Introduction
Rates of self-injurious thoughts and behaviors (SITBs) increase dramatically across adolescence (Nock et al., 2008, Reference Nock, Borges and Ono2012), with approximately 23% of high school-aged youth endorsing suicidal ideation (i.e., SI; Orri et al., Reference Orri, Scardera, Perret, Bolanis, Temcheff, Séguin, Boivin, Turecki, Tremblay, Côté and Geoffroy2020) or nonsuicidal self-injury (i.e., NSSI; Brown & Plener, Reference Brown and Plener2017; CDC, 2019; Peterson et al., Reference Peterson, Freedenthal, Sheldon and Andersen2008). Developmental trajectories of SITBs often include onset of, and transitions among, multiple types of SITBs, including both SI and NSSI. However, studies frequently examine singular SITB outcomes or otherwise consider SITBs as a single construct, making it difficult to identify whether, and which, risk pathways may be specific to particular SITBs. Identifying risk pathways for distinct SITBs in adolescence may aid in understanding the dynamic, potentially transactional developmental progression of SITBs during this period (Oppenheimer et al., Reference Oppenheimer, Stone and Hankin2018).
Interpersonal stress and SITBs in adolescence
Epidemiological data support the relevance of developmental psychopathology approaches for understanding increases in SITBs in adolescence. Consistent with the principle that risk factors may have different effects on functioning based on developmental phase (Cicchetti & Rogosch, Reference Cicchetti and Rogosch2002), adolescence is characterized by particular changes in socioaffective and biological domains that may increase risk for SITBs during this period. Recent theoretical work posits that adolescent SITBs may occur in the context of failures in stress regulation among adolescents with atypical psychophysiological stress responses, perhaps particularly in response to interpersonally themed stress (Miller & Prinstein, Reference Miller and Prinstein2019). Indeed, normative social changes in adolescence are associated with increases in interpersonal stress across both familial and peer domains (Collins & Laursen, Reference Collins and Laursen2004; Rudolph, Reference Rudolph, Eds. and Rudolph2014), and these experiences, such as peer victimization and poor family cohesion, are associated with both nonsuicidal and suicidal SITB outcomes (King & Merchant, Reference King and Merchant2008; Massing-Schaffer et al., Reference Massing-Schaffer, Helms, Rudolph, Slavich, Hastings, Giletta, Nock and Prinstein2019; Valencia-Agudo et al., Reference Valencia-Agudo, Burcher, Ezpeleta and Kramer2018).
There is particularly strong evidence for the impact of peer-related stressors on risk for adolescent SITBs. Experiences of peer-related stress such as low peer support, poor quality peer relationships, and peer victimization are associated with both NSSI and SI (Madjar et al., Reference Madjar, Ben Shabat, Elia, Fellner, Rehavi, Rubin, Segal and Shoval2017; Tatnell et al., Reference Tatnell, Kelada, Hasking and Martin2014; van Geel et al., Reference van Geel, Goemans and Vedder2015; for a review, see Cheek et al., Reference Cheek, Reiter-Lavery and Goldston2020). While more limited, studies examining family stress have also found associations between lower perceived family support or belongingness and NSSI (Tatnell et al., Reference Tatnell, Kelada, Hasking and Martin2014) and SI (Glenn et al., Reference Glenn, Kleiman, Kandlur, Esposito and Liu2022), although family conflict may be more robustly associated with SITBs in younger, preadolescent samples (i.e., 9- and 10-year-olds; DeVille et al., Reference DeVille, Whalen, Breslin, Morris, Khalsa, Paulus and Barch2020). When examined simultaneously, peer stress (e.g., peer victimization) has been shown to be a stronger predictor of adolescent SITBs than parent or family relational factors (Victor et al., Reference Victor, Hipwell, Stepp and Scott2019). Indeed, interpersonal stress in these two environments may differentially impact emotional and behavioral functioning in adolescence, as these environments become increasingly separate (Brown, Reference Brown, Lerner and Steinberg2013) and the peer environment becomes more salient (Nelson et al., Reference Nelson, Jarcho and Guyer2016; Somerville, Reference Somerville2013; van den Bos, Reference van den Bos2013). Few studies, however, have directly compared experiences of interpersonal stress within peer versus family contexts in relation to distinct SITB outcomes, which may clarify the extent to which interpersonally themed risk for SITBs in adolescence may be particularly elevated when stress is experienced in specific domains (i.e., the peer environment).
Physiological stress reactivity and SITBs
While many adolescents experience significant interpersonal stress and conflict, only a subset go on to experience or engage in SITBs. Adolescence is a period characterized by increased physiological reactivity to stressors compared to childhood and adulthood (Gunnar & Quevedo, Reference Gunnar and Quevedo2007; Stroud et al., Reference Stroud, Foster, Papandonatos, Handwerger, Granger, Kivlighan and Niaura2009). Dysregulated physiological arousal and regulatory processes when combined with stressful interpersonal relationships may help explain increases in rates of SITBs during adolescence. Heart rate (HR), measured in beats per minute (bpm), is predominantly influenced by coordination of the sympathetic and parasympathetic branches of the ANS, which activates when an organism faces environmental threat and challenge, including psychosocial stress (Cacioppo et al., Reference Cacioppo, Tassinary and Berntson2017). Cardiovascular dysfunction, as reflected by higher cardiac arousal at rest (e.g., high resting HR; see Deutz et al., Reference Deutz, Woltering, Vossen, Deković, van Baar and Prinzie2019), has been proposed to be a putative mechanism associated not only with morbidity and mortality (Khan et al., Reference Khan, Kunutsor, Kalogeropoulos, Georgiopoulou, Newman, Harris, Bibbins-Domingo, Kauhanen, Gheorghiade, Fonarow, Kritchevsky, Laukkanen and Butler2015; Qiu et al., Reference Qiu, Cai, Sun, Li, Zuegel, Steinacker and Schumann2017; Zhang et al., Reference Zhang, Shen and Qi2015), but also with a range of psychiatric disorders (Alvares et al., Reference Alvares, Quintana, Hickie and Guastella2015; Clamor et al., Reference Clamor, Hartmann, Köther, Otte, Moritz and Lincoln2014; Kandola et al., Reference Kandola, Ashdown-Franks, Stubbs, Osborn and Hayes2019; Kemp, Brunoni, et al., Reference Kemp, Brunoni, Santos, Nunes, Dantas, De Figueiredo, Pereira, Ribeiro, Mill, Andreão, Thayer, Benseñor and Lotufo2014; Kemp, Quintana, et al., Reference Kemp, Quintana, Quinn, Hopkinson and Harris2014; Latvala et al., Reference Latvala, Kuja-Halkola, Rück, D’Onofrio, Jernberg, Almqvist, Mataix-Cols, Larsson and Lichtenstein2016; Nelson et al., Reference Nelson, Byrne, Sheeber and Allen2017, Reference Nelson, Sheeber, Pfeifer and Allen2020, Reference Nelson, Flannery, Flournoy, Duell, Prinstein and Telzer2022; Paulus et al., Reference Paulus, Argo and Egge2013) , including SITB (Kaess et al., Reference Kaess, Hooley, Klimes-Dougan, Koenig, Plener, Reichl and Cullen2021). Higher resting HR during late adolescence has been shown to be associated with higher risk for adulthood internalizing disorders in a prospective study of over 1 million males (Latvala et al., Reference Latvala, Kuja-Halkola, Rück, D’Onofrio, Jernberg, Almqvist, Mataix-Cols, Larsson and Lichtenstein2016).
There is also evidence linking cardiovascular functioning, including both HR and heart rate variability (HRV), and SITBs (for a review, see Kang et al., Reference Kang, Patriquin, Nguyen, Oh, Rufino, Storch and Najafi2020). Among adults, research indicates a direct association between resting HR and completed suicide, independent of depressed mood (Lemogne et al., Reference Lemogne, Thomas, Consoli, Pannier, Jégo and Danchin2011). Similar cardiovascular patterns (e.g., higher resting HR, lower HRV) are associated with higher suicide risk in large nonclinical and clinical samples with diverse psychiatric presentations (Lee et al., Reference Lee, Baek, Cho and Hong2021; Chang et al., Reference Chang, Bjørngaard, Tsai, Bjerkeset, Wen, Yip and Gunnell2016) and may be associated with SITBs above and beyond covariates such as psychiatric comorbidities (e.g., depression; Chang et al., Reference Chang, Chang, Chen, Kuo, Lu and Huang2012; Tsypes et al., Reference Tsypes, James, Woody, Feurer, Kudinova and Gibb2018). While studies in adolescents are fewer, one study found evidence for lower baseline HRV, and greater HRV reactivity during a negative mood induction, among adolescents with histories of self-harm (i.e., including nonsuicidal and/or suicidal self-harm; Crowell et al., Reference Crowell, Beauchaine, McCauley, Smith, Stevens and Sylvers2005). In contrast, similar alterations in cardiac functioning have not been found in other adolescent samples (i.e., youth with NSSI histories; Koenig et al., Reference Koenig, Rinnewitz, Parzer, Resch, Thayer and Kaess2017). Further research is needed to establish whether cardiovascular dysfunction is related to risk for SITBs, including prospectively, and whether these patterns may be specific to certain SITB outcomes (e.g., self-injurious thoughts vs. behaviors, with or without suicidal intent).
Several additional studies in adolescents provide further support for associations between alterations in stress responses, including resting cardiac rate, and risk for psychopathology. Higher resting heart rate was found to be associated with greater internalizing symptoms in a sample of over 5,000 adolescents (Nelson et al., Reference Nelson, Flannery, Flournoy, Duell, Prinstein and Telzer2022), and related research shows that disrupted hypothalamic–pituitary–adrenal (HPA) axis responses during adolescence are associated with psychopathology, including SITB outcomes (e.g., Eisenlohr-Moul et al., Reference Eisenlohr-Moul, Miller, Giletta, Hastings, Rudolph, Nock and Prinstein2018; van Heeringen, Reference van Heeringen2012; van Heeringen & Mann, Reference van Heeringen and Mann2014). Of note, recent prospective studies show that adolescents’ physiological responses to interpersonally themed stress are associated with future SITBs (i.e., 9–18 months later; Giletta et al., Reference Giletta, Hastings, Rudolph, Bauer, Nock and Prinstein2017; Massing-Schaffer et al., Reference Massing-Schaffer, Helms, Rudolph, Slavich, Hastings, Giletta, Nock and Prinstein2019; Miller et al., Reference Miller, Eisenlohr-Moul, Giletta, Hastings, Rudolph, Nock and Prinstein2017). In the context of heightened social stress, cardiac regulation (i.e., resting HR) may potentiate risk for SITBs via changes in affect or emotion regulation, as prefrontal and subcortical brain regions that subserve emotional responsivity and regulation also control cardiac responses (Lemogne et al., Reference Lemogne, Thomas, Consoli, Pannier, Jégo and Danchin2011) and undergo dramatic structural and functional alterations during adolescence (Dahl et al., Reference Dahl, Allen, Wilbrecht and Suleiman2018).
Finally, understanding who engages in SITBs likely involves multiple interacting constructs, such that dysfunctions or disruptions across several processes are implicated in emergence of NSSI or SI. Building on developmental psychopathology perspectives, which emphasize transactions between a person and their environment, physiological (i.e., cardiovascular) dysfunction may predict heightened SITB risk only among adolescents experiencing elevated stress in one (e.g., peer or family) or multiple (e.g., peer and family) interpersonal environments. Additionally, in line with the National Institutes of Health Research Domain Criteria (RDoC; NIH, 2016) framework, examination of transdiagnostic processes assessed across multiple units of analysis (e.g., physiology, self-report) is needed to advance understanding of SITB risk (Glenn et al., Reference Glenn, Cha, Kleiman and Nock2017, Reference Glenn, Kleiman, Cha, Deming, Franklin and Nock2018). Alterations in socioemotional and physiological stress responses in adolescence overlap with RDoC’s ‘social processes’ and ‘arousal/regulatory systems’ domains and offer opportunity to test interactions among social and physiological constructs measured across self-report and objective (i.e., physiology) methods.
The current study
The current prospective longitudinal study tested whether adolescents experiencing socially themed or affiliative stress and heightened baseline arousal/regulation (i.e., cardiac) processes are at elevated risk for future SITBs. Given normative increases in socioemotional sensitivity and recalibration of stress–response systems in adolescence, which may increase vulnerability to stressful experiences, we hypothesized that baseline levels of cardiac arousal/regulation would moderate prospective associations between both peer and family affiliative conflict and SITBs (i.e., SI and/or NSSI). Specifically, we hypothesized that neither greater affiliative conflict nor higher resting HR alone would predict SITBs, but that the combination of these interpersonal and physiological factors would predict greater frequency of SI and/or NSSI over 1-year follow-up. We had no a priori hypotheses regarding differential prediction of SI versus NSSI, as prior work examining cardiovascular functioning and SITBs in adolescence is limited and has yielded mixed findings when examining nonsuicidal and suicidal SITB outcomes (Crowell et al., Reference Crowell, Beauchaine, McCauley, Smith, Stevens and Sylvers2005; Koenig et al., Reference Koenig, Rinnewitz, Parzer, Resch, Thayer and Kaess2017).
Method
Participants
The current study included data from a large scale longitudinal research project conducted from 2017 to 2021. Secondary data analysis for the current study came from a total of 147 adolescents (72 males, 48.98%) who completed two sessions of data across 1 year (session 1: M age = 12.34, SD = 0.58; session 2: M age = 12.8, SD = 0.53) and 116 adolescents completing data at 2 year (session 3: M age = aged 13.7, SD = 0.53; due to attrition; see Table 1 for participant demographics).
Participant demographics
Recruitment and assessment procedures
Participants were recruited from a small, diverse, and relatively low socioeconomic status (SES) community in rural North Carolina. Participants were recruited from a larger multiwave longitudinal study including over 830 adolescents. Completion of most of the larger study survey in participant schools at session 1 was a prerequisite for participating in the current study. To be eligible, participants had to be in regular education classes, in the 6th or 7th grade at session 1, and at least 11 years and 10 months old at session 2. Participants were excluded if they had ever been diagnosed with a learning disability, such as dyslexia, ever had a seizure, had head trauma, or if participants had braces, a permanent top retainer, or some kind of permanent dental work at the time, as part of the study required a functional magnetic resonance imaging (fMRI) scan. Parent consent and adolescent assent for the study were obtained prior to participation.
Participation took place in three sessions. Session 1 and 3 took place in school, 1 year apart, during which participants completed measures of SITBs. Participants that completed the school assessment at session 1 and met criteria for participation were invited to participate in one lab assessment at session 2 to fill out a survey that asked questions about health, friends, and family, and collected baseline cardiac measurements. The first and last school assessments were separated by an average of 11.28 months (SD = 0.20). Sessions 1 and 2 occurred within the same year and are referred to hereafter as Wave 1 (W1). Session 3 is referred to as Wave 2 (W2). Primary reasons for attrition at W2 included students’ withdrawal from the study, school transitions (e.g., withdrawal, transfer, graduation), and scheduling conflicts.
Measures
Interpersonal conflict
Peer and family conflict were collected using the Multicultural, multidimensional assessment of parent–adolescent conflict (Ruiz et al., Reference Ruiz, Gonzales and Formoso1998) at W1 (session 2). The measure consists of 8 items that were asked of both close friends and family over the past month including, “You and your friends/family disagreed with each other,” “You and your friends/family ignored each other,” “You and your friends/family had a serious argument or fight,” and “You and your friends/family yelled or raised your voices at each other” and were rated on a Likert scale ranging from 1 (Almost never) to 5 (Almost always). This measure had acceptable reliability for peer (α = 0.89) and family (α = 0.90) conflict.
SITB
Nonsuicidal self-injury
NSSI frequency was assessed at W1 (session 1) and W2 using a questionnaire adapted from prior research (Prinstein et al., Reference Prinstein, Nock, Simon, Aikins, Cheah and Spirito2008). Items assessed frequency of engagement in five NSSI behaviors without intent to die in the past year (i.e., “Cut or carved my skin on purpose,” “Hit myself on purpose,” “Inserted objects under my nails or skin on purpose,” “Burned my skin on purpose,” “Scraped or picked at skin to the point of drawing blood”) on a 5-point scale (0 = Never, 1 = 1–2 times, 2 = 3–5 times, 3 = 6–9 times, and 4 = 10 or more times). NSSI frequency was computed as a total score.
Suicidal ideation
Suicidal ideation frequency was assessed using the SQ-N (Heilbron & Prinstein, Reference Heilbron and Prinstein2010) at W1 (session 1) and W2. Participants rated the frequency they had experienced eight types of suicidal thoughts in the past year (i.e., “I thought about death,” “I thought about how I would kill myself,” “I thought that killing myself would solve my problems,” “I wished I had the nerve to kill myself,” “I thought about telling people I plan to kill myself,” “I thought my life was too rotten to continue,” “I thought it would be better if I were not alive,” and “I wished I were dead”) on a 5-point scale (0 = Never, 1 = A few times, 2 = Couple times per month, 3 = About once per week, and 4 = Almost every day). SI frequency was computed as a total score.
Pubertal development scale
Participants completed the Pubertal Development Scale (PDS) at W1, which is a self-report measure that uses pictures depicting body hair growth, voice change, and facial hair growth for males and body hair growth, breast development, and menarche for females in order to categorize participants on a range from prepubertal to postpubertal and has been shown to have good reliability (Petersen et al., Reference Petersen, Crockett, Richards and Boxer1988).
Socioeconomic status
Parents completed a 9-item scale at W1 assessing the extent to which they had experienced difficulties meeting their economic needs in the prior 3 months (Conger et al., Reference Conger, Wallace, Sun, Simons, McLoyd and Brody2002). Items included questions pertaining to payment of bills (e.g., “How much difficulty did you have paying your bills?”) rated on a 5-point scale (0 = No difficulty at all, 4 = A great deal of difficulty), as well as questions assessing the extent to which statements about basic necessities pertained to them (e.g., “You had enough money to afford the kind of food you needed,” “You had enough money to afford the kind of utilities (e.g., electricity, phone, gas, water) you needed”) rated on a 5-point scale (0 = Not true at all, 4 = Very true). These latter items were reverse scored. This questionnaire has been shown to have good reliability (Tsai et al., Reference Tsai, Telzer, Gonzales and Fuligni2013).
Heart rate
Resting HR was collected at W1 (session 2) and measured in beats per minute and collected using the LifeSource UB-351 Wrist Digital Blood Pressure Monitor that conforms to the European Directive 93/42 EEC for Medical Products and is used for diagnostic purposes (AND, 2021). Prior to single measurement, participants were asked to sit up straight with their legs uncrossed and their feet flat on the floor. The device was placed around participants’ left wrist and they were asked to rest their elbow on a table and have their arm up, so that their wrist was at heart level. Heart rate was recorded after participants had been sitting for at least 5 min.
Covariates
We collected PDS, gender, race, and SES as potential covariates. In addition, we controlled for NSSI and SI at W1, in respective models. We only used covariates that were significantly associated with outcome variables to prevent overfitting. Specifically, we controlled for PDS in NSSI models, while we controlled for PDS and gender in SI models.
Statistical analyses
All statistical analyses were conducted with R Studio, version 4.0.2. Statistical significance was defined using 95% confidence intervals and p-values < .05. Exploratory analyses including histograms as well as skew and kurtosis statistics were run for each variable to check for normality. SITB measures were winsorized to ± 3 SD to correct for outliers. Peer and family conflict measures were log10 transformed to correct for skew. Peer/Family conflict and resting HR were centered in the interaction model to reduce potential for multicollinearity (Jaccard et al., Reference Jaccard, Wan and Turrisi1990).
2.8% of data were missing from the final sample, ranging from 0% (age) to 10.88% (SI). To assess whether data were missing completely at random (MCAR) we performed parametric (p < 0.001) and nonparametric (p = 0.638) tests using the MissMech package (Jamshidian et al., Reference Jamshidian, Jalal and Jansen2014). Despite that data were not MCAR, based on recent recommendations (Matta et al., Reference Matta, Flournoy and Byrne2018), in order to account for missing data, we used multiple imputation (100 imputations) using the mice package (van Buuren, Reference van Buuren2020) (see Supplemental Material for missing data by variable).
To conduct analyses, we ran a series of models, including nonhierarchical negative binomial models using the MASS package (Ripley et al., Reference Ripley, Venables, Bates, Hornik, Gebhardt and Firth2020) to examine the associations between measures of social conflict, resting HR, and SITB. In addition, we ran zero-inflation Poisson and hurdle models (see Supplementary Material) using the pscl package (Jackman, Reference Jackman2020). In models, we examined peer conflict, parent conflict, resting HR, peer conflict × resting HR, and parent conflict × resting HR as simultaneous predictors of SITB outcomes after controlling for relevant covariates. We ran two separate models based on SITB outcomes, one for NSSI and one for SI. We specifically chose negative binomial models as the main hypotheses required interaction effects. As described in depth by McCabe et al. (Reference McCabe, Halvorson, King, Cao and Kim2020), nonlinear and count models require special considerations that have to be taken into account when interpreting interaction effects in generalized linear models, and the description of nonlinear probabilities and counts are not equal to the product terms between predictor variables as is the case for linear approaches (McCabe et al., Reference McCabe, Halvorson, King, Cao and Kim2020). Therefore, linear models may be inappropriate for evaluating nonlinear and count outcomes and tend to violate assumptions of heteroscedastic and non-normal residual values, as these variables are discrete quantities that are bounded by zero and take on integer values. In order to take the nontraditional approach outlined by McCabe et al. (Reference McCabe, Halvorson, King, Cao and Kim2020) that allows for the appropriate examination of interaction effects in generalized linear models of nonlinear probabilities and counts, we used the modglm package, which is currently compatible with negative binomial, but not zero-inflation Poisson or hurdle models. This package defines interactions as the change in a marginal effect of one variable as a function of change in another variable with the use of partial derivatives and discrete differences to quantify these effects (McCabe et al., Reference McCabe, Halvorson, King, Cao and Kim2020). Furthermore, this package provides a graphical depiction of the interaction point estimates computed observation-wise that are plotted against the model predicted outcome (Hanmer & Ozan Kalkan, Reference Hanmer and Ozan Kalkan2013; McCabe et al., Reference McCabe, Halvorson, King, Cao and Kim2020; Norton et al., Reference Norton, Wang and Ai2004). The interaction effect was probed and plotted using marginal effects with the ggeffects package (Lüdecke et al., Reference Lüdecke, Aust, Crawley and Ben-Shachar2020). All models were adjusted for relevant covariates that were significantly associated with outcome variables to prevent overfitting, which included PDS for NSSI models and PDS and SES for SI models.
Results
Descriptive statistics
Descriptive statistics for primary study variables are presented in Table 2. At W1, 52 participants (35.37%) endorsed NSSI and 60 participants (40.82%) endorsed SI. There was not a significant association at W1 between HR and NSSI (r = −0.02, p = 0.784) or SI (r = −0.02, p = 0.812). There was no significant difference in resting HR for participants that endorsed either NSSI or SI compared to participants that did not endorse NSSI or SI (p < 0.05). At W2, 40 participants (27.21%) endorsed NSSI and 63 participants (42.86%) endorsed SI. As shown in Figure 1, NSSI (t(274.93) = 1.77, p = 0.077) and SI (t(289.30) = −0.66, p = 0.508) trajectories were stable across waves.
NSSI and SI trajectories by gender. Note. Figure shows raw data as individual points, distribution density of observed values, and boxplot. NSSI = non-suicidal self-injury; SI = suicidal ideation; all values are winsorized; green = males; red = females.
Descriptive statistics
Note. For NSSI and SI variables by gender, counts (N) reflect the number of participants who endorsed NSSI and SI, and percentages (%) reflect percentages of the total sample. Where reported, ranges reflects the range of values for each variable prior to winsorizing and log transformation. NSSI = nonsuicidal self-injury; SI = suicidal ideation; W1 = Wave 1; W2 = Wave 2; **p < .01; ***p < .001.
NSSI model
In main effect models, peer conflict, family conflict, and resting HR at W1 were not associated with NSSI at W2 (see Table 3). Similarly, there was not a significant interaction between family conflict and resting HR at W1 on NSSI at W2 (see Table 4) after controlling for NSSI at W1 and PDS. In contrast, as shown in Table 4 and Figure 2, there was a significant positive interaction between peer conflict and resting HR at W1 on NSSI at W2, after controlling for NSSI at W1 and PDS, such that the combination of higher peer conflict and higher resting HR was associated with greater NSSI at W2, even after controlling for multiple comparisons (p = 0.021). This model explained a substantial amount of the variance in NSSI (R 2 = 0.48). The interaction was probed and indicated that the interaction effect was significant at the hypothetical mean of all predictor variables (β = 0.080, SE = 0.021, t = 3.875, 95% CI [0.040, 0.120]), indicating that the marginal effect of peer conflict on NSSI was stronger among those with higher resting HR. The average interaction effect across observations also was significant (β = 0.150, SE = 0.054, 95% CI [0.045, 0.256]). The interaction effect ranged from −0.02 to 1.68 across observations. The marginal effect of peer conflict on NSSI was stronger among those with higher HR across the sample and was statistically different from zero 61.90% of the time with 98.00% of the interactions positive and 2.00% of the interactions negative. Figure 3 shows the interaction point estimates computed observation-wise plotted against the model predicted outcome.
Resting heart rate moderates the effect of peer conflict on NSSI.
Interaction point estimates computed observation-wise plotted against the model predicted outcome.
Main effect models
Social conflict, resting heart rate, and NSSI
SI model
In main effect models, peer conflict, family conflict, and resting HR at W1 were not associated with SI at W2 (see Table 5). Interactions of peer conflict and of family conflict with resting HR at W1 were not significantly associated with SI at W2 (see Table 5).
Social conflict, resting heart rate, and SI
Discussion
Building on RDoC and developmental psychopathology frameworks, the current study investigated whether interactions among interpersonal and physiological stress-related vulnerabilities in adolescence were associated with nonsuicidal and suicidal SITBs longitudinally. Specifically, we investigated whether adolescents’ experiences of interpersonal stress (i.e., affiliative conflict) across peers and family domains, in combination with heightened physiological arousal/regulation (i.e., baseline cardiac arousal), predicted greater frequency of NSSI and SI 1 year later. Results provided partial support for hypotheses.
Consistent with hypotheses, social conflict (i.e., peer or family) and resting HR were not associated with SITBs in main effects analyses, suggesting that neither greater affiliative conflict nor heightened cardiac arousal/regulation processes alone are sufficient to predict adolescents’ future engagement in SITBs. Interpersonal difficulties, including conflict with peers or family, are implicated in onset or maintenance of NSSI and SI across theoretical (Joiner, Reference Joiner2005; Nock & Prinstein, Reference Nock and Prinstein2004; Van Orden et al., Reference Van Orden, Witte, Cukrowicz, Braithwaite, Selby and Joiner2010) and empirical work (Adrian et al., Reference Adrian, Zeman, Erdley, Lisa and Sim2011; Glenn et al., Reference Glenn, Kleiman, Kandlur, Esposito and Liu2022; Oppenheimer et al., Reference Oppenheimer, Stone and Hankin2018; Prinstein et al., Reference Prinstein, Guerry, Browne, Rancourt and Nock2009; Turner et al., Reference Turner, Cobb, Gratz and Chapman2016), but relatively few longitudinal studies have examined the temporal sequence of these associations. Regarding NSSI, for example, engagement in NSSI itself may lead to escalations in interpersonal (i.e., family or peer) conflict. One study found that individuals who disclosed engagement in NSSI, and thus may have experienced interpersonal reinforcement for NSSI, also reported more frequent NSSI (Turner et al., Reference Turner, Cobb, Gratz and Chapman2016). This suggests a potentially complex or cyclical relationship between interpersonal conflict and NSSI over time that may not emerge in main effects models, and which may require consideration of both aversive (i.e., conflict) and desirable (i.e., support) interpersonal contingencies. It is likely that mechanisms linking interpersonal conflict with development of suicidal outcomes such as SI are similarly complex, such that additional risk or protective factors (e.g., parent history of SI; Oppenheimer et al., Reference Oppenheimer, Stone and Hankin2018; ) should be considered when examining risk prospectively. Alternatively, interpersonal conflict may predict subsequent NSSI or SI over short timescales (e.g., minutes, hours, or days; Czyz et al., Reference Czyz, Horwitz, Arango and King2019; Turner et al., Reference Turner, Cobb, Gratz and Chapman2016) but may be more weakly associated with SITBs over longer time periods (e.g., 1 year), as tested in this investigation.
Our findings shed light on a potentially more complicated picture of the relationship between interpersonal stress and SITB risk by showing that social conflict is associated with increased risk for some types of SITBs only in conjunction with biological vulnerabilities. As hypothesized, the interaction between peer conflict and resting HR at baseline significantly predicted greater frequency of NSSI, but not SI, at follow-up, such that adolescents who experienced greater peer conflict and who exhibited a higher resting HR were at greater risk for NSSI across 1 year of early adolescence. This finding replicated across negative binomial, zero-inflation poisson, and hurdle models and is consistent with a cross-sectional study that found that adolescents with deficiencies in serotonergic functioning, in conjunction with interpersonal dysfunctions (i.e., negative parent–child dyadic interactions), were at heightened risk for self-injurious outcomes (i.e., a composite SITB variable, including NSSI and suicide attempts; Crowell et al., Reference Crowell, Beauchaine, McCauley, Smith, Vasilev and Stevens2008). Our longitudinal finding provides temporal specificity by demonstrating that an interaction of biological vulnerabilities (i.e., cardiac arousal) and interpersonal difficulties (i.e., peer conflict) may precede subsequent engagement in NSSI in adolescence. In contrast, the interaction of peer conflict with cardiac arousal was not associated with SI. The combination of greater peer conflict and higher HR may be specifically associated with NSSI given the potential of NSSI behaviors to provide immediate changes in physiological indices of arousal (Franklin et al., Reference Franklin, Hessel, Aaron, Arthur, Heilbron and Prinstein2010), which may be pertinent for individuals with higher baseline levels of cardiac arousal.
Recent theoretical work posits that adolescents’ SITBs may be most likely to occur in the context of failures in acute stress regulation (Miller & Prinstein, Reference Miller and Prinstein2019), and our finding suggests that similar mechanisms may underlie risk for nonsuicidal (i.e., NSSI) outcomes longitudinally. Findings from developmental affective neuroscience suggest that neural changes during puberty may increase adolescents’ sensitivity to their social environment (Casey et al., Reference Casey, Jones and Hare2008), which may underlie increased physiological reactivity to stressors (Stroud et al., Reference Stroud, Foster, Papandonatos, Handwerger, Granger, Kivlighan and Niaura2009), prolonged stress response recovery (Gunnar et al., Reference Gunnar, Wewerka, Frenn, Long and Griggs2009; Stroud et al., Reference Stroud, Foster, Papandonatos, Handwerger, Granger, Kivlighan and Niaura2009), and deficits in cognitive control in response to emotionally evocative stimuli (Cohen et al., Reference Cohen, Dellarco, Breiner, Helion, Heller, Rahdar, Pedersen, Chein, Dyke, Galvan and Casey2016; Somerville et al., Reference Somerville, Hare and Casey2011). The pubertal transition is also associated with significant changes in peer relationships and increases in peer-related stressors (Prinstein & Giletta, Reference Prinstein, Giletta and Cicchetti2016), which are often associated with SITBs among adolescents (Juvonen & Graham, Reference Juvonen and Graham2014; King & Merchant, Reference King and Merchant2008; Massing-Schaffer et al., Reference Massing-Schaffer, Helms, Rudolph, Slavich, Hastings, Giletta, Nock and Prinstein2019). Higher cardiac arousal may be a marker of maladaptive physiological stress responses that, for adolescents experiencing greater stress in the peer environment, increases risk for engagement in NSSI behaviors, which may serve to regulate aversive physiological or affective responses to interpersonal stress (Franklin et al., Reference Franklin, Hessel, Aaron, Arthur, Heilbron and Prinstein2010; Nock & Mendes, Reference Nock and Mendes2008).
Although the study had a number of significant strengths, including prospective prediction of both NSSI and SI in a diverse sample of adolescents, results should be interpreted in light of several limitations. First, rather than using gold-standard electrocardiogram, the current study used one single measurement of resting HR with the LifeSource UB-351 Wrist Digital Blood Pressure Monitor. While the device conforms to the European Directive 93/42 EEC for Medical Products and is used for diagnostic purposes, future research should use an electrocardiogram or wearable device to record multiple measures of HR during a resting baseline to ensure accurate values (Nelson & Allen, Reference Nelson and Allen2019). Second, our measure of SI captured both passive and active ideation in a community sample (i.e., passive consideration of not being alive, as well as more serious consideration of acting on suicidal thoughts), and prediction of more acute ideation severity may yield different effects. Third, although analyses were adjusted for covariates, we did not correct for other potentially relevant covariates (e.g., medications, BMI, health conditions, sleep, physical activity, depressive symptoms, perceived stress, smoking status) for which data were not collected. Fourth, the sample size was not significantly larger than other studies on adolescent SITB. Future studies should make sure to recruit larger sample sizes. Fifth, we used a limited assessment of SITB by focusing on SI and NSSI and results may not generalize to adolescents along the spectrum of SITBs (especially to those who are engaging in suicidal behaviors). Future research should make sure to examine multiple aspects of SITB. Lastly, results from our large, diverse adolescent sample may improve understanding of SITB risk in the general adolescent population but may not generalize to high-risk or clinically acute samples.
Conclusion
This longitudinal study tested theoretical conceptualizations of SITBs as potential responses to social stress in adolescence, specifically among adolescents exhibiting physiological vulnerabilities (i.e., heightened baseline arousal/regulation) that may be markers of maladaptive stress responses. Findings indicated that peer-related stressors may increase vulnerability for subsequent engagement in one type of self-injurious behavior (i.e., NSSI) among adolescents with biological vulnerabilities (e.g., heightened cardiac arousal) that may indicate increased physiological sensitivity to stress in the social environment. In contrast, the interaction of family conflict with cardiac arousal was not associated with SITBs, consistent with evidence that developmental changes may render conflict with peers, specifically, a particularly salient and impactful stressor in adolescence (Nelson et al., Reference Nelson, Jarcho and Guyer2016; Somerville, Reference Somerville2013). Interactions among social (i.e., peer-environmental) and biological (i.e., individual-level) vulnerabilities may be specific to nonsuicidal self-injurious behaviors (vs. suicidal thoughts), as neither peer conflict, cardiac arousal, nor their interaction were associated with SI. Promoting positive peer relationships and increasing social support systems, perhaps especially among adolescents with heightened physiological arousal/regulatory processes, may protect against onset of some types of SITBs in adolescence.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0954579423000251
Funding statement
Preparation of this manuscript was supported by grants from the National Institute of Mental Health to Mitchell J. Prinstein (R01 MH107479) and Eva H. Telzer (R01 DA039923), as well as grants from the NSF to Eva H. Telzer (BCS 1539651) and Matthew G. Clayton (1000255141).
Conflicts of interest
None.
