TNF-α levels in first-episode, drug-naïve patients with major depressive disorder: a case-control study and meta-analysis

Abstract

Background

Major Depressive Disorder (MDD) is a common psychiatric condition with a multifactorial etiology that includes inflammatory mechanisms. Tumor Necrosis Factor-α (TNF-α), a pro-inflammatory cytokine, regulates several cellular functions in health and has been associated with reduced neuronal survival and increased inflammation in cortical areas related to reward processing, motivation, and decision-making in MDD. Although TNF-α has been studied in MDD, its precise role remains unclear, highlighting the need for further investigation. We therefore conducted a case-control study and a meta-analysis to better characterize peripheral blood TNF-α levels in first-episode, drug-naïve MDD patients compared with healthy controls.

Methods

In the case-control study, plasma TNF-α levels were measured in first-episode, drug-naïve patients with major depressive disorder (MDD) and healthy controls (HCs) matched on age, and sex, using high-sensitivity enzyme-linked immunosorbent assays (ELISAs). For the meta-analysis, we selected relevant case-control studies from systematic searches of PubMed, Embase, PsycINFO, and Web of Science. A random effects model was used to calculate combined standardized mean differences (SMD) and ratio of means (RoM) for result verification.

Results

A total of 63 MDD patients and 63 matched HCs were included in the case-control analysis. Plasma TNF-α levels were significantly higher in MDD patients than in HCs (z = -2.12, p < 0.05). For the meta-analysis, 16 studies comprising 866 patients and 759 controls were included. Pooled results demonstrated significantly higher TNF-α levels in MDD patients compared with HCs (SMD: Hedges’ g = 0.65, 95% CI: 0.28 to 1.01, p < 0.001). The RoM indicated a 30% higher mean concentration in patients (RoM = 1.30, 95% CI: 1.20 to 1.42, p < 0.001). However, substantial cross-study heterogeneity was observed (SMD: I² = 91.2%, p < 0.001; RoM: I² = 95.5%, p < 0.001).

Conclusions

The results of both the case-control and meta-analysis portions of our study suggest that TNF-α levels are higher in first-episode, drug-naïve MDD patients than in healthy individuals. These findings imply that TNF-α plays an important role in the pathophysiology of MDD. Future work should examine sources of heterogeneity across studies on inflammatory factors in depression, and assess potential therapeutic targets associated with TNF-α.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-025-07685-4.

Introduction

Major Depressive Disorder (MDD) is one of the most common psychiatric disorders, with a lifetime prevalence estimated at ~ 12–15% and a one-year prevalence of ~ 6% worldwide [1–3]. Understanding the etiology of MDD may help increase clinical cure rates and reduce disease burdens for patients, their families, and society. Accumulating evidence has implicated inflammatory dysregulation in the disease’s pathogenesis. Six large-scale meta-analyses of cross-sectional and longitudinal studies have demonstrated associations between depressive states and elevated levels of several inflammatory markers, including C-reactive protein (CRP), interleukin-1β (IL-1β), interleukin-6 (IL-6), and tumor necrosis factor-α (TNF-α) [4–9].

The inflammatory system is a critical mediator linking chronic stress, immune dysfunction, and depressive pathophysiology. Chronic psychological stress, which is a well-established MDD risk factor, activates the hypothalamic-pituitary-adrenal (HPA) axis and induces low-grade systemic inflammation. This inflammatory state is characterized by upregulations in pro-inflammatory cytokines including IL-1β, IL-6, and TNF-α. Tumor necrosis factor-α (TNF-α) plays a pivotal role in immune regulation and exerts broad effects on neuropsychiatric homeostasis. Beyond its established influence on apoptosis, synaptic plasticity, and neurotransmitter metabolism, recent evidence highlights additional pathways through which TNF-α may contribute to major depressive disorder (MDD). One important mechanism involves activation of the tryptophan–kynurenine pathway: by upregulating indoleamine 2,3-dioxygenase (IDO), TNF-α reduces serotonin availability while promoting the accumulation of neuroactive metabolites such as quinolinic acid, which are neurotoxic and pro-inflammatory [10]. Moreover, TNF-α has been shown to enhance oxidative and nitrosative stress by stimulating reactive oxygen and nitrogen species production, thereby amplifying cellular damage and neuroinflammation [11, 12]. Both kynurenine pathway dysregulation and increased oxidative/nitrosative stress have been directly associated with depressive symptomatology and clinical MDD [13]. These converging findings emphasize the multifaceted role of TNF-α in linking immune activation to the pathophysiology of depression. Preclinical studies suggest that elevated TNF-α levels may compromise the integrity of the blood-brain barrier, promote neuronal apoptosis, and disrupt serotonergic pathways, all of which can contribute to the emergence of depressive symptoms.

Increasing evidence suggests that elevated TNF-α levels are associated with MDD. However, because many existing studies have included patients with varying illness durations, ages of onset, treatment histories, and comorbidities, it remains challenging to fully disentangle the temporal relationship, even though substantial evidence supports a directional effect of inflammation on the onset and persistence of depression. For example, several antidepressants have been shown to modulate inflammatory markers. Additionally, inflammatory processes are involved in the onset and persistence of MDD [14], and inflammation can make patients more susceptible to treatment-resistant depression and worsen depressive symptoms [15–17], further complicating interpretations.

The present research landscape underscores the need for more focused studies examining the role of inflammation in MDD. First-episode drug-naïve patients are a critical population for investigating early pathophysiological changes in MDD, because they present the opportunity to study the “pure” biological basis of the condition without the confounding effects of prior treatments or prolonged disease courses. If pro-inflammatory cytokines such as TNF-α are indeed elevated in these patients, it would strengthen the hypothesis that inflammation plays a causal role in the onset of depression. Understanding the role of these cytokines in MDD could help drive the develop of novel therapeutic interventions targeting inflammation, including anti-TNF agents or other immunomodulatory therapies.

Previous case-control studies and meta-analyses on cytokine changes in MDD have been limited by inconsistent methodology, and have not focused on first-episode, drug-naïve patient populations. To address these research gaps, we conducted an integrated case-control study and meta-analysis focusing specifically on TNF-α levels in first-episode, drug-naïve MDD patients, The study aimed to clarify TNF-α’s role in MDD pathophysiology and evaluate its potential as a biomarker for early-stage disease.

Methods

Case-control study

Participants

In the case-control portion of the study, we recruited 63 first-episode, drug-naïve patients with MDD (29 males and 34 females) with a median age of 33 years (range: 26–41 years) from the First Affiliated Hospital of Zhengzhou University and Wuhu Fourth People’s Hospital in China. This was a cross-sectional case-control study, and TNF-α levels were measured once at baseline during the initial psychiatric assessment. Inclusion criteria for patients were: (1) age 18–65 years; (2) meeting DSM-5 diagnostic criteria for a first episode of MDD (confirmed by the Structured Clinical Interview for DSM-5 Disorders, SCID-5 [18, 19]; and (3) a Hamilton Depression Rating Scale (HAMD) 24-item [20] score ≥ 20 at baseline (indicating at least moderate depression). First episode was defined as the first lifetime Major Depressive Episode; this was verified using the SCID-5 lifetime module and review of medical records to ensure no prior depressive or manic/hypomanic episodes. We screened potential patients for eligibility through clinical interviews and chart reviews. All participants were drug-naïve, meaning they had never received antidepressants, antipsychotics, or mood stabilizers before enrollment.

For each patient in the MDD group, one healthy control (HC) was recruited, yielding 63 HCs matched 1:1 with patients on age and sex. Healthy controls were volunteers recruited from the community (through advertisements) and the hospital’s health examination center. Inclusion criteria for HCs were: (1) age 18–65 years; (2) no current or past psychiatric disorders as assessed by the Mini International Neuropsychiatric Interview 5.0 (MINI 5.0); and (3) no family history of mental disorders in first-degree relatives. All control candidates underwent a screening interview to rule out any psychiatric conditions and ensure they met all criteria. Matching on age and sex was performed to control for these potential confounding factors. This matching procedure resulted in identical sex distribution in both groups (29 males and 34 females in each), which was by design to eliminate sex-related biases in our analyses.

Both patients and controls were recruited between January, 2021 and December, 2024. They were identified and approached either during clinic visits (for patients) or via advertisement responses and routine health check-ups (for controls). All individuals who expressed interest underwent a screening process: for patients, two independent trained psychiatrists administered the SCID-5 to confirm the MDD diagnosis and assess for any exclusion criteria; for controls, a trained interviewer administered the MINI to ensure no psychiatric diagnoses. We initially assessed a larger pool of potential participants; those who did not meet inclusion criteria or who met any exclusion criteria (outlined below) were not enrolled. After screening, a total of 63 eligible MDD patients and 63 eligible matched HCs were included in the study. Healthy controls were matched 1:1 to MDD patients based on age (± 3 years) and sex. BMI was not a matching variable but was kept within a reasonable similarity range (± 2 kg/m²) during recruitment to reduce extreme imbalances.

The exclusion criteria for all the participants are as follows: 1) Serious medical illnesses: any severe chronic physical illness such as uncontrolled diabetes mellitus, uncontrolled hypertension, significant cardiac disease, or any active malignancy; (2) Neurological or immunological disorders: any severe central nervous system disorder (e.g., stroke, epilepsy, neurodegenerative disease, or traumatic brain injury) or serious immune-related disorder (e., acute and chronic systemic disorders associated with abnormal immune-inflammatory responses, including infections, HIV, allergies, pregnancy or the postpartum period, rheumatologic diseases, and immune-mediated diseases); (3) Recent use of immunomodulatory medications: use of any immunosuppressants or immunomodulating drugs in the recent period (for example, corticosteroids or other immune-modifying therapies within the last 30 days); (4) Psychotropic medication history: any prior use of antidepressants, antipsychotics, mood stabilizers, or other psychiatric medications. All MDD patients had never been treated with such medications (drug-naïve), and HCs similarly had no history of psychiatric medication use; (5) Substance abuse: diagnosis of drug or alcohol dependence according to DSM-5, or a positive history of substance abuse. Additionally, any use of illicit drugs (e.g., marijuana, amphetamines) within the past 30 days was an exclusion criterion; (6)Acute infection: although not explicitly listed in the original criteria, we also ensured that no participant had an acute infection at the time of evaluation or in the recent weeks prior to enrollment. Individuals with fever or signs of active infection were rescheduled or excluded, to avoid acute inflammatory responses (which could affect cytokine levels) from confounding the results.

All participants underwent standardized screening to exclude acute or recent infections, given the sensitivity of TNF-α to inflammatory states. Screening procedures included a brief physical examination (temperature measurement, evaluation for fever, chills, sore throat, cough, and other infectious symptoms) and review of recent medical history. Complete blood count (CBC) with differential and C-reactive protein (CRP) were obtained for all participants to identify acute inflammatory responses. Individuals with elevated CRP, leukocytosis, or abnormal neutrophil counts were excluded. In addition, participants were asked about antibiotic use or any physician-diagnosed infection within the preceding two weeks, and those reporting such history were excluded or rescheduled. No enrolled participant had evidence of active infection or recent antibiotic exposure.

Upon enrollment, demographic information (age, sex, education, etc.) and basic clinical data (such as BMI and routine laboratory tests) were collected for all participants by trained research assistants. Patients’ clinical psychiatric assessments included confirmation of the MDD diagnosis and evaluation of depression severity. Two board-certified psychiatrists independently confirmed each patient’s diagnosis using the SCID-5, ensuring reliability of the MDD diagnosis and ruling out any other psychiatric comorbidities (e.g., no patient met criteria for bipolar disorder, anxiety disorders, or other Axis I disorders). Depressive symptom severity was further evaluated with the Hamilton Depression Rating Scale-24 (HAMD-24). In cases of diagnostic disagreement, a consensus was reached through discussion.

This study was conducted at two clinical sites. The case-control component at the First Affiliated Hospital of Zhengzhou University was approved by its Institutional Ethics Committee (Approval No. 2020-KY-211, approved December 2020). The component conducted at the Fourth People’s Hospital of Wuhu was approved by its institutional review board (Approval No. 2024-KY-2072-002, approved January 2024). All data collection for both sites (January 2021–December 2024) occurred entirely within the period covered by valid ethical approval. Written informed consent was obtained from all participants prior to enrollment.

Laboratory measurements

All participants fasted for at least 10 h before having venous blood samples collected from their antecubital veins. The time from venipuncture to centrifugation was strictly controlled and did not exceed 30 min.Samples were immediately transferred into EDTA tubes, stored on ice, and centrifuged at 3,000 rpm for 10 min at 4 °C to obtain plasma. Sample handling and ELISA reproducibility: Serum samples were centrifuged within 30 min of venipuncture, aliquoted, and stored at − 80 °C. All samples were stored for no longer than 6 months before analysis. To minimize batch effects, all case and control samples were processed and analyzed in a single batch, and all ELISA plates were run during one analytical session using the same lot of reagents. Each plate contained manufacturer-provided standards and internal quality controls to ensure consistency of quantification. According to the manufacturer’s documentation for the specific lot used in this study, the kit is produced under a standardized process in which batch-to-batch variability is below the reportable threshold; therefore, lot-specific intra-assay and inter-assay coefficient of variation (CV%) values are not provided. No sample fell outside the assay’s reportable range. TNF-α levels were measured using a double antibody sandwich enzyme-linked immunosorbent assays (ELISA) kit (Epizyme Biomedical, HJ110, China). All assays were performed twice to ensure reliability. If the two replicate measurements differed by more than 10%, the assay was repeated, no samples required exclusion for this reason. Absorbance was measured at 450 nm using a microplate reader (Thermo, USA). To minimize batch effects, all plasma samples from MDD patients and healthy controls were assayed using the same lot of reagents, processed under identical laboratory conditions, and run on the same instrument in a single analytical session. The TNF-α detection range was 15.62–1000 pg/mL, with a lower detection limit of 6 pg/mL. All measured values fell within the detection range.

Statistical analyses

All statistical analyses were performed with IBM SPSS Statistics version 25.0 (IBM Corp., Armonk, NY, USA). Continuous variables were tested for normality with the Shapiro-Wilk test. Normally distributed data were expressed as means ± standard deviations (SD), and non-normally distributed data were presented as medians (interquartile ranges, IQR). Categorical variables were presented as counts (percentages, %).

For comparisons between MDD patients and healthy controls: Independent sample t-tests were used for normally distributed continuous variables (e.g., BMI, HAMD-24 scores). Mann-Whitney U tests were used for non-normally distributed continuous variables (e.g., age, TNF-α levels). Chi-square tests (χ² tests) were used for categorical variables (e.g., sex).

The TNF-α ELISA kit used in this study had a reportable detection range of 15.62–1000 pg/mL. All measured values fell within this range, and therefore no imputation or censoring procedures were required. Our pre-specified rule—had out-of-range values occurred was to assign the lower detection limit for values below the minimum and to re-assay samples exceeding the upper limit.ding variables.

The required sample size for the case–control component was determined a priori. Based on prior meta-analytic evidence suggesting a moderate effect size (d ≈ 0.5) for TNF-α differences between MDD and healthy controls, a power analysis using G*Power 3.1 indicated that 64 participants per group would be sufficient to achieve 80% power at α = 0.05. Our final sample exceeded this requirement (MDD: 73; HC: 71), ensuring adequate statistical power.

Meta-analysis

A literature search and pooled analysis were performed according to the latest version of the Preferred Reporting Item for Systematic Reviews and Meta-Analyses (PRISMA) statement [21]. The PubMed, Embase, PsycINFO and web of science databases were searched for studies that reported measurements of TNF-α in drug-naïve first-episode patients with MDD compared with healthy controls. The search window was between the establishment of the database and February 28, 2025. Two investigators (Sa Xiao and Mingli Lou) independently performed literature searches and title/abstract screening, made final decisions on eligibility after full-text review, and extracted data. The detailed search strings are listed in the supplementary materials which are part of the online version of this article.

Inclusion and exclusion criteria

The inclusion criteria for the meta-analysis were as follows: (1) studies including a case-control group; (2) studies with a group of patients with DSM [15] or International Statistical Classification of Diseases(ICD) [22] diagnoses of MDD; (3) studies with a healthy control group; (4) studies that assessed TNF-α levels.

The exclusion criteria were as follows: (1) studies including participants with medical and/or psychiatric comorbidities; (2) studies including pregnant women or women in the postpartum period; (3) Case reports or case series (N < 10); (4) studies looking at TNF-α levels in specimens or tissues other than peripheral blood; (5) studies including participants with serious medical comorbidities (e.g., diabetes, hypertension, cardiac disease, malignancy); (6) animal studies.

Data extraction and processing

Two investigators (Sa Xiao and Mingli Lou) imported the included studies into EndNote literature management software, used EndNote’s automatic check function to remove duplicated records, and manually screened titles and abstracts to exclude those with heterogenous participant groups or inconsistent study designs. The full texts of the remaining studies were then reviewed to determine final inclusion.

Disagreements that arose during document selection or data extraction were resolved through consensus discussion. Unresolved issues were referred to the corresponding author for a final decision.

The Newcastle-Ottawa Quality Assessment Scale (NOS) [23] was used to determine study quality. Studies with scores < 5 were classified as low-quality.

Statistical analysis

Data from the meta-analysis was assessed via two methods to understand continuous outcomes. The first method harnessed mean differences (MD) and standardized mean differences (SMD). the second method used ratio of means (RoM) to explore peripheral blood TNF-α concentration levels in MDD patients in comparison to healthy controls. Each effect size measure has its own strengths and weaknesses. MD are most useful when all studies measure the outcome in the same unit, as they provide a clear interpretation of absolute differences. SMD, expressed as Hedges’ g, are needed when studies use different measurement scales, as they provide unbiased effect sizes (ES) which are adjusted for small sample sizes. This allows for between-study comparability but also increases heterogeneity. RoM represent relative changes between groups and can sometimes reduce heterogeneity, particularly when the data are positive and normally distributed. Analyses that combine all of these approaches are more robust, because conclusions can be cross-validated with different ES metrics. Specifically, integrating MD, SMD, and RoM in a meta-analysis provides several advantages, including greater robustness due to result validation from different ES measures, broader applicability across data types and study designs, reduced heterogeneity, particularly when measurement units differ, and more comprehensive interpretation.

All analyses were completed using the Stata MP software (version 17.0; Stata Corp, College Station, TX, USA). Statistical significance was considered at α = 0.05. Study heterogeneity was assessed using Cochran’s Q test and I² statistics. When p >0.1 and I ² ≤ 50%, heterogeneity was considered small, and the fixed-effects model was used for data analysis. When p < 0.1 or I ² >50%, there was statistical heterogeneity, the random-effects model was used. To further explore potential sources of heterogeneity, we conducted meta-regression analyses using study-level covariates (e.g.,sample source, test type and age). Result stability was testing with sensitivity analyses via a piecewise exclusion method which considered differences in the combined effect sizes of the two models. Funnel plots and Egger′s tests were used to determine whether the included studies had publication bias (Egger′s test p >0.1 suggested no publication bias) [24]. The Duval and Tweedie nonparametric trim-and-fill methods [24] were performed to further assess the potential publication bias. All of these techniques were employed to ensure rigorous evaluation and interpretation of the synthesized data.

Results

Case-control study

A total of 91 patients with depressive symptoms were initially screened; 6 were excluded because they did not meet diagnostic criteria or declined to participate, and an additional 22 were excluded due to the absence of suitable matched controls (sex, age or outside the predefined tolerance range of ± 3 years for age, BMI or outside the predefined tolerance range of ± 2 kg/m² for BMI). Ultimately, 63 first-episode, drug-naïve MDD patients were included. For the control group, 82 healthy individuals were screened; 6 were excluded because of psychiatric history, family history, or refusal, and 13 were excluded due to unmatched age or sex. This process yielded 63 matched healthy controls. Their demographic and clinical characteristics are summarized Table 1. There were no statistically significant differences between the MDD and HC groups in terms of age (p = 0.849), sex (p = 1.00), or BMI (p = 0.639). TNF-α levels were also significantly elevated in the MDD group compared to the HC group (z = -2.12, P = 0.034) (Table 1).

Table 1.

Demographic characteristics of the MDD group and HC group

Meta-analysis

Search results

A total of 5,544 studies were retrieved from our literature search. After excluding 359 duplicates, 4,794 records were removed via title/abstract screening. Subsequently, 381 studies that did not meet the inclusion criteria were excluded. A total of 16 studies were included in the final analysis [25–40]. The specific inclusion and exclusion process is displayed in Fig. 1. Study characteristics (including research country, patient and control sample sizes) are provided in Table 2.

Fig. 1.

Flowchart of study selection for systematic review and meta-analysis

Table 2.

Studies investigating analytes involving TNF-α meeting inclusion criteria

Additional study characteristics, such as sex, age, BMI, and control-matching methodology, catchment area and sample source (serum or plasma) are provided in Supplementary Table S1.

Study characteristics

A total of 16 studies (with 866 patients and 759 HCs) were included after systemic literature review as compare to prior meta-analysis [41] (with 5 studies including 287 patients and 281 HCs). The larger sample size can provide more reliable results. All studies reported peripheral blood TNF-α concentrations (pg /mL) in first-episode, drug-naïve MDD patients and HCs. The detection methods included ELISA multiple immunoassays and biochip array technologies. The studies were conducted in Asia, Europe, and the Americas.

Meta-analysis results

SMD results

Figure 2 displays a meta-analysis of peripheral blood TNF-α levels in first-episode, drug-naïve MDD patients compared to HCs SMD. The pooled analysis demonstrated significantly elevated TNF-α levels in MDD patients (random-effects model: Hedges’g = 0.65, 95% CI: 0.28–1.01; z = 3.498, p < 0.001). Substantial heterogeneity was observed across studies (Q = 170.11, df = 15, p < 0.001; I² = 91.20%). The robustness of our findings was confirmed through sensitivity analyses (Supplementary Figure S1), demonstrating that the removal of individual studies sequentially had no substantial impact on either the magnitude or statistical significance of the effect size estimates. Egger’s regression test did not show any evidence of publication bias (p = 0.251) (Supplementary Figure S2). A sensitivity analysis using the trim-and-fill method was performed with 0 imputed studies, which produced a symmetrical funnel plot (Supplementary Figure S3). Using the trim-and-fill method, the Hedges’g was 0.65 (95% CI, 0.29–1.01; p < 0.001). Correction for potential publication bias thus did not alter the significant association.

Fig. 2.

Forest plots of studies which measured TNF-α levels in patients with drug-naïve first-episode MDD compared to HCs

RoM results

Figure 3 displays RoM analysis which compared peripheral blood TNF-α levels between first-episode, drug-naïve MDD patients and HCs. The pooled RoM indicated that MDD patients had significantly increased TNF-α levels (random-effects model: RoM = 1.30, 95% CI: 1.20–1.42; z = 6.027, p < 0.001), Corresponding to a 30% higher mean concentration. There was also substantial heterogeneity which persisted across studies (Q = 334.95, df = 15, p < 0.001; I² = 95.5%). The RoM results also corresponded with SMD findings (Hedges’g = 0.65, p < 0.001), confirming robust evidence of TNF-α elevation in MDD patients. Sensitivity analysis (Supplementary Figure S4) also confirmed the robustness of our findings. Egger’s test suggested potential publication bias (p = 0.084 < 0.1) (Supplementary Figure S5). However, but after correction by the trim-and-fill method (the funnel plot was symmetrical after filling one study) (Supplementary Figure S6). the corrected mean ratio (RoM = 1.29, 95%CI 1.19–1.41, p < 0.001) still showed a significant core effect. It indicates that publication bias does not affect the reliability of the conclusion that “the level of TNF-α in patients with MDD is significantly higher than that in healthy people.

Fig. 3.

Forest plots of studies which measured TNF-α levels in drug-naïve first-episode MDD Patients compared to HCs

Meta regression showed no significant correlation between effect sizes in terms of sample source (serum/plasma/other, p > 0.70), test type (ELISA/other, p > 0.06), region (continent) (Europe/America/Asia, p > 0.09), and age (adult/underage, p > 0.74).

Although both methods suggested there was substantial between-study heterogeneity, the convergent results significantly strengthened the reliability of our findings.

The meta-analysis revealed that there were significantly elevated TNF-α levels in in first-episode, drug-naïve MDD patients compared to HCs (Hedges’g = 0.65, 95% CI: 0.28 to 1.01)—representing a moderate effect size that remained consistent across different statistical methods. The RoM (1.30, 95% CI: 1.20 to 1.42) also indicated a 30% higher mean TNF-α concentration in MDD patients, potentially reflecting neuroinflammatory pathophysiology. Convergent evidence from both the SMD and ratio-based metrics underscores TNF-α’s potential as an early biomarker of MDD. RoM-derived quantitative estimates (1.3 fold elevation) may also inform clinical decision-making and therapeutic development.

Discussion

This investigation bridges a critical knowledge gap in cytokine profiling in MDD patients through a focused examination of first-episode, drug-naïve MDD patients. In the case-control portion of our study, our use of strict exclusion criteria and diagnostic validation via semi-structured interviews enhanced the methodological rigor. By systematically excluding pharmacologically-treated individuals, we effectively controlled medication-related impacts on peripheral inflammatory markers, leading to clearer insights into the pathophysiology of MDD. Our case-control analyses revealed increased plasma TNF-α concentrations in first-episode, drug-naïve MDD patients compared to HCs (p = 0.034). Such an increase, observed at the initial stage of illness, suggests that elevated TNF-α is a core pathophysiological feature of early-stage MDD rather than a downstream effect of prolonged disease or pharmacotherapy. These case–control findings, coupled with our controlled study design, underscore TNF-α’s potential role as a mediator of depression pathogenesis and as an early-stage biomarker of MDD.

Our meta-analysis of 16 studies further reinforces the case–control findings, demonstrating a moderate TNF-α elevation in first-episode, drug-naïve MDD (Hedges’ g = 0.65, 95% CI: 0.28–1.01), corresponding to ~ 30% higher mean levels in patients than in controls (RoM = 1.30, 95% CI: 1.20–1.42). Compared with prior meta-analyses in mixed MDD samples, our results indicate a similar or slightly stronger effect size, but with greater specificity. For example, Dowlati et al. [5] reported a weighted mean difference of 3.97 pg/mL across 13 studies, Liu et al. [42] an SMD of 0.57, and Haapakoski et al. [43] a smaller effect (d ≈ 0.40) limited by extensive heterogeneity. As noted by Köhler et al. [8], cytokine syntheses often show large variability (I² up to 98%), and our heterogeneity levels are comparable.The novelty of our study lies not only in the larger sample size but also in its focus on first-episode, drug-naïve patients, which minimizes confounding by medication or chronicity and strengthens causal inference about early disease biology. Furthermore, by reporting both SMD and RoM, we provide complementary and clinically interpretable effect estimates, while meta-regression and sensitivity analyses address methodological sources of heterogeneity.

For the SMD method, no missing studies were imputed (0 studies imputed). After adjustment, the funnel plot was symmetrical, and the pooled effect size (Hedges’ g) was 0.65 (95% CI: 0.29–1.01, P < 0.001), which was consistent with the original results. For the RoM method, the funnel plot regained symmetry after imputing 1 missing study, and the adjusted pooled effect size (RoM) remained significant at 1.29 (95% CI: 1.19–1.41, P < 0.001). By simulating a scenario without publication bias, this process verified the robustness of the pooled effect size. It not only eliminated the interference of publication bias on the conclusion that “circulating TNF-α levels are significantly higher in first-episode, drug-naive MDD patients than in healthy individuals” but also demonstrated the rigor of the study methodology, thus providing an objective and balanced interpretation of the study evidence. Collectively, these refinements enhance the evidence base and underscore TNF-α’s potential as an early biomarker of MDD.

In our study specifically, TNF-α was measured in EDTA-treated plasma using an ELISA, whereas some other studies have quantified TNF-α in serum. Plasma-based assays are known to produce higher absolute cytokine concentrations than serum-based assays because the anticoagulant (EDTA) prevents clotting-related cytokine depletion and helps stabilize cytokines post-collection. Prior methodological work demonstrates that TNF-α in EDTA plasma remains stable for several hours, whereas serum values may be lower or more variable due to clotting processes [44]. This likely contributed to the relatively high absolute TNF-α values 、observed in both patients and controls in our cohort. Accordingly, while direct comparison of absolute concentrations with serum-based studies should be made cautiously, the within-cohort case–control differences remain robust and clinically meaningful.

Another important consideration is the equal sex distribution in our sample, which was an intentional result of matching each MDD patient to a control by sex (as well as by age and BMI) to eliminate sex as a confounding variable. This balanced design enhances internal validity and comparability between groups, but it does not reflect the known epidemiological prevalence pattern of MDD, wherein females are affected at higher rates than males [45, 46]. As a result, our findings cannot directly inform sex-specific differences in depression-related inflammation. Given that certain immune-inflammatory responses, including TNF-α levels, may differ by sex [47], future studies with larger and more sex-representative samples are warranted to determine whether the observed TNF-α elevation in MDD manifests differently in men versus women.

TNF-α is involved in several components of MDD’s pathophysiology. The C-Jun N-terminal kinase (JNK) signaling pathway can be activated by J TNF-α, and may be related to both neuronal apoptosis and MDD [48–50] Second, TNF-α also activates the nuclear factor kappa B(NF-κB) signalling pathway, which drives inflammatory responses. Dysregulation of this pathway may lead to inflammatory cytokine-associated depression [51, 52]. TNF-α additionally mediates TNF receptor 1(TNFR1) and TNF receptor 2(TNFR2) receptor-driven emotional regulation in the hypothalamus, hippocampus, amygdala, and prefrontal cortex, which may influence MDD symptoms [53]. TNFR1 and TNFR2 are also expressed in astrocytes, and astrocyte-specific gene expression is altered in post-mortem brain tissue from MDD patients [54, 55], implying that changes in astrocytic gene expression may partially underlie MDD onset. TNF-α also hand impairs synaptic plasticity, leading to cognitive deficits and memory impairments in cytokine-associated depression [56, 57], and increases the permeability of the blood–brain barrier [58] which may increase the presence of inflammatory mediators and peripheral immune cells in the central nervous system and drive MDD symptoms [59–64].

Pro-inflammatory factors can also activate the HPA axis, leading to an increase in glucocorticoid levels and the onset of depression [65]. TNF-α activates corticotropin-releasing hormone (CRH), which upregulates adrenocorticotropic hormone (ACTH) and increases glucocorticosteroid (GCS) levels. This process overstimulates the HPA axis, disrupts negative feedback, and triggers depression [66]. Additionally, under stress, cortical microglia secrete TNF-α and other substances, leading to neuroinflammation, stimulation of the HPA axis, and disruption of feedback inhibition, which overall decreases 5-HT levels and induces depression [67–69]. In addition, emerging evidence suggests that heightened inflammatory activity may contribute to suicidal behavior. Pro-inflammatory cytokines have been linked to alterations in monoaminergic neurotransmission, reduced neuroplasticity, and dysregulation of stress-response systems, all of which may increase vulnerability to suicidal thoughts and actions. Although clinical trials remain limited, recent findings indicate that anti-inflammatory strategies may hold potential for reducing suicide risk. This aligns with growing interest in inflammation-targeted interventions as part of comprehensive suicide-prevention approaches [70].

Taken together, the present and previous research findings all support the “cytokine hypothesis” of depression, which proposes that systemic, cytokine-mediated inflammation influences depressive symptoms both directly and indirectly. For example, TNF-α may activate indoleamine 2,3-dioxygenase (IDO), an enzyme involved in the kynurenine pathway, leading to reduced serotonin availability and increased production of neurotoxic metabolites. This mechanism may partially explain the mood disturbances and cognitive dysfunction seen in MDD patients.

Our findings have significant clinical implications. The fact that TNF-α levels are elevated in first-episode, drug-naïve MDD patients suggests that TNF-α may be able to serve as a biomarker for early-stage depression. Identifying biomarkers like this may facilitate the development of personalized treatment approaches and improve diagnostic accuracy. Our results also highlight the potential of targeting inflammation as a novel therapeutic strategy for MDD. Anti-inflammatory medications, such as TNF-α inhibitors, have shown promising results in preclinical studies and small clinical trials. For example, infliximab, a monoclonal antibody targeting TNF-α, showed antidepressant effects in patients with high baseline inflammatory levels. Previous studies have systematically reviewed and meta-analyzed the efficacy and safety of anti-inflammatory drugs in treating MDD. The results indicate that these drugs exert antidepressant effects in MDD patients and are generally well tolerated [71]. Based on the above reviews and safety evaluations, future studies should implement randomized, biomarker-enriched trials targeting patients with high inflammation or elevated TNF-α (or other inflammatory markers) to optimize treatment efficacy and safety.

Our meta-analysis has several limitations. First, both our hospital-based sample and the studies included in the meta-analysis were cross-sectional. TNF-α levels were measured only once at baseline, without longitudinal tracking. Consequently, the temporal direction of the association between TNF-α and depressive status cannot be determined. The findings do not imply causality, and reverse causation or unmeasured confounding cannot be ruled out. Second, the included studies were highly heterogeneous, which may limit the generalizability of the findings. Although subgroup and sensitivity analyses were conducted, residual heterogeneity cannot be excluded. Third, the small sample sizes of some of the included studies may have limited statistical power and biased the results. Fourth, differences in demographic factors, such as age, sex, and ethnicity, were not consistently controlled for across studies, which may have influenced the findings. Fifth, information on the duration of untreated depression was not available. As this variable was inconsistently reported across original studies, its potential impact on inflammatory markers could not be evaluated. Sixth, only TNF-α was measured without a broader immune panel or detailed metabolic/lifestyle correlates; this constrains interpretation of immune system dysregulation and should be addressed in future studies using multi-marker profiling with standardized pre-analytics. Seventh, the meta-analysis was not registered (e.g., in PROSPERO), which imposes certain limitations regarding transparency and reproducibility [71].Finally, the lack of standardized protocols for cytokine measurement poses a challenge to data synthesis and interpretation.

Conclusions

This meta-analysis provides robust evidence that first-episode, drug-naïve MDD patients have significantly elevated TNF-α levels compared to HCs. These findings underscore the role of inflammation in early MDD pathophysiology, and highlight TNF-α as a potential biomarker and therapeutic target for the disease. Addressing the inflammatory component of depression may lead to novel, effective interventions for this debilitating conditions.

Future research should address the limitations identified in this meta-analysis. Larger, multicenter studies with standardized patient enrollment and serum measurement methodologies are needed to validate our findings and improve reproducibility. Longitudinal studies should be conducted to clarify the temporal relationship between TNF-α levels and the onset of depressive symptoms.

Further investigations into the genetic and epigenetic factors regulating TNF-α expression may also provide insights into individual differences in inflammatory responses to stress. Research such as this could inform the precision medicine treatments for MDD. Finally, clinical trials evaluating the efficacy of anti-inflammatory therapies in first-episode, drug-naïve MDD patients are also needed.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

MDD

Major Depressive Disorder

TNF-α

Tumor Necrosis Factor- alpha

HCs

Healthy controls

ELISAs

Enzyme-linked immunosorbent assays

SMD

Standardized mean differences

RoM

Ratio of means

CRP

C-reactive protein

IL-1β

Interleukin-1β

IL-6

Interleukin-6

HPA

Hypothalamic-pituitary-adrenal

DSM-5

Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition

HAMD-24

The Hamilton Depression Rating Scale-24

ELISA

Enzyme-linked immunosorbent assays

SD

Standard Deviation

IQR

Interquartile ranges

PRISMA

The Preferred Reporting Item for Systematic Reviews and Meta-Analyses

NOS

Newcastle-Ottawa Quality Assessment Scale

MD

Mean differences

ES

Effect sizes

JNK

The C-Jun N-terminal kinase

NF-κB

Nuclear factor kappa B

TNFR1

TNF receptor 1

TNFR2

TNF receptor 2

CRH

Corticotropin-releasing hormone

ACTH

Adrenocorticotropic hormone

GCS

Increases glucocorticosteroid

IDO

Indoleamine 2,3-dioxygenase

Author contributions

CJL reviewed literature, collected clinical data, evaluated scale, and drafted the original manuscript. YW, SX and MLL conducted literature screening and quality assessment, performed data extraction and analysis, collected clinical data and evaluated scale. CJL and JRC recruited subjects, collected clinical data and evaluated scale. YRX provided financial support. MBD was responsible for collected clinical data, evaluated scale and the supervision and coordination of research as a whole. YRX provided financial support, reviewed literature, designed the study, and was responsible for the supervision and coordination of research as a whole. All the authors reviewed the various drafts of the manuscript and have approved the final version of the manuscript.

Funding

YRX’s involvement in this research was funded by Key projects jointly built by provinces and ministries (SBGJ202302071), Research Projects of Medical Education in Henan Province (WJLX2023049), Special Project for Nursing Scientific Research (Nursing Team) of the First Affiliated Hospital of Zhengzhou University (HLKY2023006) and Key Scientific Research Projects of Higher Education Institutions in Henan Province (25A320047).

Data availability

Data is provided within the manuscript or supplementary information files. Raw data supporting the obtained results are available at the corresponding author.

Declarations

Ethics approval and consent to participate

Informed consent was obtained from all the participants and their legal guardians. This study was conducted at two clinical sites. The case-control component at the First Affiliated Hospital of Zhengzhou University was approved by its Institutional Ethics Committee (Approval No. 2020-KY-211, approved December 2020). The component conducted at the Fourth People’s Hospital of Wuhu was approved by its institutional review board (Approval No. 2024-KY-2072-002, approved January 2024). All data collection for both sites (January 2021–December 2024) occurred entirely within the period covered by valid ethical approval. Written informed consent was obtained from all participants prior to enrollment. The study was performed in accordance with the Declaration of Helsinki.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.