Cited by 432Tryptophan and brain disordersStress Responses and CortisolNeuroinflammation and Neurodegeneration Mechanisms

The DOI matches its Crossref record; 81 of the 81 references it cites were found in a registry or an index; 5 of the 63 sources its claims cite were checked for support.

Inflammation-related biomarkers in major psychiatric disorders: a cross-disorder assessment of reproducibility and specificity in 43 meta-analyses

Open access
Yuan, Ning, Chen, Yu, Xia, Yan, Dai, Jiacheng, Liu, Chunyu·2019·Translational Psychiatry

Abstract

Inflammation is a natural defence response of the immune system against environmental insult, stress and injury, but hyper- and hypo-inflammatory responses can trigger diseases. Accumulating evidence suggests that inflammation is involved in multiple psychiatric disorders. Using inflammation-related factors as biomarkers of psychiatric disorders requires the proof of reproducibility and specificity of the changes in different disorders, which remains to be established. We performed a cross-disorder study by systematically evaluating the meta-analysis results of inflammation-related factors in eight major psychiatric disorders, including schizophrenia (SCZ), bipolar disorder (BD), autism spectrum disorder (ASD), major depression disorder (MDD), post-trauma stress disorder (PTSD), sleeping disorder (SD), obsessive–compulsive disorder (OCD) and suicide. A total of 43 meta-analyses involving 704 publications on 44 inflammation-related factors were included in the study. We calculated the effect size and statistical power for every inflammation-related factor in each disorder. Our analyses showed that well-powered case–control studies provided more consistent results than underpowered studies when one factor was meta-analysed by different researchers. After removing underpowered studies, 30 of the 44 inflammation-related factors showed significant alterations in at least one disorder based on well-powered meta-analyses. Eleven of them changed in patients of more than two disorders when compared with the controls. A few inflammation-related factors showed unique changes in specific disorders (e.g., IL-4 increased in BD, decreased in suicide, but had no change in MDD, ASD, PTSD and SCZ). MDD had the largest number of changes while SD has the least. Clustering analysis showed that closely related disorders share similar patterns of inflammatory changes, as genome-wide genetic studies have found. According to the effect size obtained from the meta-analyses, 13 inflammation-related factors would need <50 cases and 50 controls to achieve 80% power to show significant differences ( p < 0.0016) between patients and controls. Changes in different states of MDD, SCZ or BD were also observed in various comparisons. Studies comparing first-episode SCZ to controls may have more reproducible findings than those comparing pre- and post-treatment results. Longitudinal, system-wide studies of inflammation regulation that can differentiate trait- and state-specific changes will be needed to establish valuable biomarkers.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

This cross-disorder systematic review of 43 meta-analyses found that well-powered studies yield more consistent inflammation-related biomarker findings across major psychiatric disorders. After filtering underpowered meta-analyses, 30 of 44 inflammation-related factors showed significant alterations in at least one disorder, with 11 changing in more than two disorders.

Key findings

  • The review included 43 meta-analyses involving 704 publications on 44 inflammation-related factors across eight disorders; 26% of meta-analyses were underpowered (<80%). After removing underpowered data, 30 of 44 factors showed significant alterations in at least one disorder, and 11 changed in more than two disorders.
  • Consistency improved after power filtering: inconsistent case–control results fell from 42% (16/38) to 16% (5/31), and inconsistent state-related results fell from 67% (24/36) to 48% (15/31). Well-powered case–control findings had 84% consistency; SCZ and BD case–control comparisons showed 100% consistency.
  • Specificity examples: IL-4 increased in BD, decreased in suicide, and showed no change in MDD, ASD, PTSD, or SCZ; IL-2 was uniquely low in suicide; sIL-6R was uniquely high in BD; sIL-2R was uniquely not changed in PTSD. IL-6 and CRP were the most commonly increased factors, each changed in four disorders; NGF was the most commonly decreased factor in MDD and SCZ.
  • Based on effect sizes, 13 inflammation-related factors would need <50 cases and 50 controls to achieve 80% power at p < 0.0016. MDD had the largest number of changes and SD the least. Clustering showed BD and SCZ closer to each other, while suicide clustered with PTSD and SD.

Methods

  • Systematic review of meta-analysis publications in PubMed, Springer, and Web of Science before March 2018, following PRISMA; inclusion required English-language meta-analyses of circulating cytokine/chemokine/inflammation-related factor levels in plasma, serum, or CNS. The final set comprised 43 meta-analyses on 44 factors in eight disorders: SCZ (n = 11), MDD (n = 16), BD (n = 6), OCD (n = 1), suicide (n = 2), ASD (n = 1), PTSD (n = 1), SD (n = 1), and cross-disorder studies (n = 4); anxiety did not meet inclusion criteria.
  • Data were partitioned into trait marker analyses (undifferentiated case–control comparisons; 24 studies), state marker analyses (6 studies), or both (13 studies). Effect sizes were extracted as standardized mean differences, statistical power was calculated with G*Power, and data with power <0.8 were filtered out. Stability across repeated meta-analyses was used as a proxy for reproducibility, and unsupervised clustering assessed cross-disorder similarity.

Limitations

  • The paper states that it is built on meta-analysis results, so it cannot recover signals missed by original studies due to technical limitations, and meta-analyses can be misled by systematic bias in individual data. Independence of samples was problematic because meta-analyses often reuse the same underlying studies, so stability was used as a proxy for reproducibility rather than true independent replication.
  • The paper states that many effect sizes may be inflated because some factors have limited studies, and independent testing is needed. State-related data carry more uncertainty than trait-related data, with frequent inconsistencies in pre- and post-treatment comparisons; first-episode SCZ findings were considered more reliable. Few prior studies reported statistical power, quality control, covariate control, or multiple-testing correction.
  • Evident from the text: most data came from peripheral blood, with only one brain study and limited CSF data; the review covered only eight psychiatric disorders and excluded anxiety; most included studies focused on candidate factors rather than system-wide measures; and longitudinal data were rare, limiting differentiation of trait and state markers.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41398-019-0570-y

Cites

81 of 81 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Clustering of disorders based on alterations of 38 inflammation-related factors identified from sufficiently powered meta-analyses in blood samples. Red cells are the increased changes in disorders (effect size (ES) > 0, P < 0.05 and power > 0.8); blue cells are the decreased changes in disorders (ES < 0, P < 0.05 and power > 0.8); yellow cells are the not significant changes in disorders (P > 0.05, power > 0.8); white cells represent missing data, either not being studied or only an underpowered study in the meta-analyses; orange cells with question marks are those with inconsistent results. “*” in the cells indicates that the significant changes are found in more than two meta-analyses. MDD, ASD, PTSD, suicide, SD, BD and SCZ have 16, 7, 3, 5, 2, 8 and 7 IRF changes, respectively
Fig. 1 Clustering of disorders based on alterations of 38 inflammation-related factors identified from sufficiently powered meta-analyses in blood samples. Red cells are the increased changes in disorders (effect size (ES) > 0, P < 0.05 and power > 0.8); blue cells are the decreased changes in disorders (ES < 0, P < 0.05 and power > 0.8); yellow cells are the not significant changes in disorders (P > 0.05, power > 0.8); white cells represent missing data, either not being studied or only an underpowered study in the meta-analyses; orange cells with question marks are those with inconsistent results. “*” in the cells indicates that the significant changes are found in more than two meta-analyses. MDD, ASD, PTSD, suicide, SD, BD and SCZ have 16, 7, 3, 5, 2, 8 and 7 IRF changes, respectivelyCC BY · PubMed Central

Verification

Claims the paper makes

50 Own result · 26 Not checked yet · 21 No basis given · 13 Own interpretation · 5 Own method · 1 Partly supported · accuracy not yet measured

Verify all 58 references — free account

Show all 71 claims

  • Probably not a substantial claim. 28
  • It may state two separate claims as one. 24
45 of 45
  • The authors state this as an assumption.

    Stable findings are more likely to be reproducible than unstable ones.

    “Even though stability is not a true measure of reproducibility, we assume that stable findings are more likely to be reproducible than unstable ones.”
  • The authors state this as an assumption.

    All data including those tested only once will have better reproducibility.

    “We applied this filter to all data, assuming that all data including those tested only once will have better reproducibility.”
  • The paper relies on this without support.

    Candidate genes are generally not the top signals in genome-wide or system-wide studies.

    “Candidate genes are generally not the top signals in genome-wide or system-wide studies.”
  • The paper relies on this without support.

    Sample size is one of the major drivers of statistical power and is a limiting factor for many studies.

    “Sample size, one of the major drivers of statistical power, is a limiting factor for many studies.”
  • The paper relies on this without support.

    Genetic studies of candidate genes have produced a large number of false positives in small-sample studies due to loose statistical criteria.

    “Genetic studies of candidate genes have produced a large number of false positives in small-sample studies due to loose statistical criteria.”
  • The paper relies on this without support.

    Inflammation levels likely also fluctuate temporally due to environmental and circadian rhythm changes regardless of health status.

    “The inflammation levels likely also fluctuate temporally due to environmental and circadian rhythm changes regardless of health status.”
  • The paper relies on this without support.

    The clinical symptoms can change through the disease course while the diagnosis remains unchanged.

    “The clinical symptoms can change through the disease course while the diagnosis remains unchanged.”
  • The paper relies on this without support.

    Hundreds of genes or proteins are involved in immunity and inflammation regulation.

    “Hundreds of genes or proteins are involved in immunity and inflammation regulation and should be studied in relation to psychiatric disorders.”
8 of 8

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 81 references

Cited by

Sorted by citations

Items 301–400 of 432

Metadata

Authors
Yuan, N., Chen, Y., Xia, Y., Dai, J., & Liu, C.
Affiliations
Central South University (CN)
Corresponding author
Liu, Chunyu
Venue
Translational Psychiatry
Published
2019-09-18
Volume
9
Issue
1
Pages
233-233
ISSN
2158-3188 (ISSN-L)
OpenAlex
W2974338024
Cited by
428
References
81
Citations by year
2026: 44 · 2025: 77 · 2024: 80 · 2023: 59 · 2022: 76 · 2021: 59 · 2020: 31 · 2019: 2
FWCI
13.79
Type
journal article
Language
en
Open access
Locations
  • Translational Psychiatry (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
  • National Natural Science Foundation of China (31871276)
  • Central South University
  • Hunan University
  • Natural Science Foundation of Hunan Province
  • National Institutes of Health
  • National Institute on Aging (R21AG045789)
  • National Institute of Mental Health (U01MH116489, R01MH110920, U01MH103340)
Topics
  • Life Sciences › Neuroscience › Biological Psychiatry › Tryptophan and brain disorders
  • Life Sciences › Neuroscience › Behavioral Neuroscience › Stress Responses and Cortisol
  • Life Sciences › Neuroscience › Neurology › Neuroinflammation and Neurodegeneration Mechanisms
Keywords
inflammation, inflammatory markers, mental illness, cross-disorder analysis, reproducibility, statistical power, underpowered studies, schizophrenia, bipolar disorder, autism spectrum disorder, major depressive disorder, post-traumatic stress disorder, sleep disorders, obsessive-compulsive disorder, suicide, IL-4, first-episode schizophrenia, meta-analysis
MeSH terms
Mental Disorders, Humans, Inflammation, Sensitivity and Specificity, Meta-Analysis as Topic, Reproducibility of Results, Biomarkers

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review