Associations between peripheral plasma cytokine levels and patterns of non-suicidal self-injury methods among Chinese outpatient adolescents: a latent class analysis

Abstract

Objectives

Non-suicidal self-injury (NSSI) exhibits a high prevalence among adolescents. It is common for adolescents to employ multiple methods of NSSI, and these behaviors can manifest in diverse combinations. This study employs latent class analysis (LCA) to identify subgroups of adolescents based on the methods of NSSI and investigates the association between peripheral plasma cytokine levels and the identified NSSI patterns.

Methods

The research enrolled 405 outpatient adolescents, ranging in age from 12 to 18 years, who had exhibited NSSI behavior. These participants completed the Adolescent Non-suicidal Self-Injury Behavior Questionnaire (ANSAQ) and assessments of plasma cytokine levels. Latent class indicators consisted of 12 NSSI variables. Logistic regression analysis was employed to examine the associations between the identified classes and the related cytokines.

Results

LCA results supported a three-class model, categorizing participants into multiple methods group (24.20%), moderate methods group (43.70%), and mild/singular method group (32.10%). Further multinomial logistic regression analysis revealed that individuals with higher levels of interleukin-2 (IL-2) and interleukin-10 (IL-10) were more likely to be classified into the multiple-methods group compared to those in the mild/singular-method group (OR = 1.233 for IL-2; OR = 1.634 for IL-10). Additionally, individuals with elevated IL-10 levels were more likely to be classified into the multiple-methods group compared to those in the moderate-methods group (OR = 1.454). Furthermore, relative to individuals in the mild/singular-method group, those with higher IL-2 levels were more likely to belong to the moderate-methods group (OR = 1.207). These findings suggest a significant association between IL-2 and IL-10 levels and the severity of NSSI behaviors.

Conclusion

The classification of adolescents based on NSSI methods indicates that elevated levels of IL-2 and IL-10 increase the risk of engaging in NSSI involving multiple methods in adolescents. This has potential implications for understanding risk factors and developing targeted prevention and treatment strategies.

Introduction

Non-suicidal self-injury (NSSI) refers to repetitive, intentional acts of self-harm performed without suicidal intent, leading to varying degrees of physical injury. These behaviors, including cutting or puncturing the skin, are not socially accepted or condoned [1]. NSSI is particularly prevalent among adolescents compared to other age groups. Globally, approximately 19.5% of children and adolescents engage in NSSI, while in China, the estimated prevalence among middle school students (ages 13–18) is 27.4% [2]. NSSI is a significant predictor of future suicidal behavior and is increasingly recognized as a major public health concern, greatly impacting the mental health and well-being of adolescents [3]. Therefore, early identification and intervention of risk factors for NSSI in adolescent populations is crucial.

Existing studies on NSSI, typically relying on surveys and interviews, have not delved deeply into the characteristics of adolescent self-injury subgroups identified through different measurement approaches. These subgroups are often distinguished based on factors such as the definition of self-harm, frequency, severity, nature of the injury, gender, and co-occurrence with suicidal ideation or attempts [4–6]. The varying combinations of these factors lead to a range of potential NSSI subtypes. Seemingly similar self-injury behaviors may actually represent distinct subtypes, each influenced by different underlying factors. For instance, even if participants show similar overall NSSI scores in terms of frequency or severity, their responses across various items may differ, and individuals may engage in different types of self-injury behaviors [7]. Failure to differentiate between these subgroups may weaken the explanatory power of a particular factor for a specific subtype, reducing its predictive value for self-injury behavior. Therefore, identifying subgroups of individuals who engage in self-injury is crucial for reducing the heterogeneity of NSSI presentations, which has important implications for enhancing our understanding of related factors and developing tailored interventions for NSSI.

Latent Class Analysis (LCA) is a prominent unsupervised machine learning technique. Unlike variable-centered modeling, LCA can identify latent subgroups that exhibit distinct behavioral patterns that may not be directly observable [8]. For instance, He et al. identified two subgroups of NSSI and the associated factors, enabling the early identification of high-risk populations for primary NSSI among adolescents diagnosed with depression [9]. Hamza and Willoughby applied LCA to classify young adults into three distinct subgroups based on patterns of NSSI and suicidal behaviors: an infrequent NSSI/not high risk for suicidal behavior group, a frequent NSSI/not high risk for suicidal behavior group, and a frequent NSSI/high risk for suicidal behavior group. The results highlight the importance of evaluating individual differences in NSSI characteristics, as well as psychosocial functioning, when assessing the risk of suicidal behavior [10]. Case et al. identified four NSSI subgroups—experimental NSSI group, mild NSSI group, multiple functions/anxious NSSI group, and automatic functions/suicidal NSSI group—while examining differences in self-esteem, social support, loneliness, suicidal ideation, and behaviors across these groups [11]. These studies underscore the heterogeneity within NSSI, highlighting the need for further exploration and differentiation of these subgroups to better understand and address their specific risk factors.

Our understanding of the neurobiology of adolescents with NSSI is still limited. Immune dysfunction plays a crucial role in the development of NSSI. However, only a few studies have investigated the relationship between NSSI and inflammation. Zheng et al. found that systemic inflammation, as indicated by elevated monocyte-to-lymphocyte ratio and platelet-to-lymphocyte ratio, was strongly associated with NSSI in adolescents [12]. Kim et al. reported that patients with mood disorders and NSSI had higher levels of serum tumor necrosis factor α (TNF-α) compared to those without NSSI, although no significant differences were observed in interleukin-1β (IL-1β) and interleukin-6 (IL-6) levels [13]. In a separate study, Osimo et al. showed that low-grade systemic inflammation was linked to NSSI in adult psychiatric inpatients [14]. Kindler et al. suggested that the self-injury group exhibited significantly higher absolute white blood cell counts and white blood cell/cortisol ratios (adjusted for body mass index and smoking) compared to the healthy control group, although no significant differences in IL-6 and C-reactive protein (CRP) levels were found between the two groups [15]. However, results across studies have been inconsistent. For instance, Russell et al. found no association between IL-6 and NSSI, and even reported a negative correlation between CRP levels and NSSI, suggesting that higher CRP levels may have a protective effect against NSSI [16]. Another recent study found that among borderline personality disorder patients, those with lower IL-6 levels were more likely to have engaged in self-injurious behavior within the past week. This link involves the endocannabinoid system’s regulation of IL-6, with anandamide and 2-arachidonoylglycerol playing key roles [17]. Despite these contradictory findings, there is increasing interest in further investigating the relationship between inflammation and NSSI.

This study employed LCA to classify adolescents with NSSI based on behavioral characteristics, thereby revealing the latent heterogeneity within this population. This approach facilitates the identification of subgroups with distinct behavioral profiles, laying the groundwork for developing more targeted and effective intervention strategies. Previous research has provided preliminary evidence suggesting a potential link between immune function and NSSI. However, systematic empirical investigations examining this relationship, particularly within adolescent populations, remain limited. Building on this, the present study further investigates differences in both behavioral features and inflammatory markers among NSSI subtypes. By examining inflammation-related biological mechanisms that may contribute to the onset of specific NSSI subgroups, our findings provide preliminary evidence elucidating the biological underpinnings of NSSI heterogeneity.

Methods

Participants

A total of 405 outpatient adolescents were recruited from Qilu Hospital (Qingdao), Cheeloo College of Medicine, Shandong University. The diagnosis was conducted by two attending physicians or clinicians of higher seniority through structured clinical interviews based on the relevant diagnostic criteria. The inclusion criteria were: (1) meeting the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria for NSSI, defined as at least five self-injury episodes in the past year; (2) male or female aged 12–18 years old; (3) Diagnosed according to the International Classification of Diseases, 10th Revision (ICD-10), including anxiety disorders, depressive episodes, and bipolar disorder; (4) first visit without medication and psychotherapy; (5) a relatively stable emotional state and normal reading, comprehension and verbal skills, as well as a willingness to complete the tests; and (6) voluntary participation in the study.

The exclusion criteria were: (1) current diagnosis of any other psychiatric disorders; (2) history of suicide attempts or suicidal ideation (based on the patient’s history of illness); (3) presence of severe physical illness, especially diseases associated with systemic inflammatory responses, such as active infections, fever, or acute or chronic endocrine disorders; (4) use of anti-inflammatory drugs, corticosteroids, or antipyretics within the past two weeks.

This study was conducted in accordance with the Declaration of Helsinki. It was approved by the Ethics Committee of Qilu Hospital (Qingdao), Cheeloo College of Medicine, Shandong University, under approval protocol number KYLL-2,024,023. All eligible participants, as well as their guardians, provided informed consent after being fully briefed on the study’s objectives and procedures.

Measures

The adolescent non-suicidal self-injury behavior questionnaire (ANSAQ)[18]

The ANSAQ is divided into two sections: the behavior section and the function section. In this study, we utilized the behavior section, which contains 12 items that assess 12 different types of self-injurious behaviors. A five-point Likert scale is used, with responses ranging from 0 to 4, corresponding to “never,” “occasionally,” “sometimes,” “often,” and “always,” respectively. Higher scores reflect more severe self-injury behaviors. In this study, the Cronbach’s α coefficient for the behavior questionnaire was 0.89.

24-Item Hamilton depression rating scale (HAMD-24)[19]

The Chinese version of the HAMD-24 is widely used in research involving Chinese populations and consists of 24 items. It is a well-established tool for assessing depressive symptoms in both clinical and research settings. For the analysis in this study, aggregate scores were used, with higher scores indicating more severe depressive symptoms, rather than diagnosing clinical depression. In this study, the Cronbach’s α coefficient for the behavior questionnaire was 0.81.

Hamilton anxiety rating scale (HAMA)[20]

The Chinese version of the HAMA has demonstrated robust reliability and validity in epidemiological studies involving Chinese adolescents. The total score is calculated by summing the individual item scores, with higher scores indicating greater severity of anxiety symptoms. In this study, the Cronbach’s α coefficient for the behavior questionnaire was 0.86.

Peripheral immunological markers

EDTA-anticoagulated whole blood samples were obtained from fasting outpatients via forearm venipuncture between 7:00 and 9:00 AM. The samples were centrifuged at 3,500 rpm for 8 min. Plasma was stored at −80℃within 4 h of collection. The levels of 12 cytokines(IFN-α, IL-5, IL-2, IL-6, IL-1β, IL-10, IFN-γ, IL-8, IL-17, IL-4, IL-12p70, TNF-α) were measured using the 12-Cytokines Assay Kit (Raisecare, Qingdao, Shandong, China)in the serum samples by multiplex bead-based flow immunofluorescence assay.

Statistical analysis

LCA was conducted using Mplus 8.0, with twelve NSSI variables (coded as no = 0, yes = 1) serving as indicators for the latent classes. The model selection process for LCA involved a comparative analysis of different models to determine the optimal number of latent classes. The determination of the optimal number of latent classes was based on several statistical criteria, including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), sample-size-adjusted Bayesian Information Criterion (aBIC), entropy, the Lo-Mendell-Rubin (LMR) likelihood ratio test, and the bootstrap likelihood ratio test (BLRT). The model with the smallest AIC, BIC, and aBIC values was preferred. In the context of LMR and BLRT, a statistically significant p-value (p < 0.05) indicated that the estimated model provided a significantly better fit to the data compared to the model with one fewer class. The entropy statistic was used to assess classification quality, with values closer to 1 indicating higher classification accuracy [21]. Subsequently, the latent classes identified from the optimal model were designated as the dependent variable, while cytokines and related factors were treated as independent variables in a multinomial logistic regression analysis to investigate the predictive influence of these factors on NSSI-based class membership.

Results

Prevalence of NSSI

Table 1 presents the types and frequencies of NSSI behaviors reported by participants. Among all reported NSSI methods, the most common behaviors included punching (59.5%), pinching (56.3%), scratching (47.2%), hitting (44.2%), and banging (43.5%). Less frequently reported behaviors were burning (7.9%), carving (23.0%), and skin rubbing (25.7%).

Table 1.

Frequency of NSSI behaviors (n = 405)

Type of NSSI (%)	None	Occasionally	Sometimes	Often	Always
1. Pinching	43.7	17.0	23.2	9.4	6.7
2. Scratching	52.8	15.8	17.5	9.4	4.4
3. Banging	56.5	21.0	12.6	7.7	2.2
4. Punching	40.5	23.0	19.5	11.4	5.7
5. hitting	55.8	16.3	14.1	10.1	3.7
6. Needle sticking	65.2	11.1	11.6	8.9	3.2
7. Cutting	61.2	11.4	8.1	10.4	8.9
8. Biting	66.7	12.3	10.9	5.9	4.2
9. Hair pulling	61.7	16.8	10.6	4.4	6.4
10. Burning	92.1	5.2	2.5	0.2	0.0
11. Rubbing skin	74.3	11.6	8.1	3.5	2.5
12. Carving	77.0	11.6	4.0	3.2	4.2

LCA

We conducted an assessment of the fit of four LCA models, which were labeled as M1 through M4, respectively. The statistical fit indices are presented in Table 2. Both the LMR and BLRT yielded significant results for the two-class and three-class models. Although the entropy values for both models were comparable, the three-class model demonstrated lower AIC and aBIC values than the two-class model. Based on these indices, the three-class model was determined to be the optimal solution. Figure 1 illustrates the three subtypes identified by the model. The conditional probabilities across the three latent classes distinctly differentiate these subtypes. Class 1, representing approximately 24.20% of the sample (n = 98), exhibited high endorsement probabilities for all NSSI methods and was designated as the “multiple methods” class. Class 2, comprising 43.70% of the sample (n = 177), displayed moderate endorsement probabilities for the majority of NSSI methods, with the exception of the burn or scald method, which demonstrated low endorsement rates. This class was labeled as the “moderate methods” class. Lastly, Class 3, accounting for 32.10% of the sample (n = 130), demonstrated low endorsement probabilities for most NSSI methods, with slightly higher probabilities for behaviors such as banging and hair-pulling. This class was designated as the “mild/singular method” class.

Table 2.

Fit indices for latent class analysis models based on NSSI behaviors

	AIC	BIC	aBIC	Entropy	LMR	BLRT	N per class
M1	6036.935	6084.982	6046.904
M2	5045.814	5145.911	5066.583	0.851	0.0000	0.0000	184/221
M3	4869.917	5022.065	4901.486	0.831	0.0068	0.0064	98/177/130
M4	4841.308	5045.506	4883.677	0.771	0.1420	0.1391	97/107/99/102

Abbreviations: AIC Akaike Information Criterion, BIC Bayesian Information Criterion, aBIC sample-size-adjusted Bayesian Information Criterion, LMR Lo-Mendell-Rubin likelihood ratio test, BLRT Bootstrap Likelihood Ratio Test

Fig. 1.

Item-class profile plot of non-suicidal self-injury for the three-class. Class 1 = multiple-methods class; Class 2 = moderate-methods class; Class 3 = mild/singular-method class

Comparison of variables among classes

The findings from LCA were employed to categorize participants into their most probable class, with class membership acting as an dependent variable to explore potential disparities in demographic attributes and clinically pertinent symptoms among different classes. Table 3 delineates statistically significant variations in age, gender, Body Mass Index (BMI), as well as scores on HAMA and HAMD across the various subgroups. In contrast, Table 4 demonstrates that there are no statistically significant differences in the levels of the 12 cytokines measured across the subgroups.

Table 3.

Demographic and clinical characteristics of study participants

Variables	Class = 1 (n = 98)	Class = 2 (n = 177)	Class = 3 (n = 130)	F/χ2	p
Age (years, mean ± SD)	14.83 ± 1.61	15.41 ± 1.66	15.57 ± 1.67	6.141	0.002
Gender [males, n (%)]	21(21.4)	63(35.6)	70(53.8)	25.707	< 0.001
BMI (kg/m2, mean ± SD)	20.26 ± 4.06	21.75 ± 4.98	21.63 ± 4.69	3.518	0.031
Smoking [no, n (%)]	92(93.9)	173(97.7)	126(96.9)	2.903	0.234
HAMA (mean ± SD)	24.27 ± 6.39	18.90 ± 6.97	14.25 ± 7.22	58.649	< 0.001
HAMD (mean ± SD)	23.52 ± 5.81	19.44 ± 6.00	14.50 ± 5.94	66.017	< 0.001

Abbreviations: BMI body mass index, HAMD Hamilton Depression Scale, HAMA Hamilton Anxiety Scale, SD standard deviation

Table 4.

Serum cytokine levels in study participants

Variables (pg/ml, Mean ± SD)	Class = 1 (n = 98)	Class = 2 (n = 177)	Class = 3 (n = 130)	F	p
IFN-α	1.95 ± 1.27	1.86 ± 1.19	1.83 ± 1.01	0.324	0.723
IL-5	2.31 ± 2.10	1.97 ± 1.51	2.39 ± 4.18	1.014	0.364
IL-2	2.38 ± 2.22	2.52 ± 3.01	2.06 ± 1.86	1.292	0.276
IL-6	1.98 ± 1.35	2.29 ± 1.68	2.33 ± 2.77	1.002	0.368
IL-1β	7.21 ± 10.74	5.07 ± 8.04	6.46 ± 11.01	1.694	0.185
IL-10	2.01 ± 1.65	1.77 ± 0.91	1.69 ± 0.91	2.415	0.091
IFN-γ	5.30 ± 5.42	5.12 ± 4.89	5.54 ± 8.35	0.165	0.848
IL-8	2.11 ± 3.75	1.79 ± 3.67	2.29 ± 6.50	0.441	0.644
IL-17	4.60 ± 5.48	4.09 ± 5.20	4.37 ± 4.28	0.349	0.706
IL-4	1.06 ± 1.13	0.97 ± 0.78	1.08 ± 1.06	0.554	0.575
IL-12p70	1.06 ± 0.40	1.09 ± 0.48	1.15 ± 0.90	0.704	0.495
TNF-α	2.35 ± 2.28	2.03 ± 1.67	2.10 ± 2.16	0.835	0.435

Abbreviations: IFN Interferon, IL interleukin, TNF tumor necrosis factor

In order to validate the measurement model and more comprehensively characterize the heterogeneity in NSSI, multinomial logistic regression was utilized to investigate the associations between the identified latent classes and various study variables. As presented in Table 5, age, gender, HAMA, HAMD, IL-2, and IL-10 were identified as significant predictors of class membership based on NSSI patterns. Specifically, individuals with relatively younger age, higher HAMA scores, and elevated IL-10 levels were found to be more likely to belong to the multiple-methods class when compared to those in the mild/singular-method class and moderate-methods classes. In addition, individuals who are female, have higher HAMD scores, and elevated IL-2 levels are more likely to be classified into the moderate-methods class or multiple-methods class, compared to those in the mild/singular-method class.

Table 5.

Associations between study variables and NSSI patterns

	Class1with Class 2a		Class 1 with Class 3b		Class 2 with Class 3b
Variables	OR	95% CI	OR	95% CI	OR	95% CI
Age	0.791*	0.662–0.946	0.747**	0.607–0.920	0.945	0.809–1.102
Gender c	0.690	0.349–1.361	0.401*	0.188–0.856	0.582*	0.340–0.995
HAMA	1.105**	1.037–1.177	1.115**	1.035-1.200	1.009	0.953–1.068
HAMD	1.059	0.984–1.141	1.204**	1.104–1.313	1.136**	1.064–1.213
IL-2	1.022	0.894–1.167	1.233*	1.032–1.473	1.207*	1.045–1.393
IL-10	1.454*	1.056–2.001	1.634**	1.129–2.363	1.124	0.844–1.495

Abbreviations: HAMD Hamilton Depression Scale, HAMA Hamilton Anxiety Scale, IL interleukin, OR odds ratio, CI confidence interval

^a Class 2 was used as the reference class; ^b Class 3 was used as the reference class; ^c Males were used as the reference

^*p < 0.05, ^**p < 0.01

Discussion

Our study used LCA to investigate the characteristics of NSSI in an outpatient adolescent sample, identifying three latent classes of adolescent NSSI: multiple-methods group, moderate-methods group, and mild/singular-method group. These findings suggest significant heterogeneity in adolescent NSSI behaviors. The multiple-methods group comprises 24.20% of the total sample and exhibits the highest conditional probabilities for all NSSI methods. This subgroup is congruent with the “Multiple methods class” of earlier studies [22]. This suggests that individuals in this group engage in more severe NSSI behaviors compared to the other two groups, characterized by a higher frequency of self-injury and a greater variety of self-injury methods. The moderate-methods group represents 43.70% of the total sample, making it the largest group. Compared to the multiple-methods group, individuals in this group exhibit moderate conditional probabilities across all NSSI methods. This group is similar to the Multiple Functions NSSI/Low Suicide Ideation group in previous studies [23], as they still demonstrate a certain degree of NSSI tendencies. The mild/singular-method group accounts for 32.10% of the total sample. This group was congruent with the ‘Experimental/Mild NSSI’ group of earlier studies [23]. Individuals in this group were characterized by low-frequency engagement in NSSI in the past year and participation in fewer methods of NSS. These findings provide valuable insights into the differences in NSSI behaviors across the three groups, laying the groundwork for developing targeted intervention strategies.

In our sample, female adolescents were identified as being at a higher risk for NSSI when compared to their male counterparts. This finding is congruent with prior research on Chinese psychiatric adolescent inpatients exhibiting non-suicidal self-injury behavior [24]. It appears that female adolescents may have a greater propensity to utilize NSSI as a coping mechanism for dealing with negative emotions. Furthermore, our study indicated that younger individuals were significantly more likely to engage in multiple methods of self-injury. This aligns with existing literature indicating that episodes of NSSI typically reach their peak during mid-adolescence and subsequently decline in late adolescence [25]. However, this study excluded individuals with a history of suicidal ideation, which may have led to an underrepresentation of older individuals who use multiple methods [26]. Accordingly, the observed association between younger age and classification into the multiple-methods self-injury group warrants cautious interpretation. Future research involving a broader clinical sample is warranted to validate these findings.

We have identified an association between NSSI and symptoms of anxiety and depression, which aligns with the findings from other investigations concerning adolescent psychiatric inpatients [24]. NSSI and depressive episodes can enter into a reciprocal reinforcing pattern, wherein NSSI serves to temporarily alleviate the negative emotions characteristic of depressive states [27]. Furthermore, adolescents experiencing anxiety and depression commonly exhibit difficulties in emotional regulation. These individuals are prone to frequent episodes of negative emotions and have a reduced tolerance for distress. When confronted with emotional upheaval, challenges in comprehending and accepting their emotions, coupled with a propensity to suppress impulsive behaviors, may play a significant role in the inception of NSSI [28].

This study provides a preliminary investigation into the role of various cytokines in the onset of non-suicidal self-injury among adolescents. The primary finding of this study is that elevated serum levels of IL-2 and IL-10 may be associated with NSSI behavior in adolescents, with this relationship appearing to be unaffected by the severity of anxiety and depressive symptoms. Inflammatory cytokines may play a role in activating the kynurenine pathway of tryptophan metabolism, dysregulating the hypothalamic-pituitary-adrenal axis, and altering monoamine metabolism. One potential mechanism involves the pathophysiology of NSSI, which is closely associated with 5-Hydroxytryptamine. On the one hand, the proinflammatory cytokine IL-2 activates indoleamine 2,3-dioxygenase (IDO), and when IDO activity increases, tryptophan, a precursor to serotonin, is depleted, leading to reduced serotonin levels [29]. On the other hand, activation of the kynurenine pathway can result in abnormal glutamate signaling via N-methyl-D-aspartate receptor activation, which negatively affects emotional regulation and impulse control [30]. The reduction in serotonin function may increase susceptibility to NSSI, particularly during periods of emotional distress [31]. Furthermore, adolescents engaging in NSSI exhibit an exaggerated cortisol awakening response, and IL-2 treatment increases plasma cortisol levels, accompanied by a temporal increase in plasma adrenocorticotropic hormone levels, thereby increasing the risk of NSSI behavior [32, 33].

IL-10, an anti-inflammatory cytokine, is typically thought to inhibit IDO activity and reduce the occurrence of NSSI. However, a notable finding in this study was the association between elevated serum IL-10 levels and an increased risk of self-injurious behavior, which contrasts with the expected patterns. Previous studies on IL-10 in patients with depression have yielded inconsistent results. Adolescents with varying suicide risk levels exhibited differences in plasma levels of pro-inflammatory cytokines and the anti-inflammatory factor IL-10, independent of depressive symptoms. Elevated IL-10 levels have been identified as a potential risk factor for suicidal behavior in depressed individuals [34]. Prior research has shown that increased serum IL-10 is associated with a higher kynurenine/tryptophan ratio (a marker of IDO activity) and lower serum tryptophan levels [35]. Our preliminary hypothesis suggests that immune changes induced by NSSI may lead to increases in both pro-inflammatory and anti-inflammatory cytokines, with pro-inflammatory cytokines being more dominant. This combined effect may enhance the activity of IDO and kynurenine 3-monooxygenase enzymes, depleting tryptophan and contributing to the occurrence of NSSI. As this is the first study to report the relationship between IL-10 and NSSI in a clinical sample, further validation in independent cohorts is required. Furthermore, this study employed an exploratory approach to investigate the associations between cytokine levels and latent class membership. Given the increased risk of false positives associated with multiple testing, readers are advised to interpret the findings with caution and consider them as preliminary observations and hypothesis-generating.

In this study, no other cytokines were found to be associated with the occurrence of NSSI. Previous studies have indicated that IL-6 levels in the NSSI group were not significantly elevated when compared to the healthy control group and the non-self-injury group, which is consistent with the findings of the present study [13, 15]. Previous research findings are conflicting and indicate that IL-6 is not a robust etiological marker of increased self-harm behavior [16]. In contrast to the findings of this study, previous research has shown that, compared to the non-NSSI group, the NSSI group exhibited a significant increase in TNF-α levels, while no notable abnormalities were observed in IL-1β levels [13]. We speculate that the differences in results may be attributed to variations in factors such as age, gender, BMI, medication use, sleep quality, type and duration of the disease, the presence of comorbid psychiatric symptoms, as well as differences in sampling and testing methods across studies. However, the contradictory results highlight the complexity and diversity of the inflammatory mechanisms associated with non-suicidal self-injury, emphasizing the need for further research to validate these findings.

Limitations

First, as a cross-sectional study, the associations observed in this research require further validation through longitudinal follow-up at the individual level in order to establish causal relationships. Second, the sampling strategy may have introduced selection bias. The use of convenience sampling could lead to the overrepresentation or underrepresentation of certain subgroups. Additionally, in an effort to reduce sample heterogeneity, individuals with current or past suicidal ideation were excluded, which may further impact the accuracy and generalizability of the findings. Third, all participants were recruited exclusively from Shandong Province, which may geographically constrain the generalizability of the results. Caution should be exercised when extrapolating these findings to adolescents in other regions of China. Fourth, reliance on self-report questionnaires introduces several potential limitations, including recall bias and fluctuations in emotional state at the time of assessment, which may result in underreporting or overreporting of symptoms and behaviors. Moreover, the inclusion of sensitive items may have influenced participants’ willingness to respond honestly, potentially affecting the validity of the data. Finally, although the ANSAQ is a validated and widely used instrument, it does not capture the full spectrum of NSSI behaviors, which may result in an underestimation of more severe or atypical forms of self-injury.

Conclusion

Classifying adolescents based on NSSI characteristics provides important clinical insights into the potential psychological and biological mechanisms underlying different NSSI behavior patterns. This study found that elevated levels of IL-2 and IL-10 increase the risk of engaging in NSSI involving multiple methods in adolescents, suggesting that inflammatory cytokines may serve as potential biomarkers for identifying high-risk adolescents. These findings contribute to a deeper understanding of the pathophysiological mechanisms of NSSI and offer a scientific foundation for the development of individualized prevention and intervention strategies.

Abbreviations

AIC

Akaike Information Criterion

aBIC

sample-size-adjusted Bayesian Information Criterion

ANSAQ

Adolescent non-suicidal self-injury behavior questionnaire

BIC

Bayesian Information Criterion

BLRT

Bootstrap likelihood ratio Test

BMI

Body mass index

CRP

C-reactive protein

CI

Confidence interval

DSM-5

Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition

HAMD

Hamilton Depression Scale

HAMA

Hamilton Anxiety Scale

ICD-10

Diagnosed according to the International Classification of Diseases, 10th Revision

IDO

Indoleamine 2,3-dioxygenase

IFN

Interferon

IL

Interleukin

LCA

Latent class analysis

LMR

Lo-Mendell-Rubin likelihood ratio test

OR

Odds ratio

SD

Standard deviation

TNF

Tumor necrosis factor

NSSI

Non-suicidal self-injury

Authors’ contributions

Xinqi Wang contributed to the conception and design of the study, conducted the data analyses, interpreted the results, and drafted the manuscript. Jingjing Xu, Wangwang Xu, and Yang Zhang were responsible for data collection. Jin Gao supervised the data analysis and provided substantial revisions to the manuscript. All authors have read and approved the final manuscript.

Funding

This research was funded by the Qingdao Science and Technology Benefit People Program, grant number 22-3-7-smjk-19-nsh.

Data availability

The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

It was approved by the Ethics Committee of Qilu Hospital (Qingdao), Cheeloo College of Medicine, Shandong University, under approval protocol number KYLL-2024023. In accordance with the Declaration of Helsinki, written informed consent was obtained from the parents or legal guardians of all underage participants.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.