Introduction
Non-suicidal self-injury (NSSI) is the deliberate and self-inflicted damage to body tissue in the absence of suicidal intent [
46]. NSSI is fairly prevalent among adolescents and young adults, as estimates suggest that among non-clinical populations, 17% of adolescents and 13% of young adults will have engaged in NSSI at least once in their lifetime [
106]. The most important risk factors of NSSI as identified in a recent meta-analysis include mental disorders, low health literacy, early life maltreatment (ELM), bullying, problem behaviors, female gender, and physical symptoms [
120]. NSSI is particularly often seen in the context of psychopathological conditions strongly characterized by impaired stress and emotion regulation, most prominently major depressive (MDD) and borderline personality disorder (BPD). Among adolescent psychiatric inpatients, 50 to 80% fulfill the clinical diagnostic criteria for NSSI disorder according to DSM-5 [
91,
126], including repetitive engagement (on more than 5 days) in NSSI over the past year [
3]. NSSI is linked with other health-risk behaviors, and critically, presents one of the strongest transitional predictors of suicide attempts among adolescents [
5,
81,
125]. Concerningly, the rates of both the prevalence of NSSI as well as attempted and completed suicide among adolescents and young adults have shown sharp increases in more recent years [
34,
84]. A better understanding of NSSI, not only with regard to psychological and etiological aspects, but also in terms of its neurobiology, is thus of utmost importance.
As noted in a recent review, NSSI is a complex behavior that emerges at the intersection of social, psychological, and biological mechanisms [
57]. The social and psychological aspects that contribute to the risk for NSSI are relatively well understood (see e.g. [
16,
120], and have guided the development of important and effective psychosocial treatments (see e.g. [
66]). In contrast, the biological aspects of NSSI have just begun to come to light (Ref. [
39], also see Kaess et al. [
57] for a review). Further research aiming to advance the understanding of biological alterations in NSSI is thus indicated.
One aspect that has gained traction as a proposed neurobiological or physiological mechanism involved in NSSI are alterations in the activity and functional flexibility of the autonomic nervous system (ANS). As pointed out by others (e.g. [
38]), the
Neurovisceral Integration Model (
NIM, Refs. [
109,
110]) proposes that physiological, emotion, and cognitive regulation processes are related to each other in the service of goal-directed behavior, as well as adaptability to changing environmental demands [
109,
110]—and that the interplay between these functions can contribute to individual differences in mental and physical health and disease. The model further summarizes the relationship between the central nervous system (CNS) and the autonomous nervous system (ANS), and proposes a common cortico-subcortical neural circuit that serves as the structural link between these regulation processes. A network of neural structures [
109,
110,
113], amongst others of prefrontal areas, including ventromedial prefrontal cortex and anterior cingulate cortex, and subcortical areas such as the hypothalamus and amygdala, are together called the central autonomic network (CAN). This central autonomic network regulates the ANS through sympathetic and vagal branches that innervate the heart [
4,
10]. The dynamic balance between the sympathetic and parasympathetic branches allows for flexible control over the response of the body (e.g. heart) to a range of external and internal stimuli. Importantly, the parasympathetic system is more dominant in maintaining resting heart rate, whereas sympathetic influence on heart rate unfolds in a relatively slower manner, parasympathetic regulation of the heart is much faster, allowing for momentary modulation of cardiac activity [
92]. Heart rate variability (HRV), a biomarker that can be derived from heart rate recordings, is the variation in time intervals between heart beats and provides an index of this parasympathetic influence on the heart. Here, as the vagus nerve is the primary parasympathetic nerve [
15], when we refer to HRV, we always refer to vagally mediated HRV.
Difficulties in stress and emotion regulation present a central dysfunctional component that is shared among clinical and non-clinical populations engaging in NSSI [
2], besides a psychological level [
11], this could also be visible at the level of decreased autonomic vagal activity and flexibility [
109,
110]. Decreased HRV has been interpreted as objective and transdiagnostic indicator of emotional dysregulation and psychopathology in numerous studies [
7,
109,
110]. Meta-analytic studies imply reduced resting-state (short-term) HRV in both adult [
59] and adolescent MDD [
63,
64], as well as adult [
63,
64] and adolescent BPD [
122]. Studies on autonomic vagal (parasympathetic) activity in NSSI have shown reduced resting-state HRV and increased HRV reactivity during stress paradigms, such as in response to negative mood induction, in para-suicidal adolescents [
23]. In adolescents engaging in NSSI, resting-state short-term HRV is inversely related with the severity of BPD symptomatology, providing evidence for generally altered cardiac autonomic vagal activity in adolescents engaging in NSSI. A small number of studies have also considered short-term HR in individuals engaging in NSSI compared with healthy controls, both at rest and in response to stress, reporting similar results of potentially altered cardiac autonomic activity (e.g. [
14,
53]). Although there is substantial interest in autonomic vagal activity in association with emotional dysregulation, evidence currently is limited, and further studies focusing on the various components of ANS activity in NSSI are thus warranted.
In recent years, in addition to the investigation of short-term tonic (resting-state) and phasic (reactivity and recovery) levels of HR and HRV, increasing interest in diurnal components of cardiac autonomic activity has emerged in the context of psychological processes and psychiatric symptoms and disorders [
48]. In line with other physiological mechanisms, such as control of core body temperature, blood pressure, or urine volume, cardiac autonomic activity is following a variational pattern with a frequency of an approximate solar day (i.e. 24 h). Diurnal variation of cardiac autonomic activity indexed by HR and HRV, with respective minimum and peak levels during nighttime [
44,
47,
49,
74], respectively, can be observed in children from 1 year of age [
83,
121], and might undergo marked developmental changes over the lifespan [
78]—including diminishing in older age [
112]. Research focusing on diurnal rhythms of cardiac autonomic activity in association with psychiatric symptoms and disorders is still scarce, and thus far, focus has been laid on adult populations and non-human primates [
47,
49]. Respective studies, however, substantiated sex-specific alterations of diurnal rhythms of HRV in association with depressive symptoms in adult general population samples [
19,
32,
111,
115], and altered rhythms of different markers of cardiac autonomic activity have been found in association with depressive symptoms and difficulties in emotion regulation in the context of adult BPD [
118].
The findings of altered diurnal rhythms of cardiac autonomic activity in association with certain psychiatric conditions somewhat align with findings from the field of chronobiology, suggesting that circadian rhythms of peripheral physiology, e.g. including body temperature, blood pressure, glucocorticoid secretion, or immune responses, are altered in association with psychopathology ([
71,
114], [
21]. Growing evidence from both pre-clinical and human studies furthermore substantiates altered circadian rhythms, from molecular genetic up to behavioral levels, in the presence of many psychiatric disorders [
61,
119]. Of note, it has been emphasized previously that research should examine disturbances within different components of the circadian system in association with NSSI [
123], to date, however, research on this aspect of NSSI disorder is still essentially lacking.
In the present exploratory study, we examined diurnal variation patterns of cardiac autonomic activity in female adolescents with NSSI, assuming to find this functional component of ANS activity to be altered in this vulnerable subpopulation. We were primarily interested in potential alterations of diurnal variation patterns of cardiac autonomic activity in the presence of NSSI disorder, and thus assessed diurnal variation patterns of cardiac autonomic activity in a sample of female adolescents fulfilling DSM-5 criteria of NSSI disorder in comparison to healthy, age-matched control females. We assumed that diurnal tendencies would maximally express themselves only under conditions which sleep is less constrained by external factors (such as, e.g. school commitments) [
45], and thus assessed cardiac autonomic activity over two consecutive days on a weekend—as opposed to normal school or working days—and under natural conditions. Based on the strong associations of NSSI with emotional dysregulation, with both MDD and BPD, and with ELM exposure, as well as based on findings of reduced 24-h HRV in association with greater difficulties in emotion regulation in adult BPD [
118] and of blunted HRV increase at nighttime in association with both acute and chronic stress exposure [
50,
58,
112], we hypothesized that: in female adolescent NSSI disorder, we would find altered diurnal variation patterns of cardiac autonomic activity characterized by reduced rhythm-adjusted mean level and amplitude of HRV, and elevations in respective parameters of HR, as well as potential phase shifts in both measures—compared to healthy, age-matched control females (H1). Furthermore, we examined diurnal variation of cardiac autonomic activity in association with dimensional clinical variables. In these secondary exploratory analyses, we assumed to find significant associations of diurnal parameters of HR and HRV with severity of BPD symptomatology (H2a), as well as the severity of ELM exposure (H2b), depressive symptomatology (H2c), and difficulties in emotion regulation (H2d).
Discussion
In the present study, based on exploratory analyses, we examined diurnal variation patterns of cardiac autonomic activity in female adolescents with repetitive NSSI, and age-matched healthy controls. We hypothesized (H1) that in NSSI disorder, we would find altered diurnal variation patterns of HR and HRV, quantified by using cosinor function parameters on 48 h of preprocessed cardiac autonomic data collected over one weekend, as compared to HC. We also examined diurnal variation patterns of HR and HRV in association with a range of dimensional clinical predictors (H2a–H2d; i.e. severity of BPD symptomatology, ELM exposure, depressive symptoms, and emotional dysregulation) in secondary analyses, and tested the robustness of our results against potential confounders (i.e. age, BMI, step count, and cardiac data quality).
In partial support of our main hypothesis (H1), (unadjusted) differential rhythmicity analyses indeed revealed significant differences in diurnal patterns of cardiac autonomic activity between the two study groups. Within the present recoding period of 48 h, the NSSI group showed significantly higher and lower rhythm-adjusted mean levels of both 24 h HR and HRV, respectively, compared to HC. Furthermore, significant alterations in amplitudes regarding both HR and HRV were observed in the NSSI group: amplitude of HR was significantly higher, while amplitude of HRV was significantly lower, compared to HC. Finally, significant acrophase shifts were observed regarding both HR and HRV, such that the NSSI group reached peak levels in both these cardiac autonomic parameters approximately 1 h later.
The present finding of significantly lower rhythm-adjusted mean levels of HRV and significantly higher rhythm-adjusted mean levels of HR in NSSI compared to HC support the notion of chronically low levels of cardiac vagal (parasympathetic) activity in NSSI relative to HC. This expands on previous findings of lower short-term autonomic vagal activity in individuals with NSSI, indexed by reduced resting-state HRV and increased HRV reactivity (e.g. [
23]). Based on the NIM [
109,
110], shared neural circuits involved in the regulation of ANS activity and emotion form a functional overlap between autonomic arousal and emotion regulation. In a further elaboration of the NIM with regard to the developmental period (
dynamic model of neurovisceral integration in development, Ref. [
62], it has been suggested that the functional interaction of the ANS and CNS is shaped early in the course of life, and that adolescence might present the most sensitive period in the development of this circuitry [
62]. In normative development, vagal influence over cardiac autonomic activity is proposed to increase, indexed by normative decreases in HR and increases in HRV, respectively. ANS maturation, in turn, is assumed to be critical for patterns of PFC maturation and associated regulatory capacities over subcortical regions to emerge, affecting stress and emotion regulation [
62]. Chronically low HRV and high HR, respectively, as observed in the present group of female adolescents with NSSI disorder compared to HC, might reflect relative absence of normative ANS maturation. ANS dysmaturation in the NSSI group might further reflect disruption in developmental patterns of PFC maturation (i.e.
cortical thinning) [
62] linked with heightened sensitivity to stressors and maladaptive coping—resulting in an increased risk of psychopathological outcomes in the long run. Critically, though, given the current lack of longitudinal research on the association between ANS maturation and developmental psychopathology [
62], no conclusions can currently be drawn as to whether the observed alterations in ANS activity patterns might be causative of NSSI behavior or merely correlational, or whether they may represent systemic adaptations as assumed, e.g. in association with ELM exposure (e.g. [
56].
The findings of significantly altered amplitude and significant phase shifts of cardiac autonomic activity in NSSI disorder somewhat align with previous studies in the field of chronobiology, suggesting disturbed circadian patterns in association with psychiatric symptoms and disorders, which are seen already in young age [
37,
41,
69,
77,
85,
87]. In chronobiological research, significant changes in amplitude and acrophase (
phase shifts) of circadian rhythm, commonly referred to as circadian rhythm disruption [
61], have been associated with increased risk for physical and psychiatric disorders [
8]. Relatedly, circadian disruption in the form of misalignment between the circadian system and daily sleep–wake behaviors were shown to adversely affect mood levels and cortical activity underlying mood regulation [
20]. Again, brain imaging studies suggest adolescence to present a sensitive period for brain maturation and particularly maturation of prefrontal regions, where dramatic changes on both structural and functional levels can be observed [
17,
33,
62,
89,
104]. Such changes might be linked with lower impulse control inhibition, poorer decision making in emotional context, greater risk-taking behavior, and heightened patterns of emotional instability in adolescence [
6,
17]. Provided that adolescence is also characterized by normative changes in the circadian rhythm of numerous physiological processes (e.g. physiologically mediated shifts toward evening preference, which also contribute to irregular sleep schedules and a general mismatch between behaviors and circadian rhythms, Ref. [
18], altered amplitudes and significant phase shifts of physiological indicators, as observed in the present sample of female adolescent with NSSI, might indicate even stronger mismatch between different circadian processes and behavior, further indicating heightened vulnerability to stress. While sleep deprivation in adolescents has been associated with deficits in emotion regulation (e.g. [
9]), there is substantial evidence that sleep problems such as disrupted sleep, poor sleep quality and shorter sleep duration increases risk of engagement in NSSI—and that such associations are particularly strong among adolescents [
60]. Interestingly, studies examining attendances in accident and emergency departments related with NSSI behavior [
43], and studies using high‑frequency experience sampling in help-seeking populations engaging in NSSI [
65], report that among adolescents, this behavior is observed most frequently in the evening hours, suggesting developmental specificities in diurnal rhythms which might also affect NSSI behavior.
The present results considering dimensional clinical predictors (H2a–H2d) using multivariate regression models suggested ELM exposure (unadjusted analyses) and the severity of BPD symptoms (adjusted) to present significant dimensional predictors of diurnal variation patterns of both HR and HRV. In participants reporting higher severity of ELM exposure, a significantly lower amplitude in HR and respective higher amplitude in HRV was observed. Furthermore, a higher severity of BPD symptomatology (NSSI subgroup) was significantly linked with lower amplitude of HR. Considering the etiology of both NSSI disorder and BPD, exposure to adverse environmental factors during critical developmental periods are considered important mediators of the respective disorder, and exposure to severe forms of early life stress in particular, such as ELM, has strongly and consistently been linked with the emergence of NSSI [
27,
35,
54,
76,
80] and BPD [
54,
72,
80].
Many neural structures, circuits, and neuro-humoral systems involved in stress and emotion regulation are altered in association with ELM exposure (see, e.g. [
12,
24,
102]), and circadian disruption might be a critical patho-mechanism linking ELM exposure with increased risk of psychopathology [
61,
119]. Multilevel interactions between the major stress and circadian systems are vital for adaptive functioning of central bio-behavioral mechanisms, while exposure to severe traumatic stress can critically alter the functional interplay between those systems [
1]. Altered circadian rhythms, indicating circadian dysregulation after severe stress exposure, may present one of the core features of trauma-related psychopathology, mediating enduring neurobiological correlates of traumatic exposure through maladaptive stress regulation [
1]. The present findings of altered diurnal parameters of cardiac autonomic activity in association with chronic severe stress exposure (i.e. ELM) as well as BPD (which by some has been considered a trauma-related disorder) align with these findings. Yet, no conclusions regarding potential alterations of endogenous rhythmicity can be drawn based in the present findings, due to critical limitations. Importantly, given the lack of control over exogenous stress related factors as a potential explanation of the present findings of altered diurnal HR and HRV in the NSSI study group, we are not able to draw any conclusions regarding potential alterations in endogenous circadian rhythms based on our study results. As outlined previously, alterations in ANS activity could present an important risk factor in adolescence for the development of more severe psychopathology. Thus, potential alterations in circadian rhythms of cardiac autonomic activity in NSSI disorder could present interesting targets for future investigation, also given the relative ease and high temporal resolution of ECG measurement. Further insights into circadian rhythms of physiological systems in adolescence in general, and potential alterations among vulnerable subgroups, might be important for the development of novel neurobiological-based interventions that have the potential to restore autonomic vagal activity and affect regulatory processes. Indeed, to our knowledge, no study to date assessed potential alterations in circadian variation patterns of cardiac autonomic activity in adolescent NSSI, nor in association with any other psychiatric disorder or respective symptoms, in childhood and adolescence. Thus, rigorous chronobiological oriented studies are warranted, assessing and examining potential alterations in circadian variation of physiological rhythms in NSSI. Crucially, a better understanding of the mechanisms of how early exposure to severe and chronic stress affects physiological regulatory systems in the long run will thereby continue to present an important research objective, allowing to further advance preventive and interventional strategies in vulnerable risk-populations, including adolescents with NSSI disorder.
While previous studies in adults and non-human primates substantiated potential alterations in diurnal variation of cardiac autonomic activity in association with depressive symptoms and disorders, as well as with difficulties in emotion regulation in the context of BPD [
47,
49,
115,
118], in the present study of female adolescents with and without NSSI disorder, we did not find significantly altered diurnal variation of HR or HRV in association with depressive symptoms or emotional dysregulation. Of note, previous studies examining the association of diurnal variation of cardiac autonomic activity with depressive symptoms [
47,
115] recruited a different target population of relatively healthy adult individuals. Furthermore, the only study examining diurnal variation of cardiac autonomic activity in clinical populations of psychiatric patients again recruited adults, and furthermore, a different analytical approach had been adopted: While we and others [
47] used cosinor to derive parameters of diurnal rhythmicity (as suggested also in existing recommendations, see [
48]), in the aforementioned study [
118], in the diurnal analysis, ECG recordings were segmented in 30 min epochs and averaged according to the sleep or wake periods, allowing to examine associations of clinical variables with mean level during these two periods, as well as with mean-level differences between sleep and wake periods, while not allowing for inferences regarding the range of oscillation or potential phase shift. Although insightful, the results of previous studies in psychiatric populations might, therefore, not readily be comparable to the findings of the study at hand. Of note, while NSSI disorder might be associated with sleep disruptions [
123], in the present study, we did not find statistically significant differences in sleep duration derived from acceleration data between study groups. Indeed, previous studies could show that sleep regularity and timing reflect well-being better than sleep duration [
8,
51,
69,
79], and thus, future studies concerned with cardiac autonomic measures and their diurnal variation in association with psychopathology might measure the timing and regularity of sleep rather than mere sleep duration, and explore and also control for associations between sleep and patterns of physiological variation.
The presented results must be considered within the context of critical limitations inherent to the study at hand, which will hopefully inform the conduct of future research. As mentioned above, a major drawback of this study is the lack of inclusion of data on experiences throughout the day. Cardiac autonomic measures fluctuate in response to a host of different factors, and no definite conclusions can be drawn with regard to typical vs. atypical diurnal variation patterns of HR and HRV without further knowledge on internal and environmental influences participants might have been exposed to over the 48 h period of cardiac data recording. Moreover, besides the covariates presently included, there might be a range of further variables that could influence (diurnal) cardiac autonomic activity, including puberty status, menstrual cycle, psychotropic medication, and alcohol, nicotine, or caffeine consumption [
73,
75,
94,
99,
99,
100,
100,
117]. Of these, the lack of control for potential medication in the NSSI group presents a particularly severe drawback of this study, and any form of medication should be considered in future studies focusing on HRV in NSSI. Of note, one previous study assessed potential covariates such as consumed units of tobacco, coffee, and alcohol on an hourly basis using EMA [
115]. Finally, in the present study, which was largely based on assumptions put forth in the NIM ([
109,
112,
113]), and in line with many previous research studies in the field (e.g. [
47,
49,
115]), we have focused on a HRV measure of autonomic vagal activity, however, sympathetic- or baroreflex-based HRV measures might provide further valuable insight into potential alterations of cardiac autonomic activity in NSSI, for example, one previous study has investigated a host of different HRV metrics in association with BPD [
118]. Thus, future research, when expanding on the present findings, might focus on different HRV measures, to gain further valuable insight. Based on the present limitations, future studies should more rigorously measure and control for potential confounders (e.g. socio-emotional stressors, medication, alcohol and nicotine consumption, etc.), potentially using high-frequency EMA, in concert with recruitment of larger and more diverse samples, and considering a variety of cardiac autonomic measures, to critically extend and further substantiate the validity and credibility of the present results.