Blood Clues to Teen Suicidal Thoughts Diverge Sharply Between Boys and Girls

Suicidal ideation among adolescents with major depressive disorder is one of the most urgent and elusive problems in psychiatry. Clinicians rely almost entirely on self-report and clinical interviews, yet suicidal thoughts are frequently underreported in routine care, leaving families and clinicians without objective tools to gauge risk. A new study published in BMC Psychiatry by Xinyu Zhang, Zhen Cao and colleagues at the First Affiliated Hospital of Xi’an Jiaotong University set out to ask whether the blood of depressed teenagers carries measurable signatures of suicidal ideation, and whether those signatures differ between boys and girls. The answer, based on 319 adolescents, is that they do, and the sex differences are striking.

The research team enrolled 319 adolescents diagnosed with major depressive disorder according to DSM-5 criteria, of whom 98 reported suicidal ideation and 221 did not. Suicidal ideation status was defined using the third item of the 17-item Hamilton Depression Rating Scale, a widely used clinician-administered instrument. The investigators then drew on two complementary classes of peripheral biomarkers: immune markers measured through complete blood counts and inflammatory indices, and metabolic markers including fasting plasma glucose, glycated hemoglobin, lipid fractions and derived indices such as the triglyceride-glucose index. This dual approach reflects a growing view in biological psychiatry that suicidal behavior cannot be understood through neural circuits alone but involves whole-body physiology, including chronic low-grade inflammation and altered energy metabolism.

The first major finding concerned metabolism. Adolescents with suicidal ideation showed a hypometabolic profile, characterized by lower glucose and lower lipid levels compared with depressed peers who did not report suicidal thoughts. In other words, the young people with suicidal ideation were not showing the elevated glucose and dyslipidemia often described in adults with depression; instead, their circulating energy substrates were shifted downward. This pattern is intriguing because the adolescent brain is in a period of intense metabolic demand, and a relative scarcity of circulating fuel could interact with the neurobiological stress systems implicated in suicidal thinking. The authors emphasize, however, that these are cross-sectional associations and do not establish whether the metabolic shift is a cause, a consequence, or a correlate of suicidal ideation.

The second major finding concerned inflammation. The suicidal ideation group exhibited elevated white blood cell counts, elevated neutrophil counts and proportions, and higher values on derived inflammatory indices such as the neutrophil-to-lymphocyte ratio, the systemic immune-inflammation index and the pan-immune-inflammation value. These indices, routinely calculated from standard blood counts, integrate information about the relative abundance of different immune cell populations and are increasingly used as accessible proxies of systemic inflammation. Elevated neutrophil-to-lymphocyte ratios in particular have been linked to inflammatory activation, and this study extends that association to suicidal ideation in depressed adolescents, a population in which such biological correlates had been poorly characterized.

The most consequential results emerged when the researchers stratified their analyses by sex. In female patients, the dominant signal was inflammatory: alterations in white blood cells, neutrophils and derived inflammatory indices predominated among girls with suicidal ideation. In male patients, the picture was reversed, with metabolic disturbances, including the changes in glucose and lipid measures, standing out more prominently. This sex-specific divergence matters because adolescence is precisely the period when sex differences in depression prevalence, symptom presentation and suicide risk widen, yet most biomarker studies pool the sexes or treat sex as a nuisance variable to be adjusted away rather than a biological axis worth interrogating in its own right.

To probe how the immune and metabolic systems relate to each other, the team conducted exploratory mediation analyses. In female patients, these analyses suggested bidirectional immune-metabolic associations, meaning the data were compatible with inflammatory markers relating to suicidal ideation through metabolic variables and vice versa. In male patients, the associations were primarily linked to metabolic pathways, reinforcing the impression that in boys the metabolic axis carries most of the signal. The authors are careful to frame these mediation results as exploratory, since mediation in cross-sectional data cannot confirm causal direction, but the pattern is consistent with the idea that the route from peripheral biology to suicidal thinking may be wired differently in each sex.

The study then moved from association to prediction. The researchers built machine learning classifiers, including logistic regression, support vector machines, random forests, gradient boosting machines, decision trees, naive Bayes, k-nearest neighbors and neural networks, to distinguish adolescents with suicidal ideation from those without. To handle the class imbalance between the 98 ideation cases and 221 controls, they used the synthetic minority over-sampling technique, and they performed feature selection with recursive feature elimination. Missing values were addressed with multivariate imputation by chained equations. The models demonstrated good discriminative ability as assessed by the area under the receiver operating characteristic curve, and decision curve analysis was used to evaluate their potential clinical usefulness.

A key strength of the machine learning analysis was its interpretability layer. Rather than treating the classifiers as black boxes, the team applied Shapley Additive Explanations, a technique borrowed from cooperative game theory that quantifies each feature’s contribution to individual predictions. The SHAP-based feature rankings supported the sex-specific patterns seen in the classical statistical analyses: inflammatory variables carried more weight in models for female patients, while metabolic variables dominated in models for male patients. This convergence between hypothesis-driven statistics and data-driven feature importance lends credibility to the central claim that suicidal ideation in depressed adolescents is accompanied by distinct, sex-divergent immune-metabolic profiles rather than a single uniform biomarker.

The implications are considerable. If replicated, routinely measured blood parameters, already collected in almost every clinical encounter, could eventually complement clinical assessment of suicide risk in adolescent depression, offering a cheap and scalable biological correlate to flag patients who warrant closer monitoring. The sex-specific findings also suggest that risk models and future intervention studies should be developed separately for boys and girls, or at minimum include sex as a moderating variable. The study was approved by the Ethics Committee of the First Affiliated Hospital of Xi’an Jiaotong University, conducted under the Declaration of Helsinki, and funded by the National Natural Science Foundation of China together with a clinical research award from the same hospital.

The authors themselves are measured about the limits of the work. The design is cross-sectional, so the biomarkers cannot yet be said to predict future suicidal thoughts or behavior, only to accompany them at a single point in time. The suicidal ideation classification rests on a single rating scale item, and the mediation analyses are exploratory. What the study does provide is a foundation: a large, well-characterized sample of adolescents with depression, a systematic comparison of immune and metabolic markers, clear evidence of sex-specific patterning, and interpretable machine learning models that performed well. The researchers call for longitudinal studies to test whether these immune-metabolic signatures can genuinely anticipate risk, and whether tracking them over time could sharpen suicide prevention in one of psychiatry’s most vulnerable populations.

Subject of Research: Immune and metabolic biomarkers of suicidal ideation in adolescents with major depressive disorder, analyzed with sex-stratified statistics and interpretable machine learning.

Article Title: Immune–metabolic signatures of suicidal ideation in adolescents with major depressive disorder: a sex-specific analysis

Article References: Zhang, X., Cao, Z., Jiang, J., Yang, D., Gao, F., Wang, C., Qi, Z., Zhang, Y., Wen, B., Ma, X., Gao, Y., & Jia, M. (2026). Immune–metabolic signatures of suicidal ideation in adolescents with major depressive disorder: a sex-specific analysis. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08725-3

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08725-3

Keywords: suicidal ideation, major depressive disorder, adolescents, biomarkers, inflammation, metabolism, sex differences, machine learning, SHAP, neutrophil-to-lymphocyte ratio, psychiatry, hypometabolism