, low empathy and guilt).^58–60 However, the developmental origins of these processes can be traced to infancy,¹³³ with some evidence for differential pathways between early fearful behavioural and physiological profiles and risk for CU traits based on environmental context.^{63 134} Focusing on 3 years and 4 years balances, a need to better understand early risk factors for DBD+CU with feasibly being able to collect physiological and attentional data from young children in response to multiple social, emotional and affiliative cues. Fourth, our measurement of CU traits in the proposed study is derived only from parent report. Follow-up studies of our cohort are necessary to leverage reports from other informants (eg, educator, teacher) or methods (eg, observation of prosocial or empathic conduct in naturalistic settings) to gain insight into the pervasiveness of CU traits across contexts. Finally, the STAR model specifies that individual differences across the full spectrum of affiliation and threat sensitivity are important for conceptualising risk for different forms of psychopathology (eg, pathological dependence when threat sensitivity and affiliation are both high; figure 2). The PEAR study focuses on the quadrant of low affiliation and low threat sensitivity to characterise risk for CU traits, but future studies are needed to explore its predictive validity in relation to other personality or psychiatric disorders assessed dimensionally.^135–138

To formulate comprehensive aetiological models of CU traits and develop targeted early interventions, we need to characterise the organisation and interaction of multiple biological and social influences early in life.^{14 17 63} Longitudinal studies that pair observational, task and report-based measures with assessments of physiology and attention can establish multidomain operationalisations of fearlessness and affiliation to advance knowledge of the biobehavioural basis of CU traits in early childhood. The PEAR study addresses these needs by combining a process-oriented, multidomain approach from developmental psychopathology^13–15 with the substantive predictions of the STAR model.¹⁷ The PEAR study aims to generate novel insights about how low affiliation, fearlessness and parenting dynamically influence the development of CU traits over time. The urgency and potential societal impact of these efforts is underscored by the staggering personal and financial costs incurred by the lifetime consequences of DBD+CU, including violence, crime and incarceration.¹³⁹

Ethics statements

Patient consent for publication

Not applicable.