Exploring the Diagnostic and Predictive Value of Oral Microbiome in Esophageal Cancer: A Systematic Review and Meta-Analysis

Authors

Jie-Chi Chen

Min-Hsun Hsu

Suh-Woan Hu

Yuh-Yih Lin

Doi

10.3390/ijms26199457

PMID: 24670651 · DOI: 10.3390/ijms26199457 · Journal: International Journal of Molecular Sciences (2025)

TL;DR

Chen et al. performed a PRISMA-guided systematic review and fixed-effect meta-analysis of 18 case–control studies (1,191 cases / 1,403 controls, published 2015–2024) examining the oral microbiome in esophageal cancer. Across studies — most using saliva and 16S rRNA gene sequencing — alpha-diversity findings were inconsistent but beta-diversity differences between cases and controls were significant in 12 of 17 studies. Several ESCC studies reported elevated Prevotella, Porphyromonas (notably P. gingivalis), and Fusobacterium (notably F. nucleatum), while two esophageal adenocarcinoma studies reported elevated Actinomyces species. A fixed-effect meta-analysis of the two studies with extractable odds ratios yielded a pooled OR of 9.50 (95% CI: 5.89–15.29) for ESCC in individuals harboring specific oral microbial signatures.

Cohort & data

  • 18 case–control studies published 2015–2024; aggregate 1,191 cases / 1,403 controls.
  • Geographic distribution: China (n=12), USA (n=4), Japan (n=1), Taiwan (n=1).
  • Histology: 11 ESCC-only, 2 EAC-only, 1 mixed ESCC+EAC, 2 unspecified, 1 precancerous esophageal lesions, 1 early-stage intramucosal ESCC.
  • Sample types: saliva (n=13), plus oral wash, tongue coating, buccal mucosa, oral swabs, oral biofilms.
  • Assay: 16S rRNA gene sequencing in all studies; V3–V4 region in 9, V4 in 5, others V1–V2 or V3–V5. Platforms: Illumina MiSeq / HiSeq 2500 / NovaSeq 6000 / MiniSeq, plus Ion Torrent (PGM, S5) and 454 Roche FLX in earlier work.
  • Cancer types covered: ESCC (Esophageal Squamous Cell Carcinoma) and ESCA (Esophageal Adenocarcinoma).
  • Search: Embase, PubMed, Web of Science (inception to 6 July 2024); 2,544 records → 1,993 after dedup → 55 full-text → 18 included. PROSPERO ID 1109590.
  • Quality assessment: CASP Case–Control Study Checklist; 6 studies received >9 green lights, only 1 received <5.

Key findings

  • Pooled risk estimate. Fixed-effect meta-analysis of two ESCC studies (Wei 2022 [PMID:35873026], Chen 2022 [PMID:36389140] — neither in current corpus) yielded an OR = 9.50 (95% CI: 5.89–15.29) for ESCC in individuals with the specified oral microbial signatures; individual study ORs ranged 9.01–10.53.
  • Beta-diversity is the more robust signal. Significant beta-diversity (UniFrac or Bray–Curtis) was reported in 12 of 17 studies; alpha-diversity (Shannon, Chao1) was inconsistent — decreased in 4 studies, unchanged in 8, modestly increased in 6.
  • Recurrently enriched genera in ESCC: Prevotella, Porphyromonas, Fusobacterium, Streptococcus. Recurrently depleted: Neisseria, Lautropia, Corynebacterium.
  • Species-level ESCC hits: Porphyromonas gingivalis (multiple studies; Chen MF 2021, Wei 2022, Peters 2017, Kageyama 2019); Fusobacterium nucleatum (Liu 2020, Wei 2022, Li M 2024); Streptococcus salivarius (Wei 2022); Prevotella intermedia / P. histicola (Liu 2020, Li M 2024); Treponema vincentii (Peters 2017, Liu 2020).
  • EAC-specific signal: two EAC studies (Hao 2022, Solfisburg 2024) reported enrichment of Actinomyces bowdenii and Atopobium parvulum, and a Streptococcus shift across the Barrett’s → dysplasia → adenocarcinoma sequence.
  • Diagnostic performance reported in only two studies: sensitivity 68.2–86.4%, specificity 58.8–96.1% for microbial-panel-based ESCC classification (Wei 2022, Chen X 2022).
  • Beyond mid-2024 update: four additional studies identified in July 2025 pre-submission re-check — an ESCC classifier reached AUC 0.87 (markers including Neisseria perflava and Haemophilus parainfluenzae); a saliva model reached AUC 0.791 with decreased Akkermansia and Escherichia-Shigella linked to elevated cancer risk.
  • Mendelian-randomization evidence: a referenced MR study found causal effects of specific oral taxa on esophageal cancer with no reverse causal effect of cancer on the oral microbiome, supporting (but not proving) a directional contribution of dysbiosis to carcinogenesis.

Genes & alterations

The review does not generate new genomic data. As background context the authors cite the canonical ESCC genomic landscape (citing Song et al. 2014, Nature — not in current corpus):

  • TP53 — mutated in ESCC.
  • RB1 — mutated in ESCC.
  • CDKN2A — mutated in ESCC.
  • PIK3CA — mutated in ESCC.
  • NOTCH1 — mutated in ESCC.
  • NFE2L2 — mutated in ESCC.

Frequently affected ESCC pathways named in the discussion: Wnt, cell cycle, Notch, RTK-Ras, and AKT signaling. Proposed microbial → host mechanisms invoke TLR4–NF-κB activation (P. gingivalis LPS), FadA-adhesin / E-cadherin / β-catenin engagement (F. nucleatum), and Bcl-2 / Bad / caspase-9 / Jak1-Akt-Stat3 modulation (P. gingivalis) — but no gene-level alteration calls are made by this review.

Clinical implications

  • Biomarker potential. Specific oral microbial signatures (Prevotella, Porphyromonas, Fusobacterium enrichment) may serve as non-invasive early-detection biomarkers for ESCC; the pooled OR of 9.50 supports clinical relevance though it derives from only two pooled studies.
  • EAC under-studied. ESCC-derived microbial markers should not be assumed to generalize to EAC; the authors call for EAC-specific cohorts with standardized methodology.
  • Therapeutic angle. He Y 2024 (cited reference 43) reported salivary microbiome shifts associated with response to chemoradiotherapy in inoperable ESCC, raising the possibility of microbiome-based response prediction or microbiome-targeted adjuncts.
  • No drugs or treatment regimens are evaluated by this paper itself.

Limitations & open questions

  • Only 2 of 18 studies had data sufficient for quantitative meta-analysis; pooled OR (9.50) rests on a thin quantitative base.
  • Heterogeneity in methodology — sample type (saliva vs. swab vs. tongue coating), 16S region (V3–V4 vs. V4 vs. V1–V2 vs. V3–V5), platform (Illumina MiSeq/HiSeq/NovaSeq, Ion Torrent, 454), and bioinformatic pipelines — limits direct comparability.
  • Confounder control is inconsistent. Few studies adjusted comprehensively for smoking, alcohol, age, PPI use, and oral health — all of which independently affect both oral microbiome composition and esophageal cancer risk.
  • Causality undetermined. Observational data cannot distinguish whether microbial shifts drive carcinogenesis, reflect tumor-induced ecological changes, or are confounded by shared risk factors. Authors flag the possibility of bidirectional causation.
  • Geographic / population bias. 12 of 18 studies are from China; generalizability to other populations is limited.
  • EAC evidence gap. Only 2 EAC-only studies (+1 mixed) — too few to draw firm conclusions for adenocarcinoma.
  • No mechanistic experimentation in this review. Mechanisms (TLR4–NF-κB, FadA–E-cadherin, butyrate/H₂S metabolism) are summarized from primary literature, not validated here.
  • Open question for cross-paper synthesis: whether oral microbiome signatures correlate with the canonical ESCC genomic alteration landscape (TP53, NOTCH1, NFE2L2 etc.) — not addressed by any included study.

Citations from this paper used in the wiki

  • “Eighteen studies published between 2015 and 2024 were obtained, involving 1191 cases and 1403 controls, mostly using oral saliva samples and 16S rRNA gene sequencing.” (Abstract)
  • “A fixed-effect meta-analysis of two studies showed that individuals with specific oral microbial signatures had significantly higher odds of developing esophageal squamous cell carcinoma (OR = 9.50; 95% CI: 5.89–15.29).” (Abstract)
  • “Notably, several investigations on esophageal squamous cell carcinoma showed higher levels of Prevotella, Porphyromonas, and Fusobacterium, while two studies on esophageal adenocarcinoma reported elevated levels of Actinomyces species.” (Abstract)
  • “On a molecular level, mutations in tumor-related genes including TP53, RB1, CDKN2A, PIK3CA, NOTCH1, and NFE2L2 are commonly found in ESCC, and abnormal epigenetic modifications are also implicated in its progression. Genomic alterations in ESCC frequently involve the Wnt, cell cycle, Notch, RTK-Ras, and AKT signaling pathways.” (Discussion, citing Song et al. 2014)
  • “Significant beta-diversity was observed in 12 out of 17 studies that reported this metric, employing methods such as UniFrac and Bray–Curtis.” (Results §3.4)
  • “Two studies reported sensitivities ranging from 68.2% to 86.4% and specificities between 58.8% and 96.1%, suggesting that combinations of microbial taxa may offer promising non-invasive screening approaches.” (Results §3.5)
  • “A recent Mendelian randomization (MR) study identified significant causal relationships between specific oral microbial taxa and esophageal cancer, while reverse MR analyses did not support a causal effect of esophageal cancer on the oral microbiome.” (Discussion)

This page was processed by crosslinker on 2026-05-11.