Prognostic and clinicopathological significance of tertiary lymphoid structure in esophageal squamous cell carcinoma: a systematic review and meta-analysis review

Abstract

Tertiary lymphoid structures (TLS), ectopic immune cell aggregates in non-lymphoid tissues, have emerged as potential predictors of esophageal squamous cell carcinoma (ESCC) outcomes. Given increasing evidence, we conducted an updated meta-analysis to systematically evaluate their prognostic and clinicopathological significance. A comprehensive literature search was performed through PubMed, Embase, Web of Science, Scopus, and Cochrane Library (up to June 2024) for studies assessing TLS associations with TNM staging and survival outcomes (OS/PFS) in ESCC. Pooled odds ratios (ORs) and hazard ratios (HRs) with 95% confidence intervals (CIs) were calculated using random-effects models. This meta-analysis incorporated seven studies comprising nine separate datasets that evaluated the impact of TLS in ESCC. The pooled analysis demonstrated a significant positive correlation between TLS presence and more advanced T stage (OR = 2.65, 95%CI: 1.86–3.78; p < 0.01) but not N stage (OR = 1.27, 95%CI: 0.85–1.89; p = 0.24). Additionally, TLS presence was significantly associated with enhanced overall survival (HR = 0.49, 95%CI: 0.41–0.58, p < 0.01) as well as progression-free survival (HR = 0.56, 95%CI: 0.45–0.69, p < 0.01). Notably, when assessed using combined HE and IHC criteria, the prognostic benefit of TLS was more pronounced, with HRs further decreasing to 0.40 (95% CI: 0.31–0.51) for OS and 0.50 (95% CI: 0.41–0.60) for PFS. These findings confirm that the presence of TLS, particularly when verified through combined HE staining and IHC, is an independent favorable prognostic factor in ESCC patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-025-14997-x.

Introduction

Esophageal cancer is an extremely aggressive malignant tumor with poor prognosis, ranking as the seventh leading cause of cancer-related deaths worldwide [1]. Esophageal squamous cell carcinoma (ESCC) is a predominant histological subtype, accounting for over 90% of esophageal cancer cases in China [2]. The prognosis of ESCC exhibits significant heterogeneity, with patients at the same TNM stage and receiving similar treatments showing significantly different survival outcomes [3]. Consequently, there is an urgent need for additional prognostic factors to improve risk stratification and guide personalized treatment strategies.

Tertiary lymphoid structures (TLS) are organized aggregates of B and T cells that form within non-lymphoid tissues [4]. They have emerged as a critical component in tumor adaptive immunity by providing a specialized tumor microenvironment that facilitates interactions between immune and tumor cells [5, 6]. The presence and maturation status of TLS within tumors have been associated with clinical outcomes, with multiple independent studies demonstrating this prognostic value remains significant after rigorous adjustment for TNM stage [5, 7]. This finding may contribute to a deeper understanding of cancer biology and the development of novel treatment strategies [8]. Recent studies also indicate that TLS may correlate with TNM staging, further highlighting their importance in understanding tumor biology and guiding therapeutic strategies [9, 10].

In this systematic review and meta-analysis, we aim to elucidate the relationship between TLS and TNM staging in ESCC, as well as evaluate the potential of TLS as a prognostic indicator and examine its implications for patient management strategies.

Material and methods

Protocol and registration

This study was registered with PROSPERO (ID: CRD42024544499; see Supplementary Table 1) and adheres to both the Meta-Analysis of Observational Studies in Epidemiology [11] reporting guidelines and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [12] statement (Supplementary Table 2).

Literature search procedure

The following databases were systematically searched for this study: PubMed (https://pubmed.ncbi.nlm.nih.gov/), Embase (https://www.embase.com/), Web of Science Core Collection (https://www.webofscience.com/), Scopus (https://www.scopus.com/) and the Cochrane Library (https://www.cochranelibrary.com/). The searches covered online English-language articles from database inception to July 31, 2025, respectively. The search strategies and keywords used for each database are listed in Supplementary Table 3. In addition to conducting an exhaustive search of relevant databases, we reviewed the references of pertinent publications and systematic reviews published within the past five years.

The evaluation was conducted using Rayyan (http://rayyan.qcri.org/), Microsoft Excel (Microsoft Corporation, Redmond, WA, USA), and Zotero software (Corporation for Digital Scholarship, Fairfax, VA, USA). Two evaluators (RL and YD) independently assessed the titles and abstracts of all articles based on the inclusion criteria. If the abstracts appeared potentially relevant, the full texts were reviewed in detail. A third reviewer (CY) was involved in the deliberation and unanimous determination of any conflict or disagreement between reviewers. All the evaluators were professional investigators trained in systematic literature research.

Inclusion and exclusion criteria

To establish the eligibility criteria for study inclusion, we adhered to the Patient, Intervention, Comparison, Outcome, and Study Design framework [13, 14]. The following inclusion criteria were identified and applied based on Population, Intervention, Comparator, Outcome, Study Design (PICOS) framework: (1) Patient: Patients diagnosed with ESCC through pathologically confirmed, with a minimum sample size of 10 patients. (2) Intervention: Presence, density, or maturity of TLS evaluated via histopathology. (3) Comparison: Investigations that compare the effectiveness of TNM staging or prognosis related to different TLS status within ESCC. (4) Outcome: For TNM staging, the reported outcomes should encompass odds ratios (OR) or detailed data delineating TNM stages at varying TLS thresholds. In the context of prognosis, studies are required to report at least one of the following critical outcomes: overall survival (OS), recurrence-free survival (RFS), early recurrence events, or late recurrence events. (5) Study design: English-language studies accessible in online databases. Research encompassing case–control studies, cohort studies, and clinical trials. Documentation of follow-up in cohort studies.

Exclusion criteria were as follows: (1) Studies limited to abstracts, case reports, or review articles; (2) Publications not written in English or those unavailable in online databases. (3) Research where Hazard ratios (HRs) or ORs cannot be obtained or estimated from the article. In addition, in instances where multiple studies involved overlapping patient cohorts, preference was given to the study that ranked highest based on a composite evaluation of methodological rigor, sample size adequacy, and recency of publication.

Outcome definition

The primary outcomes were defined as: (1) the association between TLS and TNM staging (T/N stage), measured by pooled ORs, and (2) the prognostic value of TLS for patient survival (OS/PFS/RFS/DFS), measured by pooled HRs. ORs for TNM staging were calculated based on dichotomized clinical classifications (e.g., T1-2 vs. T3-4; N0 vs. N +). Given the substantial clinical and methodological overlap between PFS, RFS, and DFS in ESCC studies—particularly their shared focus on disease progression events—we elected to harmonize these related endpoints under the PFS designation. For survival outcomes, multivariable-adjusted HRs were preferentially extracted from included studies to control for potential confounders. When adjusted HRs were unavailable, univariate HRs were extracted but subsequently analyzed separately in sensitivity analyses to assess potential confounding effects.

Data extraction

Structured forms were designed and used to retrieve relevant data from the selected studies. Two researchers (RL and YD) independently extracted the following information from the included studies: Author, Year of publication, country/region, study interval, Design, sample size, sex, Follow-up time, Laboratory, TLS assessment criteria, and diagnostic method details [15]. Inter-rater reliability was rigorously assessed using Cohen's κ coefficient, which measures agreement between raters beyond chance, with interpretive thresholds applied (κ = 0.40–0.60: fair agreement; κ = 0.60–0.75: good agreement; > 0.75: excellent agreement) [16]. Finally, an evaluator (CY) conducted a critical quality check.

Study quality and bias assessment

Three evaluators (CY, RL, and YD) independently assessed the methodological quality of the included studies. The included studies were all cohort studies, which were individually scored using the Newcastle–Ottawa Scale, ranging from 0–9 [17]. The scores were further categorized into three groups: 0–3, 4–6, and 7–9, which corresponded to low, moderate, and high-quality studies, respectively [17]. Any disagreement between the 3 evaluators (CY, RL, and YD) was resolved through discussion and consensus, and finally calibrated by the fourth evaluator (YG).

Data synthesis and statistical analysis

Meta-analysis statistics

The data extracted for the meta-analysis were analyzed using R software Version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria), with the significance threshold set at p < 0.05. Heterogeneity among studies was assessed using the I² statistic and Cochran’s Q value. The I² > 50% and p < 0.05 typically indicate substantial heterogeneity, and we adopted a random-effects model for our pooled analyses to account for variability beyond sampling error. Heterogeneity was assessed using the I² statistic, categorized as follows: 0%−40% (low), 30%−60% (moderate), 50%−90% (substantial), and > 75% (high) according to the Cochrane Handbook for Systematic Reviews (Version 6.4, 2023).

Subgroup analysis

To ascertain subgroup variations and potential sources of the observed heterogeneity, we conducted a series of prespecified subgroup analyses. These analyses included evaluating the relationship between sample size (> 200 versus ≤ 200), source of literature, quality assessment (moderate versus high), TLS criteria (High versus Low expression and Presence versus Absence), and laboratory methods (HE versus HE + IHC). Given the study's objective to provide a comprehensive outcome that reflects the effect of TLS presence on delaying disease progression by integrating progression-free survival (PFS), relapse-free survival (RFS), and disease-free survival (DFS), the subgroup analyses also evaluated the predictive impact of TLS on these distinct yet related survival indicators.

Statistical analysis

For statistical analyses, IBM SPSS Statistics (Version 26.0, IBM Corp., Armonk, NY, USA) was utilized to compute ORs and corresponding 95% CIs. Sensitivity analyses were conducted to examine the robustness of the pooled results by sequentially omitting one study at a time to observe the impact on overall findings. Potential publication bias was graphically assessed using funnel plots. Furthermore, Egger's regression test for asymmetry and Begg’s rank correlation test were performed to quantitatively evaluate any possible publication bias [18, 19].

Results

Study selection

The initial retrieval yielded 3971 potentially relevant studies. Following the removal of 378 duplicates, 197 records Marked as ineligible by automation tools, and one record removed for other reasons, an initial screening of titles and abstracts from the remaining 2901 studies resulted in the exclusion of 2838 studies that did not meet the eligibility criteria. The κ coefficient indicating inter-rater reliability was measured at 0.733. Full-text review of 63 studies led to further exclusion of 53 publications based on predefined criteria, achieving almost perfect agreement (κ = 0.710) among evaluators. Finally, ten studies were considered eligible for data extraction and included in this systematic evaluation and meta-analysis [20–26]. The details of the process are shown in Fig. 1.

Fig. 1.

Flowchart of literature search and study selection process

Study characteristics

Table 1 summarizes the characteristics of the ten studies included in this review [20–29]. These studies were sourced from five different databases and encompassed participants from a diverse set of countries, including the United States, Japan, and China, spanning the period from database inception to July 31, 2025. Each study adopted a cohort design, with sample sizes ranging from 87 to 650 subjects, cumulatively involving 1,772 participants. The results of the study quality assessment showed that six studies [21–23, 27–29] were rated as high quality and four [20, 24–26] were considered moderate. Overall, the risk of bias across the studies was assessed to be low to moderate. The primary methodologies employed for evaluating the presence and maturity of TLS were hematoxylin–eosin staining and immunohistochemistry.

Table 1.

Main characteristics of studies included in this meta-analysis

References	Country	Study interval	Design	Quality Assessment	Patients	Laboratory	Detection Method	Cut-off Criterion	Citation
Chen et al., (2024)	USA	2006–2023	Cohort study	High	87	HE	TLS density	Optimal threshold	[21]
Deguchi et al., (2022)	Japan	2011–2014	Cohort study	Medium	236	HE + IHC	TLS density (CD20 +)	Upper quartile	[20]
Hayashi et al., (2023)	Japan	2001–2017	Cohort study	High	316	HE	TLS count	Median	[22]
Li et al., (2022)	China	2017–2018	Cohort study	Medium	185	HE	TLS count	> 4 TLSs	[25]
Ling et al., (2022)	China	2008–2017	Cohort study	Medium	650	HE + IHC	TLS density (CD20 + CD21 + CD23 +)	Presence	[26]
Nakamura et al., (2023)	Japan	2008–2020	Cohort study	Medium	180	HE + IHC	TLS density (CD20 +)	Presence	[24]
Wang et al., (2023)	China	2013–2017	Cohort study	High	118	HE + IHC	TLS density (CD4 + CD8 + CD20 +)	Optimal threshold	[23]
Huang et al., (2025)	China	2009–2014	Cohort study	High	147	HE + IHC	TLS count (CD3 + CD20 + CD23 +)	≥ 3 TLSs	[27]
Huang et al. (2024)	China	2014–2021	Cohort study	High	359	HE + IHC	TLS count (CD3 + CD20 + CD23 +)	Presence	[28]
Jiang et al. (2025)	China	2008–2017	Cohort study	High	291	HE	TLS count (GC + structures)	≥ 3 TLSs	[29]

Laboratory methods: HE Hematoxylin & Eosin, IHC Immunohistochemistry

Study outcomes

Relationship between TLS and TNM stage

In the analysis of the relationship between TLS and TNM staging, our data indicates a potential positive association between TLS presence and both the T and N stages. Regarding the T stage, our data shows a statistically significant positive correlation between TLS presence and higher T stage (OR = 2.65, 95%CI: 1.86–3.78; p < 0.01, Fig. 2A). The observed relationship is accompanied by moderate heterogeneity, evidenced by an I² statistic of 58.2% and confirmed via Cochran's Q test (p = 0.02). However, sensitivity analysis employing the leave-one-out method revealed that excluding Deguchi’s study led to a marked reduction in heterogeneity (OR = 3.07, 95% CI = 2.27–4.15; p < 0.01; I² = 25%, Supplementary Fig. 1A). Additionally, no evidence of significant publication bias was observed through funnel plots along with Egger’s and Begg’s tests (p for Egger’s test = 0.60; p for Begg’s test = 0.71) as shown in Supplementary Fig. 1B.

Fig. 2.

Forest plots show the association between TLS and (A) T stage and (B) N stage in ESCC. OR, odds ratio; 95% CI, 95% confidence interval. TLS, tertiary lymphoid structures; ESCC, esophageal squamous cell carcinoma; Diamond markers represent pooled effect sizes

For the N stage, while the direction of the association also appears to be positive, it does not reach statistical significance (OR = 1.27, 95%CI: 0.85–1.89; p = 0.24, Fig. 2B). This association exhibits substantial heterogeneity, with an I² statistic of 68.2%, corroborated by Cochran's Q test (p < 0.01). Sensitivity analysis confirmed that the exclusion of any individual study did not significantly alter heterogeneity estimates (Supplementary Fig. 1C). Moreover, there was no indication of publication bias based on funnel plots and Egger’s and Begg’s tests (p for Egger’s test = 0.30; p for Begg’s test = 0.54, Supplementary Fig. 1D).

Relationship between TLS and prognosis

The analysis demonstrated a significant association between the presence of TLS and both extended OS (HR = 0.49, 95%CI: 0.41–0.58, p < 0.01, Fig. 3A) and improved PFS (HR = 0.56, 95%CI: 0.45–0.69, p < 0.01, Fig. 3B). The relationship with OS exhibited low heterogeneity (I² = 35%, Cochran's Q test p = 0.12), whereas the correlation with PFS showed moderate heterogeneity (I² = 51%, Cochran's Q test p = 0.02).

Fig. 3.

Forest plots demonstrate the prognostic impact of TLS on (A) overall survival and (B) progression-free survival in ESCC patients. HR, hazard ratio; 95% CI, 95% confidence interval; OS, overall survival; PFS, progression-free survival

Sensitivity analyses confirmed the robustness of these findings, as the exclusion of any single study did not substantially alter the results (Supplementary Fig. 1E, G for OS and PFS, respectively). Additionally, assessments for publication bias using funnel plots, Egger’s, and Begg’s tests revealed no significant bias for either outcome (Egger’s p = 0.26, Begg’s p = 0.47 for OS; Egger’s p = 0.28, Begg’s p = 0.42 for PFS, Supplementary Figs. 6 and 8).

Subgroup analyses

Subgroup analyses were performed to elucidate potential sources of heterogeneity in the association between TLS and patient prognosis. Regarding OS, the results maintained consistency across different countries and evaluation, yet significant variations emerged among subgroups stratified by patient characteristics (p < 0.01) and laboratory techniques (p = 0.02) (Fig. 4A and Supplementary Figs. 2). A more robust association between TLS and patient prognosis was evident in studies that had smaller sample sizes (HR = 0.33, 95%CI: 0.24–0.46), used "High vs. Low" as the dichotomization threshold for TLS (HR = 0.33, 95%CI: 0.22–0.48), and employed both HE staining and IHC for TLS identification (HR = 0.43, 95%CI: 0.34–0.54). In contrast, this relationship was less pronounced in studies with larger sample sizes (HR = 0.56, 95%CI: 0.46–0.68) or those that identified TLS using HE staining alone (HR = 0.57, 95%CI: 0.44–0.73).

Fig. 4.

Subgroup analysis of the association between TLS and (A) OS and (B) PFS, stratified by study characteristics. HE, hematoxylin and eosin staining; IHC, immunohistochemistry; HR, hazard ratio; 95% CI, 95% confidence interval

For PFS, subgroup analyses revealed significant heterogeneity by country (p = 0.03), sample size (p = 0.04), and laboratory methods (p = 0.03) (Fig. 4Band Supplementary Fig. 3), with stronger prognostic associations observed in smaller cohorts (< 200 patients: HR = 0.43, 95% CI: 0.32–0.58; I² = 0%), Japanese studies (HR = 0.43, 95% CI: 0.32–0.58; I² = 0%), and HE + IHC-based TLS detection (HR = 0.50, 95% CI: 0.41–0.60; I² = 0%). Weaker associations were noted in larger cohorts (> 200 patients: HR = 0.62, 95% CI: 0.53–0.72; I² = 63.5%), Chinese studies (HR = 0.62, 95% CI: 0.53–0.72; I² = 56.5%), and HE-only methods (HR = 0.68, 95% CI: 0.55–0.84; I² = 69.7%).

Discussion

TLS are organized aggregates of immune cells that form within non-lymphoid tissues because of chronic inflammation and are associated with immune cell infiltration in the tumor immune microenvironment [6]. TLS is generally considered to be linked with better prognosis and clinical outcomes following immunotherapy [5]. Considering that TNM staging is a critical standard for evaluating disease severity and predicting prognosis in cancer patients, this systematic review and meta-analysis was conducted to investigate the relationship between TLS and both clinicopathological features and prognosis in ESCC.

Pooled evidence from ten studies (n = 2569) indicates that TLS presence/increased density was typically associated with advanced T classification stages (meta-regression p < 0.01). Paradoxically, patients with TLS exhibited significant OS and PFS in our meta-analysis [pooled HR for OS: 0.49 (95% CI: 0.41–0.58); PFS: 0.56 (95% CI: 0.45–0.69)]. These survival benefits remained statistically significant after adjustment for confounding variables in multivariate Cox proportional hazards analyses. This apparent contradiction may be reconciled by considering the dynamic role of TLSs in tumor-immune interactions: although TLSs progressively accumulate during tumor evolution as organized immune aggregates in response to chronic inflammation [6], they also establish a pro-immunogenic niche capable of sustaining anti-tumor responses through enhanced antigen presentation, T/B cell activation, and coordinated immune cell recruitment [30–32]. The emergence of TLSs signifies a sustained immune response to malignancy by establishing a localized microenvironment that bolsters anti-tumor immunity [33, 34]. Consequently, TLS density reflects not merely tumor burden but rather the host's capacity to mount an effective local immune response—ultimately explaining why their presence independently predicts favorable prognosis, where robust immune engagement may counteract the typically poor outcomes associated with higher T-classification.

Furthermore, we investigated the impact of different research methodologies on prognosis. Our findings suggested a potential trend toward better prognostic performance with TLS quantification methods compared to binary classification based on TLS presence or absence, though this difference did not reach statistical significance. While a simple determination of whether TLSs are present may provide some level of prognostic insight, grouping patients according to high or low TLS expression yields more accurate predictive power. Additionally, our study found that combining HE staining with IHC for TLS diagnosis offers enhanced prognostic prediction. Indeed, a more comprehensive characterization of TLSs could provide deeper insights into their mechanisms of exerting antitumor effects and improving patient outcomes. As the investigation of TLSs advances, studies are increasingly focusing on the finer details of these structures, including their maturity, intra- and peritumoral distribution, density, cellular composition, and secretion of transcription factors [6, 35]. Emerging evidence suggests that TLSs characterized by intratumoral location, higher maturity, larger area, or greater density tend to correlate with improved prognosis. Conversely, peritumoral and immature TLSs have been associated with either no impact on prognosis or even adverse outcomes in certain cancers [36, 37]. Consequently, future investigations should delve further into elucidating the complex relationship between TLS characteristics and patient prognosis, which may potentiate antitumor immune responses and thereby optimize clinical outcomes.

Although our study provides valuable insights, several limitations must be acknowledged. First, the limited number of included studies (n = 10) reduces the statistical power and may limit the generalizability of results. Second, although we verified the robustness of the results through sensitivity analysis, the heterogeneity among studies may have affected the stability of the results. Third, variations in TLS quantification methods (e.g., digital vs. manual assessment) and cutoff definitions across studies may have introduced measurement bias. In addition, the causal relationship between TLS and TNM staging could not be determined because all studies were observational studies. Future studies are needed to further validate these findings through randomized controlled trials and prospective studies should investigate whether TLS-directed therapies could improve clinical outcomes in ESCC patients.

Conclusion

Our study demonstrated that the presence of TLS correlated with advanced T-stage disease yet independently predicted improved survival in ESCC. The histopathologically validated, quantitative TLS classification system (integrating H&E-based morphology with IHC confirmation) demonstrated optimal prognostic stratification.

Abbreviations

TLS

Tertiary lymphoid structure

ESCC

Esophageal squamous cell carcinoma

OS

Overall survival

PFS

Progression-free survival

OR

Objective response (or Odds ratio, if applicable)

HR

Hazard ratio

CI

Confidence interval

HE

Hematoxylin and eosin staining

IHC

Immunohistochemistry

Authors’ contributions

Chu-ting Yu: Literature screening, Data analysis, Manuscript writing (original draft), Critical revision. Ye Gao: Literature evaluation, Manuscript review & editing. Ru-yue Liu: Literature screening, Manuscript writing (original draft). Yu-ang Ding: Literature screening. Luo-wei Wang: Literature evaluation, Manuscript review & editing, Supervision.

Funding

This study was supported by the National Natural Science Foundation of China (Grant No. 82370677) and the Shanghai Municipal Education Commission—Shanghai Education Development Foundation "Chenguang Program".

Data availability

Data sharing is not applicable to this article as no datasets were generated or analysed during the current study. The datasets analyzed in this systematic review and meta-analysis were derived from published studies, which are fully cited in the References section. For complete methodological transparency, the original data extraction sheets (including quality assessment records and raw effect size calculations) may be made available by the corresponding author upon reasonable request, contingent upon compliance with any copyright or licensing restrictions applicable to the original studies.

Declarations

Ethics approval and consent to participate

Not applicable.

Competing interests

The authors declare no competing interests.