Putative Neuroimmune Mechanisms and Candidate Biomarkers of rTMS Efficacy in Improving Negative Symptoms of Schizophrenia: Evidence Boundaries and Translational Pathways

Abstract

Negative symptoms in schizophrenia often persist for prolonged periods, severely impairing patients’ social functioning, while the therapeutic benefits of existing pharmacological treatments remain limited. Recent meta-analyses suggest that high-frequency stimulation of the left dorsolateral prefrontal cortex (L-DLPFC) provides short-term adjunctive benefits for negative symptoms, with effects ranging from small to moderate. However, treatment response is influenced by the stimulation parameters, target localization, and patient characteristics; thus, overall therapeutic efficacy remains heterogeneous. In patients with schizophrenia, peripheral immune–inflammatory abnormalities may occur, and some inflammatory markers are associated with disease stage, cognitive function, and negative symptom severity. Repetitive transcranial magnetic stimulation (rTMS) can modulate the prefrontal cortex and related brain networks, but its effects on immune-inflammatory processes, neuroendocrine–autonomic function, and neuroplasticity remain to be further elucidated. This review summarizes the immune–inflammatory mechanisms underlying negative symptoms, the clinical efficacy of rTMS, and the biological link between them, and discusses the potential of inflammatory markers for patient stratification and treatment-response prediction. In the future, standardized symptom assessment, stimulation parameters, and target localization are warranted, alongside repeated biological sample collection, prespecified stratified analyses, and external validation. Until inflammatory markers have been sufficiently validated with respect to predictive performance, reproducibility, and clinical net benefit, they should not be used alone to determine rTMS treatment eligibility, select stimulation parameters, or guide stratification decisions.

Introduction

Schizophrenia is a major psychiatric disorder that leads to long-term disability. Its core clinical manifestations include positive symptoms, negative symptoms, and cognitive impairment.¹ Compared with positive symptoms, negative symptoms more frequently exhibit a chronic, persistent course and are closely associated with recovery of social functioning, occupational ability, quality of life, and long-term prognosis.² Currently, dopamine D2 receptor antagonists remain the cornerstone of pharmacological treatment for schizophrenia; however, their improvement of negative symptoms is limited.^1,3 Therefore, exploring novel adjunctive therapeutic approaches and their underlying biological mechanisms holds substantial clinical significance. rTMS can modulate cortical excitability and distributed brain network activity, thereby influencing psychiatric symptoms.⁴ Existing randomized controlled trials and meta-analyses suggest that high-frequency repetitive transcranial magnetic stimulation (HF rTMS) over the L-DLPFC is associated with a modest improvement in overall negative symptoms in patients with schizophrenia. However, the overall therapeutic effect remains limited, and it appears to be influenced by factors such as stimulation frequency, treatment course, and target selection.^5–7 Therefore, current research has shifted from simply determining whether rTMS is effective to identifying patients who may benefit and the optimal stimulation parameters. In parallel, neuroimmunology provides a new perspective for explaining treatment heterogeneity. Patients with schizophrenia may exhibit aberrant peripheral inflammatory markers, which are associated with cognitive impairment, functional deficits, and certain negative symptoms phenotypes.^8–10 Inflammatory processes may contribute to disease pathophysiology by modulating neurotransmitter metabolism, microglial activity, complement-mediated synaptic pruning, blood–brain barrier integrity, and neuroplasticity.^11,12 Based on this, the present article addresses three related issues. First, it asks which inflammatory findings show relatively stable links with negative symptoms phenotypes. Second, it evaluates how strongly the clinical efficacy of rTMS, as well as its immune-related mechanisms, are supported by the current evidence base. Third, it considers what methodological standards candidate inflammatory biomarkers must satisfy if they are to serve as stratification tools for rTMS. This study does not set any unverified clinical thresholds. Instead, it aims to define the limits of the existing evidence and to outline a translational pathway that can be tested in prospective studies.

Scope of the Review and Evidence Appraisal

This article adopts a critical narrative review framework to synthesize research evidence across domains including negative symptoms in schizophrenia, rTMS, immune–inflammatory mechanisms, and biomarker stratification. The interpretation of evidence is primarily based on study design, consistency of findings, clinical relevance, and the extent to which potential confounding factors were controlled. Evidence from other psychiatric disorders and animal-model studies is used only to support the biological plausibility of the proposed mechanisms, and is not treated as direct evidence for negative symptoms in schizophrenia or for the clinical efficacy of rTMS. To distinguish the degree of certainty across different findings, we categorize the evidence into three tiers: (i) relatively established evidence, supported by meta-analyses, randomized controlled trials, and independent replications; (ii) emerging evidence, characterized by observed clinical or biological signals but constrained by limited sample sizes, heterogeneity across studies, and insufficient external validation; and (iii) hypothetical evidence, derived primarily from indirect clinical observations, animal studies, or mechanistic inferences, and therefore requiring validation through targeted direct investigations. This framework is intended to standardize evidence interpretation and to provide a reference point for subsequent prospective research.

Immune-Inflammatory Mechanisms Underlying Schizophrenia Negative Symptoms

Aberrant immune–inflammatory activity constitutes one of the key findings in schizophrenia biological research. Accumulating evidence from meta-analyses indicates that patients show altered peripheral inflammatory protein profiles. However, the specific patterns vary according to illness stage and clinical status. Early meta-analytic evidence indicates that dysregulation of interleukin-1β (IL-1β) and interleukin-6 (IL-6) is more frequently observed in patients experiencing acute exacerbation, those with first-episode psychosis, or those who have not received treatment. In contrast, alterations in tumor necrosis factor-α (TNF-α) and interferon-γ (IFN-γ) may be detected across multiple phases of the illness course.⁸ Recent network meta-analyses have further demonstrated that IL-1β, IL-6, TNF-α, and C-reactive protein (CRP) are elevated in both acute-phase and chronic-phase patients compared with healthy controls. By contrast, IFN-γ alterations show a more pronounced stage-dependent pattern.¹³ These findings do not necessarily indicate a continuously activated single immune pathway; instead, the disease process may involve, in combination, infectious triggers, autoimmune mechanisms, elevated inflammatory mediators, and genetic susceptibility to immune dysregulation.¹⁴ Inflammation-related alterations are not specific to schizophrenia. Aberrant cytokine networks with partial overlap have been reported across schizophrenia, bipolar disorder, and major depressive disorder, suggesting that immune–inflammatory dysregulation may represent a transdiagnostic pathological dimension rather than a disease-specific hallmark.⁹ Although peripheral inflammatory markers cannot directly reflect central nervous system inflammation, they can provide indirect information about the individual’s systemic immune state.¹⁵ Therefore, the clinical interpretability of any single cytokine is likely limited. Integrating peripheral measures with the negative symptoms dimension, cognitive performance and social functioning, as well as neuroimaging and brain-network characteristics, may be more informative for identifying patient subgroups with distinct biological profiles.

Inflammation-Related Pathways as Candidate Mechanisms Underlying Schizophrenia Negative Symptoms

Aberrant inflammation may further disrupt brain-network function relevant to schizophrenia negative symptoms by impairing immunometabolic balance and synaptic regulation. Based on this premise, we integrate a candidate mechanistic framework comprising genetic susceptibility, peripheral immune dysregulation, dysfunction of the tryptophan-kynurenine pathway, complement-microglia-mediated synaptic pruning, dysfunction of the prefrontal-striatal-limbic circuitry, and alterations in the gut–brain axis. We use this framework as a basis for understanding the heterogeneity of negative symptoms and for identifying potential rTMS targets (Figure 1).

Figure 1.

Candidate immune–inflammatory mechanisms underlying negative symptoms in schizophrenia and potential intervention points for rTMS. This figure was created by Figdraw2.0, https://www.figdraw.com. This schematic summarizes the candidate links between genetic factors and peripheral inflammation, gut–brain axis dysregulation, inflammation-driven kynurenine pathway activation, neuroimmune and synaptic mechanisms, and abnormalities in the prefrontal-striatal-limbic network with the negative symptom dimension. Potential intervention sites of rTMS and priority directions for future research are also indicated. Solid lines represent associations supported by existing evidence, whereas dashed lines denote mechanistic pathways that remain to be validated.

Abbreviations: CRP, C-reactive protein; IL, interleukin; TNF-α, tumor necrosis factor-alpha; IDO, indoleamine 2,3-dioxygenase; KYNA, kynurenic acid; QUIN, quinolinic acid; NMDA, N-methyl-D-aspartate; nAChR, nicotinic acetylcholine receptor; C3, complement component 3; C4, complement component 4; rTMS, repetitive transcranial magnetic stimulation; TSPO-PET, translocator protein positron emission tomography.

The tryptophan-kynurenine (TRP-KYN) pathway is a key link between inflammation and neurotransmitter metabolism. Its metabolic products participate in the regulation of glutamatergic neurotransmission, redox homeostasis, and synaptic plasticity. In antipsychotic-naïve, first-episode psychosis patients, higher serum kynurenic acid (KYNA) levels were associated with more severe negative symptoms, higher IL-1β levels, and poorer short-term clinical improvement.¹⁶ In addition, another study reported that an increased QUIN/KYNA ratio was primarily related to higher Positive and Negative Syndrome Scale (PANSS) total scores and general psychopathology, without showing clear specificity for negative symptoms.¹⁷ Treatment status, clinical subtype, and clozapine exposure may also influence the associations between relevant metabolic markers and symptom profiles.¹⁸ The gut–brain axis may also influence central functions via microbial metabolites, intestinal barrier integrity, and immune signaling. Evidence from bidirectional Mendelian randomization suggests that certain gut microbiome characteristics have a bidirectional causal relationship with schizophrenia.¹⁹ A small-sample study reported microbiome alterations only after low-frequency rTMS combined with risperidone; however, in the absence of an rTMS-only control arm, these changes cannot be attributed to rTMS itself.²⁰ In addition, the microglia–complement–synaptic pruning pathway provides a further biological explanation for these relevant central immune mechanisms. Microglia participate in immune surveillance, synaptic remodeling, and brain network maturation; nevertheless, Translocator protein positron emission tomography (TSPO-PET) findings vary with the tracer, quantification method, and disease stage, so they cannot be interpreted as persistent microglial overactivation.²¹ Genetic studies show Complement Component 4 (C4) structural variants and higher Complement Component 4A (C4A) expression increase schizophrenia risk.²² Patient-derived cellular models have also shown increased complement deposition and enhanced microglia-mediated synaptic engulfment.^11,23 Aberrant synaptic pruning may affect the prefrontal-striatal-limbic circuitry and establish candidate links to deficits in motivation, reward processing, and social behavioral abnormalities; however, the association with specific dimensions of negative symptoms has not yet been robustly and consistently replicated.

Overall, the aforementioned pathways are currently more suitable as a candidate explanatory framework linking inflammation, abnormal brain networks, and negative symptoms. Whether they mediate the therapeutic response to rTMS remains to be determined and should be tested in clinical studies with a clearly defined temporal sequence.

Clinical Associations Between Schizophrenia Negative Symptom Phenotypes and Aberrant Inflammatory Markers

Negative symptoms are not a homogeneous clinical entity. From a clinical typology perspective, primary and secondary negative symptoms have different etiologies: the former more likely reflect the core pathology of the disorder, whereas the latter may be attributable to factors such as depression, extrapyramidal reactions, sedation, avoidance of positive symptoms, or environmental limitations. Persistent negative symptoms emphasize the stable presence of symptoms across the course of illness, whereas the deficit syndrome refers to a specific clinical subtype characterized by primary, enduring negative symptoms. From a dimensional standpoint, the Brief Negative Symptom Scale (BNSS) is typically summarized into two higher-order domains: reduced emotional expressivity (EXP) and motivation and pleasure (MAP), which–compared with the PANSS negative symptom subscale–may be better suited for identifying mechanistic differences across negative symptom dimensions.^24–27 Therefore, when examining the relationship between inflammation and negative symptoms, it is essential to first specify the clinical phenotype and symptom dimensions under investigation, rather than making broad inferences based solely on total scores.

On the basis of existing evidence, it appears likely that patients with the deficit syndrome may exhibit a more pronounced pro-inflammatory phenotype; however, the relevant findings have largely been derived from cross-sectional studies and have not yet been adequately replicated using consistent phenotypic definitions and longitudinal designs. Compared with non-deficit patients, some studies have reported higher levels of pro-inflammatory markers in deficit patients. TNF-α concentration has been shown to be positively correlated with emotional blunting, alogia, and the total score of negative symptoms.²⁸ CRP has also been reported to be independently associated with the deficit syndrome.²⁹ Similarly, a first-episode schizophrenia study conducted in China found that patients with negative symptom subtypes had elevated TNF-α levels, which were positively correlated with negative symptom severity.³⁰ However, the deficit syndrome is a clinical category, whereas the BNSS captures separable symptom dimensions; therefore, the findings above are not sufficient to draw direct conclusions regarding dimension-specific inflammatory mechanisms. At the level of specific symptom dimensions, the current evidence remains limited. The association between TNF-α and emotional blunting and alogia is directionally consistent with the reduced emotional expressivity (expressive deficits) domain; however, it has not yet been sufficiently replicated in independent samples. In contrast, the relationship between IL-6 and motivation-related symptoms appears less consistent. In clinically stable patients with early-onset schizophrenia, IL-6 levels were negatively correlated with the total PANSS negative symptom score, as well as with avolition and social withdrawal.³¹ Another study reported that IL-6 was more strongly associated with depressive symptoms in the early stage of illness, whereas, as the duration of illness increased, its association with the severity of negative symptoms became stronger.^32,33 Therefore, the existing evidence does not support a simple mapping of any single inflammatory cytokine to a fixed negative symptom dimension. A more plausible interpretation is that the inflammation–symptom relationship may be jointly influenced by factors such as the stage of illness, treatment status, and the composition of symptom profiles.

The clinical interpretation of CRP should remain cautious. Meta-analyses have shown that CRP levels are generally higher in patients with schizophrenia than in healthy controls, although antipsychotic treatment is not necessarily associated with a significant change.¹⁰ Although some studies have suggested that CRP may have some clinical utility, its associations with negative symptoms, positive symptoms, and general psychopathology have been inconsistent.^34,35 Associations between CRP and IL-6 levels and default mode network dysfunction further suggest that inflammatory signaling may be linked to brain-network functional states.³⁶ In addition, inflammatory markers have been associated with cognitive and social-cognitive impairments, although it remains unclear whether these findings correspond directly to dysfunction related to negative symptoms. In early psychosis, higher IL-6 levels have been associated with poorer mentalizing ability, and TNF-α was reported to predict worse facial emotion recognition performance.³⁷ Moreover, IL-6 and interleukin-8 (IL-8) have been linked to innate immune activation and cognitive performance.³⁸ Impaired working memory has also been reported in association with multiple inflammatory markers.³⁹ Collectively, these findings suggest that inflammation-related alterations may involve cognitive processing, social cognition, and brain-network efficiency, and may co-occur with functional impairment. However, their specific role in the development of negative symptoms remains to be established.

Overall, existing evidence suggests that CRP, IL-6, TNF-α, and other inflammatory mediators are associated with schizophrenia negative symptoms and cognitive impairment; however, the direction and magnitude of these associations may be influenced by factors such as disease stage, medication status, and clinical phenotypes (Table 1). Given the limited efficacy of current pharmacotherapies for ameliorating primary negative symptoms, adjunctive neuromodulatory strategies such as rTMS warrant further investigation. Nevertheless, whether rTMS can improve negative symptoms via immune–inflammatory pathways remains to be established by direct evidence. The following section reviews the clinical evidence for rTMS and potential sources of heterogeneity in treatment response.

Table 1.

Representative Studies on the Clinical Associations of Inflammatory Factors with Negative Symptoms and Cognitive Impairment in Schizophrenia

Author (Year)	Study Design	Sample (n)	Illness Stage/Med Status	Inflammatory Markers	Negative Symptom/Cognitive Assessment	Main Correlations or Findings	Ref.
Fernandes et al, 2016	Cross-sectional and longitudinal studies, meta-analysis: including patients with schizophrenia and healthy controls, n=85,000	Mean age 16–57 years	Included both antipsychotic-naïve and medicated patients	CRP	PANSS; cognitive measures in a subset of included studies	Higher CRP levels in patients with schizophrenia than in healthy controls; evidence for an association between CRP and negative symptoms was inconsistent	[10]
Dickerson et al, 2007	Cross-sectional study; patients with schizophrenia (n=413)	CRP <5.0 mg/L: 39.5±11.3 years, male/female 201/85; CRP ≥5.0 mg/L: 42.7±10.6 years, male/female 55/72	406 participants were receiving antipsychotic medication	CRP	PANSS; RBANS	Worse cognitive performance in participants with CRP ≥5.0 mg/L; no significant association between CRP and PANSS negative subscale scores	[35]
Goldsmith et al, 2018	Cross-sectional study; deficit schizophrenia (n=17), non-deficit schizophrenia (n=39), healthy controls (n=28)	Deficit: 44.7±10.7 years, male/female 7/10; non-deficit: 40.9±12.9 years, male/female 22/17; controls: 40.5±12.6 years, male/female 15/13	All participants were receiving antipsychotic medication	TNF-α, IL-6	PANSS; Deficit syndrome assessment; affective flattening, alogia	TNF-α and IL-6 were associated with deficit schizophrenia; TNF-α predicted affective flattening, alogia, and PANSS negative symptoms	[28]
Wang et al, 2022	Cross-sectional study; deficit schizophrenia (n=51), non-deficit schizophrenia (n=90), healthy controls (n=67)	Deficit: 29.92±9.27 years, male/female 24/27; non-deficit: 31.94±9.30 years, male/female 34/56; controls: 29.43±6.38 years, male/female 30/37	All participants were receiving antipsychotic medication; olanzapine-equivalent dose: deficit 11.13±7.00 mg/d, non-deficit 9.45±6.67 mg/d, p=0.292	CRP, IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12, IL-17, TNF-α, IFN-γ	PDS; PANSS; RBANS	CRP correlated with PANSS negative scores (r=0.276, p < 0.05); the overall pro-inflammatory cytokine burden correlated with negative symptoms (r=0.435, p<0.001); Anti-inflammatory factors were negatively correlated with negative symptoms (r=−0.435, p<0.001); CRP was an independent correlate of deficit syndrome.	[29]
Chen et al, 2024	Cross-sectional study; clinically stable early-onset schizophrenia (n=33), adult-onset schizophrenia (n=41), healthy controls (n=41)	Early-onset: 38.79±9.21 years, male/female 23/10; adult-onset: 46.14±7.26 years, male/female 24/17; controls: 42.61±10.30 years, male/female 23/18	Stable antipsychotic treatment for ≥1 year; chlorpromazine-equivalent dose: early-onset 653.37±300.04 mg/d, adult-onset 704.85±311.41 mg/d, p=0.475	IL-6	PANSS; expressive deficit dimension; avolition/social withdrawal dimension	In the early-onset group, IL-6 was negatively correlated with PANSS negative scores (r=−0.387, p < 0.05); with avolition/social withdrawal (r = −0.387, p < 0.05). After adjustment for covariates, IL-6 remained associated with the PANSS negative symptom score (R2 = 0.151, p < 0.05).	[31]
Zhao et al, 2024	Prospective observational study; first-episode drug-naïve schizophrenia (n=142) and healthy controls (n=100); 6-week olanzapine follow-up	First-episode: 39.04±10.82 years; controls: 41.31±12.39 years	Drug-naïve at baseline; then treated with olanzapine for 6 weeks; chlorpromazine-equivalent dose recorded	IL-2, IL-4, IL-8, IL-10, TNF-α	PANSS	IL-2 and IL-10 were positively correlated with PANSS subscales (all p < 0.05); IL-8 correlated with PANSS negative scores (r = −0.172, p < 0.05); IL-10 was an independent risk factor for treatment outcome (OR = 2.327, p < 0.05).	[40]
Catalán et al, 2025	Cross-sectional study; clinical high-risk individuals (n=35), first-episode psychosis (n=39), healthy controls (n=33)	High-risk: 22.4±5.55 years; first-episode psychosis: 27.9±9.27 years; controls: 27.5±3.77 years	-	IL-6, TNF-α	PANSS; social cognition tests	Not primarily focused on negative symptoms; IL-6 predicted poorer mentalization (β=0.28, p < 0.05); TNF-α predicted poorer facial emotion recognition (β = −1.37, p < 0.01).	[37]
Patlola et al, 2025	Cross-sectional study; schizophrenia (n=98) and healthy controls (n=202)	Schizophrenia: 43.5±12.3 years, male/female 64/34; controls: 36.2±12.3 years, male/female 117/85	All participants were receiving antipsychotic medication	IL-6, IL-8, IL-10, IL-12, TNF-α, IFN-γ, CRP; TLR	Neurocognitive tests (attention/processing speed, working memory, logical memory, etc)	Patients showed elevated IL-6, IL-8, IL-10, TNF-α, and CRP; IL-6 was associated with worse attention/processing speed and working memory; TNF-α was associated with impaired logical memory; IL-6 and IL-8 may mediate the relationship between TLR activity and cognitive impairment.	[38]
Corsi-Zuelli et al, 2025	Cross-sectional study; early-stage schizophrenia (n=201) and longer-duration schizophrenia (n=94)	Age not reported in the abstract; males: early-stage 72.8%, longer-duration 67.3%	Early-stage group: stable antipsychotic treatment; longer-duration group: clinically stable; analyses adjusted for chlorpromazine-equivalent dose	IL-6	Negative symptoms; depressive symptoms; cognition and brain structural measures	The association between IL-6 and symptoms varied by illness stage: in early-stage patients, higher IL-6 was mainly associated with depressive symptoms; in longer-duration patients, higher IL-6 was more strongly linked to negative symptoms(p < 0.05)	[32]
Moga et al, 2026	Cross-sectional study; schizophrenia (n=41) and healthy controls (n=20)	Schizophrenia: 42.10±11.60 years, male/female 23/18; controls: 40.70±10.12 years, male/female 10/10	-	IL-6, TNF-α, IL-10; lncRNA MALAT1, NEAT1, MEG3	BNSS; PANSS	IL-6 correlated with PANSS total score (r=0.48, p<0.01) and PANSS negative score (r=0.34, p < 0.05); Elevated IL-6 combined with decreased NEAT1 effectively distinguished patients with severe negative symptoms	[33]
Yan et al, 2015	Observational study; first-episode schizophrenia (n=90) and healthy controls (n=35)	First-episode schizophrenia: 31.36±12.92 years, male/female 60/30; controls: 33.10±13.24 years, male/female 25/10	Drug-naïve before treatment; after diagnosis, risperidone 1–6 mg/day for 2 months; trihexyphenidyl or benzodiazepines as needed	IL-6, TNF-α, IL-1β	PANSS	TNF-α and IL-1β were positively correlated with PANSS negative scores (r=0.554, p<0.001; r=0.278, p < 0.05, respectively)	[30]

Abbreviations: PANSS, Positive and Negative Syndrome Scale; BNSS, Brief Negative Symptom Scale; RBANS, Repeatable Battery for the Assessment of Neuropsychological Status; PDS, Proxy for the Deficit Syndrome; CRP, C-reactive protein; WBC, white blood cell count; NLR, neutrophil-to-lymphocyte ratio; IL, interleukin; TNF-α, tumor necrosis factor-alpha; IFN-γ, interferon-gamma; TGF-β1, transforming growth factor-beta 1; TLR, Toll-like receptor; lncRNA, long non-coding RNA; MALAT1, metastasis-associated lung adenocarcinoma transcript 1; NEAT1, nuclear paraspeckle assembly transcript 1; MEG3, maternally expressed gene 3; OR, odds ratio; β, standardized regression coefficient; R², coefficient of determination; NR, not reported; Ref., reference.

Clinical Evidence for rTMS in the Treatment of Negative Symptoms in Schizophrenia

Negative symptoms in schizophrenia often persist for prolonged periods and are closely associated with functional impairment, while improvements achieved through current pharmacological interventions remain limited. As a non-invasive neuromodulatory approach capable of modulating cortical excitability and related brain network activity, rTMS has increasingly been used as an adjunctive treatment for negative symptoms. Prior randomized controlled trials and systematic reviews generally suggest that rTMS may confer a degree of symptom improvement; however, findings across studies are not fully consistent. Such discrepancies may be related to differences in stimulation targets, frequency, dose, sham stimulation settings, and patients’ clinical characteristics.^41,42 Therefore, this section will summarize the existing evidence with respect to overall efficacy, stimulation parameters, symptom dimensions, and the durability of treatment effects, and will discuss potential factors influencing treatment response.

Basic Principles of rTMS and Commonly Used Stimulation Targets

rTMS can induce electric currents within the cortex through time-varying magnetic fields, thereby modulating the excitability of local neurons and related brain network activity.^4,43 Its effects are jointly influenced by stimulation parameters–such as frequency, intensity, number of pulses, and inter-train intervals–as well as target location, coil geometry, and individuals’ baseline cortical excitability. In general, high-frequency stimulation (eg, 5, 10, and 20 Hz) is more likely to enhance cortical excitability, whereas low-frequency stimulation (eg, 1 Hz) often exhibits inhibitory effects. However, this classification is not fixed and may depend on factors including the functional state of the target region, delivered dose, disease stage, and inter-individual variability.^44,45 Therefore, rTMS parameters should be interpreted in conjunction with disease characteristics and the functional state of the targeted brain region. In studies of negative symptoms in schizophrenia, the L-DLPFC is one of the most commonly used stimulation targets.^4,6 The L-DLPFC is involved in working memory, cognitive control, motivational regulation, emotional processing, and social cognition. Dysfunction of this region, as well as abnormal connectivity with the striatum and limbic system, has been proposed to be associated with negative symptoms such as avolition, diminished experience of pleasure, reduced emotional expression, and social withdrawal.^46–48 Taken together, the L-DLPFC is both functionally relevant and modifiable, and thus represents an important target in contemporary neuromodulation research on negative symptoms.

Immunological Mechanisms Underlying the Effects of rTMS on Negative Symptoms: Current Evidence and Boundaries

Regarding whether rTMS can improve negative symptoms of schizophrenia through immune-inflammatory pathways, the hierarchy and reliability of the available evidence are not uniform. A cross-disease systematic review and meta-analysis found that rTMS was associated with reduced TNF-α levels and increased BDNF levels. Qualitative synthesis also suggested a possible reduction in IL-1β, whereas IL-6 showed no consistent change.⁴⁹ However, because the included studies involved heterogeneous populations, diagnostic categories, and stimulation parameters, and were not restricted to schizophrenia, these findings cannot be directly extrapolated to schizophrenia negative symptoms. Direct clinical evidence in schizophrenia remains limited. In one study evaluating rTMS combined with risperidone, the treatment regimen was associated with reductions in pro-inflammatory factors and increases in anti-inflammatory and neurotrophic factors; however, the study primarily assessed cognitive function and aggressive behavior. Moreover, the combined-treatment design precludes attributing these effects to rTMS alone.⁵⁰ Taken together, the current cross-disease literature suggests that rTMS may modulate selected inflammatory markers, but whether a comparable immunomodulatory effect occurs in schizophrenia remains to be established.

Stimulation of the L-DLPFC can modulate local cortical excitability and influence functional connectivity with the striatum, limbic system, and hypothalamus. One proposed pathway is that changes within central networks affect peripheral immune status via stress responses and autonomic nervous system activity. However, in schizophrenia, current evidence supporting the stress–autonomic–immune pathway is still largely based on indirect inferences.^51–53 Cellular and animal studies further suggest that rTMS may modulate microglial-related signaling, inflammasome activity, and BDNF-dependent synaptic plasticity.^54–56 These findings support the biological plausibility of the proposed mechanisms, but they do not directly demonstrate that the same processes lead to clinical improvement of negative symptoms in patients with schizophrenia. Notably, cytokines released by microglia do not exert a purely pro-inflammatory or anti-inflammatory effect. For example, TNF-α and IL-6 release induced by 10 Hz magnetic stimulation may participate in synaptic plasticity processes.⁵⁴ Therefore, the effects of rTMS may not simply involve lowering inflammatory cytokine levels, but rather reshaping neural–immune signaling and its dynamic balance (Figure 2).

Figure 2.

Putative neuroimmune pathways through which rTMS may improve negative symptoms in schizophrenia. This figure was created by Figdraw2.0, https://www.figdraw.com. HF rTMS over the L-DLPFC may exert its effects through modulation of prefrontal networks, thereby influencing the HPA axis, autonomic/vagal pathways, and microglia-associated NF-κB/NLRP3 signaling. These mechanisms are proposed to be linked to the restoration of peripheral inflammatory balance, improvements in neurotrophic support and metabolism, and recovery of plasticity within the prefrontal–striatal–limbic circuitry. Blue, orange, and green denote the relatively established, emerging, and hypothetical tiers of evidence, respectively. Upward arrows (↑) indicate increased expression/activation of the indicated process or molecule, whereas downward arrows (↓) indicate decreased expression/inhibition.

Abbreviations: HF rTMS, high-frequency repetitive transcranial magnetic stimulation; L-DLPFC, left dorsolateral prefrontal cortex; HPA, hypothalamic–pituitary–adrenal; NF-κB, nuclear factor kappa-light-chain-enhancer of activated B cells; NLRP3, NLR family pyrin domain-containing protein 3; TNF-α, tumor necrosis factor-alpha; IL, interleukin; BDNF, brain-derived neurotrophic factor; VEGF-A, vascular endothelial growth factor-A; FGF-2, fibroblast growth factor-2.

Clinical Efficacy of rTMS in Improving Negative Symptoms of Schizophrenia and Its Heterogeneity

Multiple meta-analyses have indicated that rTMS, as an adjunctive treatment for negative symptoms of schizophrenia, can achieve small-to-moderate improvements in symptom severity. Variability across studies highlights the importance of stimulation parameters and procedural factors. The observed heterogeneity is mainly associated with differences in sham-stimulation procedures, discrepancies in participant characteristics, inconsistencies in stimulation dose and target localization, and variation in the selection of outcome measures.^7,57,58 At the parameter level, evidence for high-frequency stimulation of the L-DLPFC is relatively more robust. The overall results of the meta-analyses are largely consistent, and subgroup analyses further indicate that higher stimulation intensity, higher frequency, a more adequate treatment course, and relatively stable target localization are all consistently associated with better outcomes.^7,59 From a clinical standpoint, these findings support further evaluation of rTMS as an adjunctive option for appropriately selected patients receiving antipsychotic treatment.

With respect to outcome assessment, most existing studies have used the PANSS negative subscale or the total Scale for the Assessment of Negative symptoms (SANS) score as primary endpoints, and some have further conducted SANS domain-level analyses. A small double-blind study showed that HF rTMS over the L-DLPFC improved the total SANS score and certain SANS domains; however, after statistical adjustment, the between-group difference in alogia was no longer significant.⁶⁰ In addition, the relationship between higher-order factors of the BNSS and changes in negative symptoms still requires better alignment through rigorous study design and data collection. In exploratory analyses of post-treatment outcomes, a secondary analysis of one trial suggested that participants with higher baseline white blood cell counts had smaller improvements in immediate memory after rTMS.⁶¹ In China, combination-therapy studies have also reported that symptom improvement was accompanied by decreases in IL-6 and TNF-α, as well as an increase in Interleukin-10 (IL-10).^62,63 However, because these studies employed combined pharmacotherapy designs, the independent contribution of rTMS cannot be reliably disentangled at either the statistical or biological level. From a translational perspective, multimodal prediction models represent a promising direction for further investigation. For example, structural magnetic resonance imaging, clinical, and genetic information has been integrated to predict negative symptoms treatment response.⁶⁴ Whether inflammatory variables provide incremental predictive value beyond clinical and neuroimaging features remains to be tested in prespecified models and independent cohorts.⁶⁵ Taken together, CRP, IL-6, TNF-α, and white blood cell count have each been associated with negative symptoms, cognitive performance, or certain post-rTMS clinical changes, although substantial heterogeneity in sample characteristics, treatment protocols, outcome measures, and inflammatory assay methods limits direct comparison across studies (Table 2).

Table 2.

Major Clinical Evidence on rTMS for Improving Negative Symptoms and Inflammation-Related Markers in Schizophrenia

Author(s) (Year)	Study Type	Sample/Groups	rTMS Parameters and Target	Primary Outcome	Inflammatory Markers	Significance	Ref.
Shi et al (2014)	Meta-analysis	Studies on rTMS for negative symptoms	10 Hz, L-DLPFC, 110% MT, ≥3 weeks optimal	rTMS improved negative symptoms; course and parameters influenced efficacy	–	Laid foundation for parameter optimization discussions	[58]
Wang et al (2025)	Systematic review and meta-analysis	17 RCTs	20 Hz, L-DLPFC; >15 sessions subgroup more pronounced	Negative symptom scores decreased, with modest effect size	–	Supports adjunctive value of HF rTMS	[7]
Chen et al (2026)	Meta-analysis	23 RCTs, 1024 patients	HF rTMS; L-DLPFC, 110% MT, ≥15 sessions superior	HF rTMS significantly improved negative symptoms	–	Supports “adequate stimulation and stable target” strategy	[59]
Wen et al (2026)	Meta-analysis	7 randomized trials, 440 patients	iTBS vs conventional rTMS	iTBS showed greater improvement in negative symptoms with comparable safety	–	Supports exploration of brief stimulation paradigms	[66]
Tsui et al (2026)	Network meta-analysis	55 studies, 1981 TRS patients	Multiple NIBS protocols; neuronavigated iTBS over L-DLPFC, etc.	Multiple NIBS protocols improved negative symptoms	–	Highlights potential of neuronavigation and precise targeting	[67]
Lu Jing-jing et al (2022)	Randomized controlled trial	rTMS group (n=55), sham group (n=55)	10 Hz rTMS; target	Improved PANSS negative symptoms and cognitive function	IL-6, IL-10, TNF-α, TGF-β1	Inflammatory changes may serve as mechanistic monitoring clues	[62]
Dong Xiao-mei et al (2025)	Single-blind RCT	Elderly chronic schizophrenia patients	Cerebellar vermis TBS-mode rTMS, 5 days for 4 weeks	Improved negative symptoms and cognitive function	IL-6, IL-10, TNF-α	Expands cerebellar targeting and inflammatory monitoring evidence	[63]
Dong M.S. et al (2024)	Multicenter machine-learning study	Active rTMS (n=45), sham group (n=47)	rTMS; focused on response prediction	Multimodal model predicted negative symptom response; balanced accuracy 94% in active group	–	Supports shift from average efficacy to individualized stratification	[64]
Ye et al (2025)	Secondary analysis of an rTMS RCT	Active rTMS (n=56), sham group (n=28)	10-Hz and 20-Hz rTMS	Treatment for 6 weeks; cognition followed to 18 weeks	WBC	Core WBC counts indicated that peripheral burden may partly affect neuroregulatory gains; however, findings primarily pertain to cognitive, not negative symptoms	[61]

Abbreviations: rTMS, repetitive transcranial magnetic stimulation; NIBS, non-invasive brain stimulation; iTBS, intermittent theta burst stimulation; TBS, theta burst stimulation; L-DLPFC, left dorsolateral prefrontal cortex; MT, motor threshold; TRS, treatment-resistant schizophrenia; RCT, randomized controlled trial; PANSS, Positive and Negative Syndrome Scale; WBC, white blood cell count; IL, interleukin; TNF-α, tumor necrosis factor-alpha; TGF-β1, transforming growth factor-beta 1; NR, not reported; Ref., reference; HF, high-frequency.

Compared with conventional HF rTMS, intermittent theta burst stimulation (iTBS) offers greater time efficiency, and neuronavigation may improve the consistency of target localization.⁶⁶ Nevertheless, because there is substantial heterogeneity across stimulation dose, target selection, electrode/coil configuration, participant characteristics, and outcome measures, these different modalities remain difficult to compare directly. Although some meta-analyses suggest potential benefits of iTBS, anodal transcranial direct current stimulation (a-tDCS) to the frontal lobe, or deep transcranial magnetic stimulation (dTMS) for negative symptoms, ranking by navigated continuous theta burst stimulation (cTBS) or iTBS in network meta-analyses should not be interpreted as demonstrating superiority over conventional HF rTMS–particularly given limited certainty of evidence and the absence of adequately powered head-to-head trials using uniform targets, stimulation parameters, and core outcomes.^66–68 Preliminary placebo-controlled studies also indicate that deep TMS may be beneficial for negative symptoms; however, the evidence remains confined to small trials.⁶⁹ Therefore, the main implications of these techniques for the present review are as follows: future studies should treat stimulation pattern, target localization, and stimulation dose as pre-specified stratification factors, and evaluate whether these variables moderate the associations between candidate inflammatory phenotypes and clinical response, rather than selecting a so-called “optimal” intervention based solely on existing rank orders.

Discussion

Evidence Boundaries

Existing evidence can be broadly categorized into three tiers. First, HF rTMS over the L-DLPFC has been suggested by multiple randomized trials and meta-analyses to provide small-to-moderate adjunctive benefits for negative symptoms. Second, peripheral inflammatory abnormalities have been repeatedly associated with certain symptom dimensions, cognition, and functioning; however, their symptom dimension specificity remains limited. Third, whether inflammation mediates rTMS efficacy, and whether baseline indicators can predict treatment response, are still largely unsupported by direct evidence.^66–68 Regarding the stimulation protocols themselves, although iTBS, tDCS, and deep TMS have shown promise in some meta-analyses, the lack of a clear ranking advantage is insufficient to establish superiority, because head-to-head studies conducted on the same target with the same outcome measures are lacking. Accordingly, such findings should be considered hypotheses and leads for subsequent research rather than evidence for clinical selection.^69–71

Key Confounding Factors and the Boundary of Clinical Translation

The inflammation–symptom relationship and its treatment responsiveness are commonly confounded by sex, antipsychotic exposure, and metabolic syndrome. These factors are not independent sources of variation; rather, they form a coupled interaction network linking inflammation–metabolism–medications. Sex is an important determinant of clinical heterogeneity in schizophrenia. It is reflected not only in differences in age at onset, symptom profiles, social functioning, and adverse drug reactions, but also–potentially–through immunoregulatory and pharmacodynamic mechanisms that shape inflammation-related phenotypes.⁷² Prior studies have demonstrated pronounced sex differences in the clinical manifestations of schizophrenia. Male patients often have an earlier onset, more severe negative symptoms, and greater impairment in social functioning. By contrast, female patients tend to have a later onset, with relatively more prominent affective symptoms, yet they are more likely to experience metabolic and endocrine adverse effects during treatment.⁷² At the immunological level, longitudinal evidence indicates that the trajectories of cytokine changes during antipsychotic treatment differ between women and men: women more frequently show dynamic increases in Interleukin-4 (IL-4), IL-6, and TNF-α, whereas men exhibit more stable overall changes, suggesting a possible sex-specific inflammatory response.⁷³ In addition, associations between peripheral inflammatory markers such as the neutrophil-to-lymphocyte ratio (NLR) and symptoms may vary by sex.⁷⁴

The complexity of these sex differences largely reflects effect modification by antipsychotic exposure itself. Variations among different agents in terms of drug exposure, treatment duration, and metabolic side effects may all influence peripheral inflammatory markers. Moreover, weight gain and metabolic adverse reactions induced by different antipsychotics are not uniform, which further amplifies heterogeneity in inflammatory phenotypes across sexes.^40,75–77 Metabolic syndrome is also an important potential confounder. Prior studies have shown that metabolic syndrome is relatively common in patients with schizophrenia. Low-grade systemic inflammation associated with metabolic syndrome, together with metabolic changes induced by antipsychotic medications, may affect the interpretation of associations among inflammatory biomarkers, clinical symptoms, functional outcomes, and treatment response.⁷⁸ Beyond sex and metabolic status, additional factors–including smoking, diet, physical inactivity, insufficient sleep, recent infections, and acute stress–may also alter inflammatory measurements. If these variables are not systematically recorded, the interpretability of the results remains limited even after statistical adjustment. Therefore, future research should specify how these factors are captured at the study-design stage and incorporate them into a standardized data framework. Based on the above principles, the minimum dataset for studies of inflammatory biomarkers in schizophrenia should include at least the following: antipsychotic type, dose, treatment duration, chlorpromazine-equivalent dose, recent changes in dose, and, when feasible, plasma drug concentrations. In addition, the following variables should be recorded: age, sex, duration of illness, smoking, substance use, body mass index (BMI) or waist circumference, fasting glucose, blood lipids, metabolic syndrome, recent infections, and the use of anti-inflammatory medications. Sleep-related variables, including sleep quality, total sleep duration, and sleep disorders, should also be collected. Notably, sleep should be repeatedly recorded and analyzed as a potential time-varying confounder, rather than being assessed only once at baseline.⁷⁹

Evidence Limitations and Future Longitudinal Research Directions

Existing studies on the inflammation–negative symptom relationship are predominantly cross-sectional, which limits the ability to establish temporal ordering and to rule out reverse causality as well as confounding factors such as medication exposure and metabolic status. Accordingly, these findings are more appropriately interpreted as reflecting “associations” rather than “causality”. Longitudinal evidence helps address some of these gaps. In the BeneMin study follow-up, TNF-α was associated with decreased expression, whereas CRP showed no significant relationship with any negative symptoms dimension. In addition, minocycline did not improve negative symptoms overall or in the subgroup with increased inflammation.⁸⁰ However, other studies indicate that CRP changes dynamically with medication exposure.⁸¹ Therefore, future research should obtain repeated measurements at multiple clinically relevant time points, such as before treatment, early in treatment, at the end of the treatment course, and during follow-up, and align inflammatory biomarkers with assessments of BNSS negative symptom dimensions, cognition, and social functioning.^82,83 This approach is intended to disentangle predictive, mediating, and concurrent changes. To organize these complex interrelationships into a framework that can be empirically tested, we developed a conceptual model (Figure 3) to guide subsequent research. This model integrates potential pathways linking baseline inflammatory burden, negative symptom dimensions, and clinical response to rTMS, while rigorously distinguishing between hypotheses supported by existing correlational evidence and hypotheses that remain to be validated as predictive or mediating mechanisms. However, in the absence of longitudinal data, stratified conclusions derived from cross-sectional analyses should be interpreted with caution.

Figure 3.

Hypothesized relationships among baseline peripheral inflammatory markers, BNSS negative symptom dimensions, and rTMS response. This figure was created by Figdraw2.0, https://www.figdraw.com. The left panel shows candidate inflammatory markers, the middle panel presents the two higher-order symptom dimensions, and the right panel shows clinical and optional mechanistic outcomes. Solid lines indicate relatively established pathways, whereas dashed lines indicate exploratory or unvalidated pathways; no proven causal relationship is implied.

Abbreviations: IL, interleukin; TNF-α, tumor necrosis factor-alpha; CRP, C-reactive protein; NLR, neutrophil-to-lymphocyte ratio; BNSS, Brief Negative Symptom Scale; HF rTMS, high-frequency repetitive transcranial magnetic stimulation; L-DLPFC, left dorsolateral prefrontal cortex.

Clinical Translation and Practical Feasibility

From a clinical-translation perspective, inflammatory biomarkers are currently more appropriately used as research stratification tools rather than as direct inputs for routine rTMS decision-making. Because high-sensitivity C-reactive protein (hs-CRP) is readily accessible and relatively inexpensive, it may serve as a candidate marker for studies and for preliminary screening.⁸⁴ Although conventional hs-CRP classifications, such as <1, 1–3, and ≥3 mg/L, have been used for cardiovascular risk assessment, these cut-off values have not been established as reproducible thresholds that are specific to schizophrenia or to rTMS.^85,86 Prior schizophrenia studies have explored the 1–3 mg/L range in an exploratory manner.⁸⁷ However, no currently validated threshold is available to guide rTMS-related clinical decision-making.⁸⁴ Therefore, these classifications should not be directly extrapolated to determine rTMS eligibility, guide treatment selection, or stratify patients. Cytokine panels such as IL-6 and TNF-α can provide additional support for mechanistic investigations and exploratory stratification. However, their results are sensitive to the timing of sample collection, fasting status, recent infection, assay platform, and sample-processing procedures, and thus they remain largely confined to research settings.^8,13 TSPO-PET is closer to capturing central immune processes; however, it is influenced by tracer characteristics, genetic polymorphisms, medication exposure, and disease stage. In addition, it is costly and not widely accessible, making it more suitable for centralized mechanistic studies or confirmatory trials.²¹ Genotyping of C4 and C4A reflects genetic risk background; however, it provides limited explanatory power regarding the current inflammatory status and response to rTMS. Therefore, it should be considered an exploratory indicator at this stage.²²

Based on this cost–feasibility continuum, we propose advancing along a discovery–validation–translation pathway. We propose that future discovery studies should measure CRP/hs-CRP, IL-6, TNF-α, and IL-10 under standardized sampling conditions and examine their associations with negative symptom measures and treatment response. Next, patients are stratified using prespecified thresholds, and the corresponding response patterns across distinct inflammatory phenotypes are evaluated. Additionally, a research-oriented patient stratification process should be established. First, among patients with clearly defined negative symptoms who have undergone standardized clinical assessment, acute infections and other factors that may influence negative symptom severity or inflammatory marker levels should be excluded. These include obvious depressive symptoms, extrapyramidal symptoms, excessive sedation, and other relevant confounders. Meanwhile, antipsychotic exposure and BMI should be recorded. Finally, in the validation cohort, enrichment, stratified randomization, or response-adaptive allocation designs are employed to determine whether this stratification strategy genuinely improves the ability to identify treatment effects and yields clinically meaningful net benefit.

Correspondingly, outcomes should jointly capture symptom severity and functional impairment. Specifically, the BNSS total score should be used as the primary endpoint, whereas the two higher-order domains should serve as key secondary endpoints and be evaluated independently. The Specific Levels of Functioning (SLOF) should be used for functional assessment and longitudinal follow-up to confirm the persistence of benefit.^88–90 From a statistical-analysis perspective, studies should clearly distinguish between confounding control, prognostic associations, treatment prediction, and mechanistic mediation. Predictive value should be assessed through tests of interactions between treatment and biomarkers, rather than relying solely on simple correlation analyses. If the goal is to develop a predictive model, it should also be externally validated in an independent cohort with respect to discrimination, calibration, incremental value, and clinical net benefit.^91–93 Building on this, adaptive designs may be employed to investigate differential treatment responses across inflammatory subgroups. For example, trials may incorporate biomarker-stratified randomization based on baseline inflammatory status, include an internal pilot followed by sample-size re-estimation, and apply response-adaptive randomization or an enrichment design for participants in the high-inflammation subgroup. Although adaptive designs hold promise, the stratification thresholds, sample reallocation rules, timing of interim analyses, stopping criteria, and type I error control should be specified a priori in the protocol to avoid post hoc selective interpretation.^94–96 Meanwhile, any biomarker-guided stratification must be implemented entirely within the established rTMS safety framework and must not compromise the routine screening requirements for seizure risk, sleep deprivation, substance use, intracranial pathology, or compatibility with implanted devices.^97,98 Building on this foundation, although emerging multimodal models that integrate multi-omics data have demonstrated predictive potential, their clinical utility still requires further multicenter validation to confirm reproducibility, predictive performance, net benefit, and cost-effectiveness.⁶⁴

Conclusion

In summary, negative symptoms in schizophrenia exhibit substantial clinical and biological heterogeneity. HF rTMS over the L-DLPFC may confer adjunctive benefit to a subset of patients; however, its efficacy is jointly determined by stimulation parameters, target localization, symptom phenotypes, and individual patient characteristics. Peripheral inflammatory abnormalities are associated with negative symptoms as well as cognitive and social functioning, yet the current evidence is insufficient to support any single inflammatory biomarker as an independent tool for treatment decision-making. Future studies should prioritize validating reproducible inflammatory predictors and should, in prospective randomized sham-controlled trials, integrate negative symptom scales such as the BNSS, functional outcomes, and prespecified biomarker analyses. These predictors should be further validated for reproducibility, external generalizability, and clinical net benefit in independent cohorts. Until then, no single inflammatory marker, nor any combination thereof, should be used alone to determine rTMS treatment eligibility, parameter selection, or stratification decisions.

Funding Statement

This study was supported by grants from the following funding programs: National Natural Science Foundation of China (82360275, 82560282, 82160269); Yunnan Provincial Department of Science and Technology Program Projects (202301AY070001-020, 202405AC350104, 202203AC100007, 202301AY070001-145); and Scientific Research Project of Yunnan Provincial Clinical Medical Center for Mental Disorders (2024YNLCYXZX0277, 2024YNLCYXZX0278).

Disclosure

The authors report no conflicts of interest in this work.