Elevated homocysteine and non-suicidal self-injury in adolescent depression: possible sex-specific clinical associations

Abstract

Background

Non-suicidal self-injury (NSSI) has become increasingly prevalent among adolescents with major depressive disorder (MDD), yet its underlying biological mechanisms remain poorly understood. This study investigated the association between serum homocysteine and NSSI in adolescents with MDD, with emphasis on sex-specific effects.

Methods

This retrospective study enrolled 141 adolescents with MDD, categorized into NSSI (n = 57) and Non-NSSI (NNSSI, n = 84) groups. Serum homocysteine was measured from fasting samples at admission. Clinical symptoms were assessed using the Hamilton Depression Rating Scale (HAMD) and Hamilton Anxiety Rating Scale (HAMA). Analysis of covariance compared homocysteine between groups. Partial correlation and hierarchical regression analyses examined associations between homocysteine and clinical symptoms, evaluating sex moderation effects.

Results

Homocysteine was significantly elevated in the NSSI versus NNSSI group (14.57 ± 7.31 vs. 11.70 ± 2.77 µmol/L, P = 0.001). In the NNSSI group, homocysteine negatively correlated with diurnal variation (r = −0.232, P = 0.034). Hierarchical regression revealed distinct sex moderation patterns across groups. In the NNSSI group, significant sex-by-homocysteine interactions were identified for sleep disturbance (β = 0.376, P = 0.013) and somatic symptoms (β = 0.127, P = 0.019). In the NSSI group, interaction effects appeared only for diurnal variation (β = 0.056, P = 0.034).

Conclusions

Elevated homocysteine may serve as a potential biological marker for NSSI in adolescents with MDD. Associations between homocysteine and clinical symptoms may exhibit sex-specific patterns that differed markedly between NSSI and NNSSI groups.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13034-026-01144-y.

Introduction

Adolescent depression represents one of the most prevalent mental disorders worldwide, with epidemiological studies demonstrating a 12-month prevalence rate of approximately 11% and a lifetime prevalence reaching 20% [1]. Recent years have witnessed growing research interest in non-suicidal self-injury (NSSI) behaviors among adolescents with major depressive disorder, particularly given the rising incidence of NSSI [2, 3] and its established role as a significant risk factor for suicidal behavior [4]. NSSI is defined as the intentional, direct destruction of one’s own bodily tissue without suicidal intent, a behavior that lacks social or cultural acceptance [5]. NSSI constitutes an increasingly critical public health concern among adolescent populations. Recent global meta-analyses indicate that adolescent NSSI prevalence ranges from 7.5% to 46.5% [6]. This substantial variability reflects multiple methodological and contextual factors, including NSSI definitions, assessment instruments, sampling strategies, and geographical or cultural contexts [6, 7]. For example, studies excluding suicidal intent yielded higher prevalence estimates (18.3% vs. 11.3%), and Asian countries showed higher rates than other continents (19.5% vs. 14.6%) [6]. Notably, NSSI prevalence is consistently higher among female than male adolescents [7]. However, the neurobiological mechanisms underlying depression comorbid with NSSI across different sexes, as well as the associated sex-specific differences, remain poorly understood, warranting systematic research in adolescent populations.

Homocysteine (HCY), an α-amino acid representing a critical intermediate in methionine metabolism, has been recognized as a risk factor for cardiovascular disease, Parkinson’s disease, and cognitive disorders [8, 9]. However, emerging evidence has revealed that HCY is also implicated in psychiatric disorders, including schizophrenia and depression [10, 11]. Studies suggest that hyperhomocysteinemia may disrupt central nervous system function through multiple mechanisms, including the induction of neuroinflammation, excitotoxicity, and oxidative stress [12–14], pathological processes that may collectively compromise neural circuits involved in emotion regulation and impulse control. Research has demonstrated that plasma Hcy levels in adults with major depressive disorder (MDD) and bipolar depression are significantly elevated compared with healthy controls [15]. A 2021 meta-analysis of 34,873 participants revealed that serum homocysteine levels in depressed patients exceeded those of healthy controls by 2.53 µmol/L [16]. Furthermore, this meta-analysis demonstrated that hyperhomocysteinemia was associated with a 34% increased risk of depression [16]. A longitudinal study involving 1,460 middle-aged and elderly individuals demonstrated that elevated HCY levels were positively associated with depressive symptoms, with this association being more pronounced in women [17], thereby suggesting that sex may modulate the neurobiological effects of HCY. A recent large-scale study of MDD patients at Beijing Anding Hospital comparing 483 individuals with suicidal ideation/behavior (MDS group) to 483 individuals without a suicide history (MDNS group) identified serum folate and HCY as potential biomarkers for suicide risk, with notable sex and age-related variations [18], further supporting the modulatory role of sex in the neurobiological effects of HCY.

However, existing research has predominantly focused on the association between HCY and suicidal behavior, with the majority of studies conducted in adult populations [18–21]. To our knowledge, no studies have directly examined the sex-specific association between HCY and NSSI behavior in adolescents with depression. This study investigated the association between HCY levels and NSSI behavior among adolescent MDD patients across sexes, and to examine correlations with clinical characteristics as well as the potential moderating effects of sex. By systematically examining the association between HCY and NSSI behavior and its sex-specific patterns in adolescent MDD populations, this study aims to provide novel insights into the neurobiological mechanisms underlying NSSI.

Methods

Patient recruitment and study design

This retrospective study included 141 adolescent patients with major depressive disorder (MDD) aged 12 to 17 years. All eligible patients were consecutively enrolled from the pediatric psychiatry inpatient unit at Hainan Anning Hospital between June 2023 and June 2025. The participant selection process is detailed in Figure S1 (Supplementary Materials). All eligible patients were consecutively enrolled during the study period, and group assignment was based on clinical NSSI status rather than randomization. The unequal sample sizes between the NSSI (n = 57) and NNSSI (n = 84) groups reflect the natural distribution of NSSI prevalence among adolescent inpatients with MDD at our institution. All eligible participants were included to maximize statistical power and to avoid selection bias that would arise from excluding NNSSI participants. Diagnoses were established through semi-structured interviews conducted by two experienced psychiatrists according to the DSM-5 criteria for MDD.

Inclusion criteria comprised: (1) fulfillment of DSM-5 diagnostic criteria for major depressive disorder; (2) age between 12 and 17 years; (3) a 17-item Hamilton Depression Rating Scale (HAMD-17) total score ≥ 20, indicative of moderate to severe depression; and (4) a Young Mania Rating Scale (YMRS) total score < 7. Exclusion criteria comprised: (1) a diagnosis of schizophrenia, bipolar disorder, or intellectual disability; (2) prior receipt of systematic psychotherapy or electroconvulsive therapy; (3) a history of neurological disorders or organic brain disease; (4) a history of alcohol or substance abuse; and (5) any medical comorbidity identified through personal history, clinical, or laboratory examination, including pregnancy or postpartum depression.

All enrolled patients had received antidepressant medication prior to admission as part of outpatient management. To account for potential confounding effects, antidepressant dosage (standardized to fluoxetine equivalents) was included as a covariate in all statistical analyses.

This retrospective study received ethical approval from the Ethics Committee of Suzhou Guangji Hospital (Su Guanglun, No. 2023-016) and adhered to the principles outlined in the Declaration of Helsinki. Given the retrospective nature of this investigation, the requirement for informed consent was waived by the Ethics Committee. Confidentiality of all patient data was strictly maintained throughout the study.

Data collection

Demographic and clinical data were collected using standardized case report forms, including age, sex, body mass index (BMI), and disease duration. BMI was calculated using the standard anthropometric formula: BMI = weight (kg) / height² (m²). Disease duration was defined as the time interval from symptom onset to the date of assessment, expressed in months. Data were collected through the hospital’s electronic medical record system, which maintains comprehensive longitudinal records, including baseline characteristics, disease trajectory, and treatment interventions. Double data entry was performed by two independent researchers to ensure data accuracy and completeness.

Clinical assessment

Hamilton Anxiety Scale (HAMA)

The 14-item Hamilton Anxiety Scale (HAMA) was employed to assess anxiety symptom severity. The HAMA assesses both psychic and somatic anxiety manifestations, yielding a total score and two subscale scores: psychic anxiety and somatic anxiety. Each item is rated on a 5-point scale (0–4), yielding a total score ranging from 0 to 56, with higher scores reflecting greater anxiety severity.The HAMA was originally developed by Hamilton [22] and has demonstrated good reliability and validity in Chinese psychiatric populations [23]. All HAMA assessments were administered by trained, board-certified psychiatrists who had undergone standardized training in clinical scale administration, demonstrating good inter-rater reliability (intraclass correlation coefficient > 0.85).

Hamilton Depression Rating Scale (HAMD-17)

The 17-item Hamilton Depression Rating Scale (HAMD-17) was employed to assess depressive symptom severity. The HAMD-17 assesses multiple dimensions of depression, yielding a total score and eight subscale scores: anxiety/somatization, weight, cognitive impairment, diurnal variation, psychomotor retardation, sleep disturbance, feelings of hopelessness, and general somatic symptoms. Items are rated on either a 3-point (0–2) or 5-point (0–4) scale depending on the item, with higher scores reflecting greater depression severity [24, 25]. The Chinese version of the HAMD-17 has demonstrated good reliability and validity in Chinese psychiatric populations [26, 27]. All HAMD-17 assessments were conducted by board-certified psychiatrists who had undergone standardized training.

Homocysteine measurement

Early morning fasting venous blood samples (3–5 mL) were collected from the elbow vein of all participants upon admission into vacuum procoagulant tubes. The samples were centrifuged at 3000–4000 rpm for 10 min at room temperature to separate the serum.

Serum HCY concentrations were determined using a Hitachi 7180 automated biochemical analyzer (Hitachi High-Technologies Corporation, Tokyo, Japan) employing the enzymatic cycling assay method. All measurements were performed in the clinical laboratory at Hainan Anning Hospital following standardized operating procedures. The intra-assay and inter-assay coefficients of variation were less than 3% and 5%, respectively, which met established clinical laboratory standards. Serum HCY concentrations were determined on the same day of blood collection as part of routine clinical biochemistry testing at admission; no stored or archived serum samples were re-analyzed. HCY data were subsequently extracted from the hospital laboratory information system (LIS) for research purposes. The laboratory reference range for serum homocysteine was as follows: HCY: 4 to 15.4 µmol/L.

Non-suicidal self-injury assessment

Consistent with DSM-5 criteria, non-suicidal self-injury (NSSI) was defined as intentional self-harm behavior without suicidal intent [28]. NSSI history was ascertained by two board-certified psychiatrists upon admission through direct clinical interviews and medical record review. Participants were stratified into two groups based on NSSI history: those with at least one self-harm episode within the preceding month (NSSI group) and those without such history (NNSSI group).

Antidepressant dosage (fluoxetine equivalents)

Current antidepressant regimens were documented at enrollment, and all dosages were standardized to fluoxetine equivalents (mg/day) using established conversion factors. Fluoxetine equivalent dosages were calculated according to the methodology established by Furukawa, T. A. et al. [29]. Detailed information regarding antidepressant medication classes, specific agents, dosage ranges, and the distribution across groups is provided in Supplementary Table S1.

Statistical analysis

All statistical analyses were conducted using SPSS version 25.0 and Python version 3.9.13. All statistical tests were two-tailed, and P < 0.05 was considered to indicate statistical significance for primary analyses. For exploratory analyses involving multiple comparisons (partial correlations and moderation analyses), Bonferroni correction was applied, with adjusted significance thresholds specified in the respective subsections.

Descriptive statistics

The normality of continuous variables was assessed using the Shapiro-Wilk test and visual inspection of Q-Q plots. Continuous variables with normal distributions were expressed as mean ± standard deviation (mean ± SD), whereas non-normally distributed variables were expressed as median (interquartile range) [M(IQR)]. Categorical variables were expressed as frequencies and percentages.

Group comparisons

Between-group comparisons of continuous variables were conducted using independent samples t-tests for normally distributed data and Mann-Whitney U tests for non-normally distributed data. Data were considered normally distributed if the Shapiro-Wilk test yielded P > 0.05. Categorical variables were compared between groups using chi-square tests. Effect sizes were calculated to assess the magnitude of group differences, with Cohen’s d employed for parametric tests and rank-biserial correlation for nonparametric tests.

To account for potential confounding variables, analysis of covariance (ANCOVA) was employed to compare serum HCY levels between groups, with covariates including age, BMI, disease duration, and fluoxetine equivalent dosage. The assumption of homogeneity of regression slopes was verified prior to ANCOVA, and partial eta squared (partial η²) was used as the effect size measure.

Correlation analysis

Partial correlation analyses were conducted separately in the NSSI and NNSSI groups to examine associations between serum HCY levels and clinical symptoms, adjusting for potential confounding factors (age, sex, BMI, disease duration, and fluoxetine equivalent dosage). Clinical variables included the HAMA total score and its two subscales (psychic and somatic anxiety), as well as the HAMD-17 total score and its eight subscales (anxiety/somatization, weight, cognitive impairment, diurnal variation, psychomotor retardation, sleep disturbance, feelings of hopelessness, and general somatic symptoms), yielding a total of 12 variables. Partial correlation coefficients (r) and their respective P values were calculated for each group. Forest plots were generated to display partial correlation coefficients and their 95% confidence intervals. To account for Type-I error inflation arising from multiple comparisons, Bonferroni correction was applied across all 12 clinical symptom dimensions examined within each group, yielding an adjusted significance threshold of α/12 = 0.0042. Bonferroni-corrected results are reported in Supplementary Table S2.

Moderation analysis

To examine whether sex moderated the association between HCY levels and clinical symptoms differentially across the NSSI and NNSSI groups, hierarchical multiple linear regression analyses were conducted separately within each group. Separate regression models were constructed for each of the 12 clinical variables, with each variable serving as the dependent variable. Predictor variables were entered hierarchically as follows: Step 1 included control variables (age, BMI, disease duration, and fluoxetine equivalent dosage); Step 2 added the main effects of sex (coded as 0 = male, 1 = female) and mean-centered HCY levels; Step 3 introduced the sex×HCY interaction term. The significance of the interaction terms was evaluated using t-statistics and their respective P values. For each model, unstandardized regression coefficients (β), standard errors (SE), t values, P values, and R² values were reported. Similarly, Bonferroni correction was applied across the 12 moderation models within each group (adjusted threshold: α/12 = 0.0042), and corrected results are reported in Supplementary Table S3.

Results

Demographic and clinical characteristics of study participants

This study included 141 adolescent patients with depression, comprising 84 participants in the NNSSI group and 57 in the NSSI group. The NNSSI group consisted of 58 females and 26 males, whereas the NSSI group consisted of 40 females and 17 males. Sex distribution did not differ significantly between groups (χ²=0.000, P = 1.000).

No significant differences were observed between groups in baseline characteristics, including age, BMI, disease duration, or medication dosage (all P > 0.05). Regarding clinical symptom severity, HAMA assessment revealed no significant differences between the NSSI and NNSSI groups in total score (22.02 ± 8.71 vs. 20.44 ± 8.99, P = 0.302), the somatic anxiety subscale (9.44 ± 5.04 vs. 9.13 ± 5.32, P = 0.731), or the psychic anxiety subscale (12.58 ± 4.33 vs. 11.31 ± 4.27, P = 0.087). Similarly, HAMD-17 assessment demonstrated no significant differences between groups in the total score (31.56 ± 9.29 vs. 29.49 ± 10.25, P = 0.223) or any subscale scores (anxiety/somatization, weight, cognitive impairment, diurnal variation, psychomotor retardation, sleep disturbance, hopelessness, and general somatic symptoms) (all P > 0.05) (Table 1).

Table 1.

Baseline characteristics of adolescents with MDD stratified by NSSI status

Variable		NSSI group (n = 57)	Non-NSSI group (n = 84)	Statistic	P value
Age		14.81 ± 1.75	14.65 ± 1.78	t = 0.502a	0.616
BMI (kg/m²)		20.45 ± 4.13	20.70 ± 4.64	t = −0.325a	0.746
Sex	Male	17	26	χ² = 0.000b	1.000
	Female	40	58
Disease duration (months)		28.53 ± 21.78	27.47 ± 22.14	t = 0.280a	0.780
Fluoxetine equivalent dosage (mg/day)		27.95 ± 14.15	29.11 ± 10.13	t = −0.566a	0.573
HAMA total score		22.02 ± 8.71	20.44 ± 8.99	t = 1.035a	0.302
Somatic anxiety		9.44 ± 5.04	9.13 ± 5.32	t = 0.344a	0.731
Psychic anxiety		12.58 ± 4.33	11.31 ± 4.27	t = 1.724a	0.087
HAMD total raw score		31.56 ± 9.29	29.49 ± 10.25	t = 1.224a	0.223
Anxiety/Somatization		6.40 ± 2.43	5.76 ± 2.48	t = 1.520a	0.131
Weight		0.56 ± 0.68	0.58 ± 0.64	t = -0.194a	0.847
Cognitive Impairment		8.00 ± 2.96	7.60 ± 3.17	t = 0.764a	0.446
Diurnal variation		0.61 ± 0.70	0.67 ± 0.65	t = −0.459a	0.647
Retardation		5.89 ± 2.01	5.62 ± 1.86	t = 0.838a	0.404
Sleep disturbance		4.63 ± 1.40	4.14 ± 1.67	t = 1.822a	0.071
Feelings of hopelessness		4.35 ± 1.99	3.95 ± 2.38	t = 1.040a	0.300
General somatic symptoms		1.18 ± 0.57	1.15 ± 0.61	t = 0.202a	0.840
Serum homocysteine (µmol/L)		14.57 ± 7.31	11.70 ± 2.77	t = 3.272a	0.001

NSSI = non-suicidal self-injury; BMI = body mass index; HAMA = Hamilton Anxiety Rating Scale; HAMD = Hamilton Depression Rating Scale

Significant differences (P < 0.05) are indicated in bold. Antidepressant classes and agent-specific dosage ranges are detailed in Supplementary Table S1

^a t-test;

^b χ² test

Serum HCY levels in the NSSI group (14.57 ± 7.31 µmol/L) were significantly elevated compared with the NNSSI group (11.70 ± 2.77 µmol/L) (t = 3.272, P = 0.001, Cohen’s d = 0.548). After adjusting for potential confounding factors (including age, body mass index, disease duration, and fluoxetine equivalent dosage), analysis of covariance (ANCOVA) confirmed a significant difference between groups (F = 10.058, P = 0.002).

Correlation analysis between HCY and clinical symptoms in the NNSSI and NSSI groups

To examine the association between HCY levels and clinical symptoms in adolescent patients with depression, partial correlation analyses were conducted separately in the NNSSI and NSSI groups after adjusting for age, sex, body mass index, disease duration, and fluoxetine equivalent dosage (Fig. 1). The analysis encompassed three HAMA measures (total score, psychic anxiety, and somatic anxiety) and nine HAMD-17 measures (total score, anxiety/somatization, weight, cognitive impairment, diurnal variation, psychomotor retardation, sleep disturbance, feelings of hopelessness, and general somatic symptoms), yielding a total of 12 clinical variables.

Fig. 1.

Forest plot depicting partial correlation analysis between serum HCY levels and clinical symptoms in the NNSSI and NSSI groups. Square markers represent the point estimates of partial correlation coefficients (r), while horizontal error bars indicate their corresponding 95% confidence intervals (CIs). The dashed line represents the zero-correlation reference line. Green circles denote statistically significant correlations (P < 0.05), whereas gray squares denote non-significant correlations (P ≥ 0.05). All analyses were adjusted for age, sex, body mass index (BMI), disease duration, and fluoxetine-equivalent dose

In the NNSSI group (n = 84), partial correlation analysis revealed a significant negative association between HCY levels and the diurnal variation subscale (r = −0.232, P = 0.034, 95% CI: −0.429 to −0.018), which persisted after adjusting for age, body mass index, disease duration, and fluoxetine equivalent dosage. No other clinical variables in this group demonstrated significant associations with HCY levels (all P > 0.05). In the NSSI group (n = 57), none of the 12 clinical variables demonstrated significant associations with HCY levels (all P > 0.05). After Bonferroni correction (α/12 = 0.0042), none of the partial correlation findings retained statistical significance. Bonferroni-corrected results are presented in Supplementary Table S2.

Effects of sex × HCY interaction on clinical symptoms in the NNSSI and NSSI groups

To investigate whether the moderating effect of sex on the relationship between HCY levels and clinical symptoms differs according to NSSI status, stratified analyses of the sex × HCY interaction were conducted separately in the NNSSI and NSSI groups, after adjusting for age, body mass index (BMI), disease duration, and fluoxetine-equivalent dose. Among the 12 clinical symptom dimensions examined, the stratified regression analyses identified three significant sex × HCY interactions (P < 0.05), suggesting that the moderating pattern of sex on HCY effects differed between the two groups.

In the NNSSI group (n = 84), two significant interactions were identified. The sleep disturbance factor demonstrated a significant sex × HCY interaction (β = 0.3761, SE = 0.1471, t = 2.5575, P = 0.0125; 95% CI [0.0832, 0.6690]), with the model explaining 14.99% of the variance. Similarly, the general somatic symptoms factor exhibited a significant sex × HCY interaction (β = 0.1273, SE = 0.0531, t = 2.3977, P = 0.0190; 95% CI [0.0216, 0.2331]). Both interactions were positive in direction.

In the NSSI group (n = 57), only the diurnal variation factor showed a significant sex × HCY interaction (β = 0.0561, SE = 0.0257, t = 2.1802, P = 0.0341; 95% CI [0.0044, 0.1078]). This positive interaction indicated that, in the NSSI group, the association between diurnal rhythm disturbance and HCY levels was significantly stronger in female patients than in male patients. Notably, the significant interaction effect in the NSSI group was exclusively observed for the diurnal variation factor, whereas no significant interactions were detected for the sleep disturbance or general somatic symptoms factors (Table 2). After Bonferroni correction (α/12 = 0.0042), none of the sex × HCY interaction terms retained statistical significance; corrected results are presented in Supplementary Table S3.

Table 2.

Sex × HCY interaction effects on clinical symptoms in NNSSI and NSSI groups

Clinical scale	NNSSI group (n = 84)				NSSI group (n = 57)
	β	SE	t	P	β	SE	t	P
HAMA total score	− 0.4110	0.8100	− 0.5073	0.613	0.3132	0.3359	0.9325	0.356
Somatic anxiety	− 0.3295	0.4837	− 0.6811	0.498	0.1062	0.1932	0.5495	0.585
Psychic anxiety	− 0.0815	0.3847	− 0.2118	0.833	0.2070	0.1667	1.2422	0.221
HAMD total score	0.3718	0.9375	0.3966	0.693	0.4133	0.3460	1.1944	0.238
Anxiety/somatization	− 0.0911	0.2265	− 0.4023	0.689	0.0981	0.0919	1.0669	0.291
Weight	0.0200	0.0592	0.3379	0.736	0.0234	0.0259	0.9005	0.372
Cognitive impairment	0.2612	0.2885	0.9051	0.368	0.0882	0.1087	0.8118	0.421
Diurnal variation	− 0.0786	0.0566	− 1.3892	0.169	0.0561	0.0257	2.1802	0.034*
Retardation	− 0.0964	0.1690	− 0.5704	0.570	0.1125	0.0742	1.5159	0.136
Sleep disturbance	0.3761	0.1471	2.5575	0.013*	0.0102	0.0520	0.1960	0.845
Hopelessness	− 0.1370	0.2155	− 0.6358	0.527	0.0328	0.0763	0.4295	0.670
General somatic symptoms	0.1273	0.0531	2.3977	0.019*	− 0.0003	0.0207	− 0.0122	0.990

All models were adjusted for age, body mass index (BMI), disease duration, and fluoxetine-equivalent dose. The interaction coefficient β represents the unstandardized regression coefficient for the sex × HCY interaction term. *P < 0.05. SE, standard error

Discussion

To the best of our knowledge, this study is the first to investigate the association between serum HCY levels and non-suicidal self-injury (NSSI) behavior, along with its sex-specific characteristics, in adolescents with major depressive disorder (MDD). The key findings are as follows. First, after controlling for multiple confounding factors, serum HCY levels were significantly elevated in adolescents with NSSI behavior compared with those without, suggesting that HCY may serve as a potential biological marker for NSSI behavior. Second, the correlation patterns between HCY levels and clinical symptoms differed markedly between groups. HCY levels showed a negative correlation with diurnal variation in the NNSSI group, whereas no significant overall correlations were identified in the NSSI group. Third, the moderating effect of sex on HCY-symptom associations exhibited distinct characteristics across groups: in the NNSSI group, sex moderated the associations between HCY and both sleep disturbance and somatic symptoms, while in the NSSI group, sex moderated only the association between HCY and diurnal rhythm. These findings provide new insights into the neurobiological mechanisms underlying NSSI behavior in adolescents with depression.

In the present study, we found that adolescents with MDD who engaged in NSSI behavior exhibited significantly higher serum HCY levels compared with those without NSSI behavior. Previous studies have reported similar findings. A Turkish case-control study found that serum HCY levels were significantly higher in the suicide attempt group than in the control group [19]. An Italian study demonstrated that higher HCY levels were significantly associated with higher suicidal ideation scores, independent of depression severity [20]. Furthermore, a study conducted at Beijing Anding Hospital involving 1,115 patients with MDD identified an age-related heterogeneous association between HCY levels and suicide risk [18]. The Korean MAKE BETTER prospective cohort study, which followed 884 patients with depression for 12 months, found that baseline HCY levels exceeding 11.1 µmol/L were significantly associated with increased suicide severity (OR = 1.60) [21]. However, to the best of our knowledge, no studies have specifically examined HCY levels in adolescents with depression presenting with NSSI behavior. In the present study, adolescents with depression who exhibited NSSI behavior demonstrated elevated serum HCY levels. Based on the existing literature, we propose that elevated HCY levels may be associated with NSSI behavior through the following mechanisms (Fig. 2). First, hyperhomocysteinemia may induce oxidative stress and neuroinflammatory cascades. HCY accumulation may promote excessive generation of reactive oxygen species (ROS), potentially leading to oxidative damage in neurons and vascular endothelial cells [13]; such oxidative stress has been hypothesized to activate microglia and astrocytes, possibly triggering the release of proinflammatory cytokines such as IL-6, TNF-α, and C-reactive protein (CRP) [12, 30]. Given that previous studies have reported an association between NSSI behavior and elevated peripheral inflammatory markers [30], inflammatory responses may represent one potential pathway linking elevated HCY levels to NSSI. However, this hypothesis requires direct validation in adolescent populations. Second, aberrant HCY metabolism may also influence NSSI behavior through disruption of neurotransmitter homeostasis. Dysregulation of HCY metabolic pathways may result in decreased production of S-adenosylmethionine (SAM), the primary methyl donor in the body, potentially affecting the synthesis and metabolism of key neurotransmitters such as serotonin, dopamine, and norepinephrine [12]. Studies in adults suggest that elevated HCY levels may be associated with reduced serotonin function, and serotonergic system dysfunction has been implicated in increased impulsivity and impaired emotion regulation [31]. Third, based on in vitro and animal studies, HCY may induce excitotoxicity through activation of N-methyl-D-aspartate (NMDA) receptors [13], which could theoretically affect the functional integrity of frontal-limbic circuits that play a crucial role in impulse control and emotion regulation [32]. However, it should be emphasized that these three mechanistic pathways are primarily based on adult studies, in vitro experiments, or animal models, and direct evidence in adolescent populations with NSSI remains lacking. The cross-sectional design of the present study precludes the establishment of causal relationships between elevated HCY levels and NSSI behavior, nor does it permit validation of the aforementioned mechanistic hypotheses. Future studies integrating neuroimaging, metabolomics, and prospective designs are needed to validate these theoretical propositions.

Fig. 2.

Proposed pathophysiological mechanisms potentially associated with elevated homocysteine in adolescent major depressive disorder with non-suicidal self-injury behavior. This diagram illustrates three hypothesized pathophysiological pathways through which elevated homocysteine (HCY) might be associated with non-suicidal self-injury (NSSI) behavior. Pathway 1 depicts the hypothesized oxidative stress and neuroinflammatory cascade, showing how HCY-induced reactive oxygen species (ROS) generation might cause cellular damage, potentially activate microglia, and possibly trigger pro-inflammatory cytokine release (IL-6, TNF-α, CRP). Pathway 2 illustrates the neurotransmitter dysregulation hypothesis, demonstrating how SAM depletion may impair the synthesis of serotonin (5-HT), dopamine (DA), and norepinephrine (NE), potentially leading to synaptic dysfunction (shown with presynaptic terminal, synaptic cleft, and postsynaptic receptors) and possibly culminating in behavioral and emotional dysregulation (depicted as a distressed face with tears). Pathway 3 demonstrates the excitotoxicity mechanisms, where NMDA receptor activation may cause calcium overload, potentially disrupt neural networks, and contribute to frontal-limbic circuit damage. All three pathways are hypothesized to converge and may collectively be associated with increased NSSI behavior risk. The mechanisms depicted represent current theoretical understanding based on available evidence and require further validation through prospective studies. HCY homocysteine, ROS reactive oxygen species, IL-6 interleukin-6, TNF-α tumor necrosis factor-alpha, CRP C-reactive protein, SAM S-adenosylmethionine, 5-HT serotonin, DA dopamine, NE norepinephrine, NMDA-R N-methyl-D-aspartate receptor, Ca²⁺ calcium ion, NSSI non-suicidal self-injury

The present study revealed markedly different correlation patterns between HCY levels and clinical symptoms across the two patient groups. In the NNSSI group, serum HCY levels demonstrated a significant negative correlation with the diurnal variation factor, indicating that higher HCY levels were associated with less severe diurnal rhythm disturbances in adolescents with depression who did not engage in self-injury. Ozdogan et al. observed an association between hyperhomocysteinemia and circadian rhythm patterns in adults with bipolar disorder [27], suggesting that HCY may be involved in the regulation of biological rhythm systems. The interpretation of the negative correlation pattern observed in the present study (i.e., higher HCY levels associated with less severe diurnal rhythm disturbances) remains unclear. This finding may reflect several possibilities: (1) the presence of unknown adaptive or compensatory mechanisms in patients without NSSI behavior; (2) the influence of unmeasured confounding factors; or (3) a statistical false positive. Given the limited sample size and cross-sectional design of the present study, the biological significance of this finding requires further validation in prospective studies with larger sample sizes. Theoretically, aberrant HCY metabolism may affect circadian rhythm systems through multiple pathways: decreased SAM levels may influence melatonin synthesis [28]. Oxidative stress and excitotoxicity may impair suprachiasmatic nucleus function [29]. However, these mechanistic hypotheses are primarily based on in vitro studies and animal models, and their relevance to adolescents with depression has not been directly validated. More importantly, the negative correlation direction we observed (i.e., elevated HCY associated with reduced rhythm symptoms) contradicts theoretical expectations, further suggesting that this finding may be influenced by uncontrolled confounding factors or may involve more complex regulatory mechanisms. Therefore, this result should be interpreted with caution, and future studies are needed to validate this finding.

No overall significant correlations between HCY levels and clinical symptoms were detected in the NSSI group. This null finding should be interpreted within the framework of sex stratification. Sex moderation analyses revealed that sex significantly moderated the relationship between HCY levels and diurnal variation in the NSSI group, suggesting that the absence of overall correlations may be partially attributable to the offsetting effects of sex-specific heterogeneity. Sex moderation analyses demonstrated between-group differences: in the NNSSI group, sex moderated the relationships between HCY levels and both sleep disturbance and somatic symptoms, whereas in the NSSI group, sex moderated only the relationship between HCY levels and diurnal rhythm. These divergent patterns suggest that NSSI status may alter the sex-specific effects of HCY. However, it should be noted that the sample size of the present study, particularly that of the sex-stratified subgroups, was limited (NSSI group: females, n = 40; males, n = 17), and some interaction effects may not have been detected due to insufficient statistical power. Therefore, these findings require validation in larger samples, and their clinical significance needs to be clarified through prospective studies. Potential mechanisms underlying the sex differences may include the following: (1) the effects of estrogen on methylenetetrahydrofolate reductase (MTHFR) enzyme activity and HCY metabolism [33, 34]; (2) the expression of estrogen receptors in the suprachiasmatic nucleus and their regulation of circadian rhythm genes [35, 36]; (3) sex differences in hypothalamic-pituitary-adrenal (HPA) axis function [37]. However, these mechanistic hypotheses are primarily based on NNSSI populations or animal studies, and their applicability to adolescents with depression, particularly those presenting with NSSI, remains unclear. The present study did not measure sex hormone levels, MTHFR genotypes, or HPA axis function indices; therefore, it was unable to directly validate these mechanistic hypotheses. Future studies should integrate genomic, endocrinological, and neuroimaging data to more comprehensively elucidate the biological basis of sex moderation of HCY effects.

Several important limitations of the present study should be acknowledged. First, the cross-sectional study design represents the most significant limitation, precluding the establishment of temporal sequence and causal relationships between elevated HCY levels and NSSI behavior. We cannot determine whether elevated HCY levels represent a cause, consequence, or correlate of NSSI behavior attributable to a shared third factor. Although the present study observed elevated HCY levels in the NSSI group and between-group differences in the correlation patterns between HCY levels and clinical symptoms, the causal direction among elevated HCY levels, NSSI behavior, and depressive symptoms cannot be determined. Longitudinal or prospective studies are essential for exploring potential causal relationships. Second, the relatively small sample size may have limited the statistical power to detect significant effects, particularly in sex-stratified moderation analyses. Third, as a single-center study with samples drawn from a single medical institution in Hainan Province, the study may be subject to selection bias, limiting the generalizability of the findings. Fourth, all participants had received antidepressant medication treatment; although fluoxetine-equivalent doses were controlled for, medication classes (e.g., selective serotonin reuptake inhibitors [SSRIs], serotonin-norepinephrine reuptake inhibitors [SNRIs], mirtazapine) were not differentiated, and these medications may have differential effects on HCY metabolism. Additionally, factors such as treatment duration and medication adherence may also influence HCY levels and clinical symptoms, but the present study did not collect detailed information on these variables. Moreover, the exclusive enrollment of inpatients may have introduced selection bias toward more severe presentations, limiting generalizability to outpatient populations; future studies incorporating drug-naïve and outpatient samples would help clarify whether medication status confounds the observed HCY–NSSI associations. Fifth, the present study measured only HCY as a single biomarker, without simultaneously assessing upstream and downstream indicators of its metabolic pathway (e.g., folate, vitamin B12, vitamin B6, SAM), genetic factors (e.g., MTHFR gene polymorphisms), inflammatory markers (e.g., IL-6, TNF-α, CRP), and sex hormone levels, or nutritional status indicators (e.g., serum albumin, dietary intake assessments). Given that adolescent nutritional habits may influence HCY metabolism through folate and B-vitamin pathways, future studies should incorporate comprehensive nutritional assessments. Furthermore, culture-specific factors — including China’s lack of mandatory staple-food folic acid fortification and the high academic psychosocial stress burden among Chinese adolescents — may have additionally influenced HCY levels in our sample [38, 39]. These factors may represent potential confounders or mediators of the association between HCY levels and NSSI, and their absence limits our understanding of the underlying mechanisms. Sixth, homocysteine levels in the present study were measured exclusively in peripheral blood (serum). Although prior research has demonstrated correlations between peripheral and cerebrospinal fluid HCY concentrations, peripheral HCY levels may not directly reflect central nervous system HCY metabolism or CNS-specific pathophysiological processes. Future studies incorporating cerebrospinal fluid sampling or neuroimaging-based metabolic assessments would provide more direct evidence regarding CNS-related mechanisms. Seventh, the absence of a healthy control group precluded our ability to assess the magnitude of HCY elevation relative to age-matched healthy adolescents, or to determine whether the observed HCY levels represent a depression-general or NSSI-specific phenomenon. Future studies should include a healthy control group to enable three-way comparisons (NSSI vs. NNSSI vs. healthy controls). Furthermore, the unequal sample sizes between the NSSI (n = 57) and NNSSI (n = 84) groups, reflecting the natural prevalence distribution, may have introduced differential statistical power across group-specific analyses. Future studies with balanced designs or propensity score matching could help address this limitation. In addition, the binary classification of NSSI (present vs. absent) does not capture variations in frequency, severity, or chronicity; future studies employing dimensional NSSI measures would provide more nuanced insights into the HCY–NSSI relationship. Finally, although Bonferroni correction was applied across the 12 clinical symptom dimensions in both partial correlation and moderation analyses (adjusted α = 0.0042), none of the exploratory findings retained statistical significance after correction, whereas the primary between-group comparison of HCY levels remained significant. These results underscore the need for caution in interpreting the exploratory correlation and moderation findings, which should be considered hypothesis-generating and require replication in larger, independent samples. Future prospective studies should explore these associations in larger, multicenter cohorts, integrating genomic, metabolomic, and neuroimaging data, and employ structural equation modeling (SEM) or mediation analysis frameworks with comprehensive biomarker panels to formally test the hypothesized causal pathways depicted in Fig. 2 and more comprehensively elucidate the mechanistic role of HCY in adolescent depression and NSSI.

Conclusion

This study demonstrated that serum homocysteine levels were significantly elevated in adolescents with major depressive disorder (MDD) exhibiting non-suicidal self-injury (NSSI), suggesting a potential biomarker. Sex-specific associations with clinical symptoms and distinct group differences provide insights into NSSI neurobiological mechanisms, though prospective validation is warranted.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

P.C. and H.Z. wrote the main manuscript text. X.Z. performed manuscript review and editing. P.C., J.L., Y.Y., W.S., and Q.C. conducted formal analysis. Y.Y. and W.S. provided supervision. D.Z. provided resources. X.Z. and H.Z. acquired funding. All authors reviewed the manuscript.

Funding

This work was supported by Suzhou Clinical Key Disciplines for Geriatric Psychiatry (SZXK202116), Suzhou Clinical Medical Center for Mood Disorders (Szlcyxzx202109), Suzhou Key Technologies Program (SKY2021063), Key Laboratory in Science and Technology Development Project of Suzhou (SZS2024016), Suzhou Municipal Science and Technology Bureau(SYW2025022, SKY2023227), Suzhou Clinical Key Disease Diagnosis and Treatment Technology Specialization(LCZX202218), Suzhou University Open Research Program(KJS2528), Lanzhou Science and Technology Bureau(2022-ZD-57), Health Commission of Gansu Province(GSWSKY2022-38), Nanjing Medical University University-Level Research Project(NMUB20250085), and Suzhou Municipal Health Young Backbone Talent Training Program with Nationwide Mentorship(Qngg2024027). These sources had no further role in the study design, data collection and analysis, writing of the report, or in the decision to submit the paper for publication.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This retrospective study received ethical approval from the Ethics Committee of Suzhou Guangji Hospital (Approval No. Su Guanglun 2023-016) and was conducted in accordance with the principles of the Declaration of Helsinki. Given the retrospective nature of the study, the requirement for written informed consent was waived by the Ethics Committee. Confidentiality of all patient data was strictly maintained throughout the study.

Competing interests

The authors declare no competing interests.