Extent of lymphadenectomy in thoracic esophageal squamous cell carcinoma: a meta-analysis of three-field versus two-field dissection.

Background
Esophageal squamous cell carcinoma (ESCC) remains one of the most lethal malignancies globally, with a 5-year overall survival (OS) rate that seldom exceeds 20% in most Western cohorts, though slightly higher in high-volume Asian centers due to earlier detection and more aggressive surgical strategies [1]. Surgery remains a cornerstone of curative treatment for thoracic ESCC, with lymphadenectomy playing a crucial role not only in locoregional control but also in accurate staging, guiding decisions on adjuvant therapy, and prognostication [2].
Two-field lymphadenectomy (2FL)—comprising mediastinal and abdominal node dissection—has long been the standard in many Western institutions [3]. However, three-field lymphadenectomy (3FL), which includes additional dissection of cervical lymph nodes, has gained traction particularly in East Asian surgical practice [4]. Proponents of 3FL argue that it offers superior nodal clearance and may improve survival, particularly in patients with subclinical cervical nodal metastases. Several retrospective studies and institutional experiences have reported improved OS and disease-free survival (DFS) with 3FL [5, 6], especially in patients with upper or middle third ESCC and positive nodes in the supracarinal or cervical regions.
Nonetheless, the broader adoption of 3FL remains controversial. Critics highlight its association with increased perioperative morbidity, particularly recurrent laryngeal nerve injury, pulmonary complications, and prolonged hospitalization. Moreover, the reported survival benefit has not been consistent across studies [7–10], and the absence of well-powered randomized controlled trials limits the generalizability of current evidence. Recent meta-analyses have also shown conflicting results, with some suggesting a modest survival advantage [11, 12], while others question its clinical significance in light of increased complications [13]. Compounding this controversy is the evolving role of neoadjuvant therapy, which may alter the biological burden of nodal disease and mitigate the need for extended surgical dissection.
Despite decades of clinical use and multiple comparative studies, no consensus has been reached on whether three-field lymphadenectomy provides meaningful oncologic benefit over two-field lymphadenectomy in thoracic ESCC. Previous systematic reviews [11, 13, 14] have been limited by small sample sizes, lack of stratified outcome reporting, or failure to comprehensively evaluate both short- and long-term outcomes, particularly in relation to postoperative complications and quality of life.
To address this persistent uncertainty, we conducted a comprehensive systematic review and meta-analysis comparing three-field and two-field lymphadenectomy in patients with thoracic ESCC. Our aim was to evaluate differences in oncologic efficacy—including overall survival, disease-free survival, and recurrence—alongside surgical metrics, nodal yield, and postoperative complications and mortality, to inform evidence-based clinical decision-making and guide future research.

Materials and methods

Study design
This systematic review and meta-analysis was conducted to compare the outcomes of 3FL versus 2FL lymphadenectomy in patients undergoing surgical resection for thoracic ESCC. The study adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines [15]. The review was registered on PROSPERO (CRD420251146571).

Search strategy
A comprehensive literature search was conducted using three electronic databases: PubMed, Scopus, and Web of Science. The search spanned all available years up to June 15th, 2025. The following keywords and MeSH terms were used in various combinations: (“esophageal cancer” OR “esophageal squamous cell carcinoma” OR “thoracic esophageal cancer”) AND (“3FL lymphadenectomy” OR “three-field dissection” OR “2FL lymphadenectomy” OR “two-field dissection” OR “lymph node dissection”). The full search query, adjusted per searched database, can be found in Table S1. Search results were exported to EndNote for duplicate removal. Additionally, a manual search of Google Scholar and reference lists of included studies was conducted to identify any further eligible articles [16].

Eligibility criteria
Studies were included if they met the following criteria:

Comparative studies evaluating 3FL versus 2FL lymphadenectomy in patients with thoracic ESCC.

Reported on at least one relevant oncologic or perioperative outcome stratified by surgical technique.

Randomized controlled trials (RCTs), prospective cohorts, and retrospective comparative studies.

Exclusion criteria included:

Studies without stratification of outcomes per lymphadenectomy field method.

Studies involving non-thoracic esophageal cancer (e.g., cervical or gastroesophageal junction).

Case reports, reviews, editorials, and conference abstracts.

Studies not reporting relevant outcomes or reporting irrelevant endpoints (e.g., gastrointestinal dysfunction unrelated to surgical technique).

Duplicate datasets published across multiple reports.

Study selection
Two independent reviewers screened titles and abstracts, followed by full-text screening of potentially eligible articles. Disagreements were resolved through discussion or by consultation with a third reviewer. The selection process is summarized in the PRISMA flow diagram (Fig. 1).

Data extraction
Data were extracted independently and in duplicate using a standardized data collection form. Extracted variables included study characteristics (author, year, country, study design), patient demographics (sample size, age, sex), and surgical details (lymphadenectomy technique, approach). Data on surgical approach (open, minimally-invasive “MI”), chemotherapy (neoadjuvant – NAC; adjuvant – AC), and radiotherapy (preoperative or postoperative) were also collected. To account for the evolving surgical trend in such cases, studies were further stratified based on time period (< 2010 vs. ≥2010).
Outcomes were categorized into primary and secondary. Primary outcomes included OS, DFS, recurrence, residual tumor status (R0/R1/R2), postoperative nodal/metastatic stage (N+/M+, stage III/IV). Secondary outcomes included the number of dissected lymph nodes, operative time, blood loss, length of hospital stay, reoperation rate, postoperative complications, and mortality (30-day, 90-day, in-hospital). Complications were analyzed as individual, study-reported outcomes rather than retrospectively reclassified using the Clavien–Dindo system, as patient-level data were unavailable and post hoc grading could result in misclassification or double counting of events. When multiple timepoints were reported (e.g., 1-, 3-, 5-year OS/DFS), each was extracted separately. When necessary, authors were contacted for additional data.

Risk of bias assessment
The risk of bias for RCTs was assessed using the Cochrane Risk of Bias 2 (RoB 2) tool [17], where a trial was assessed over 5 domain and given an overall rating of high risk, low risk, or of some concerns. For non-randomized studies, the National Institute of Health (NIH) tool was used, where a study was given a rating of good (> 20 score), fair (15–20 score), or poor (< 15 score) [18]. Quality assessments were performed independently by two reviewers, with discrepancies resolved by consensus.

Statistical analysis
Data synthesis was performed using STATA (version 18). For dichotomous outcomes, odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. For continuous outcomes, mean differences (MDs) with 95% CIs were pooled using inverse variance methods. Where necessary, medians and interquartile ranges were converted to means and standard deviations using established formulas.
Heterogeneity was assessed using the I² statistic and Cochran’s Q test. An I² > 50% indicated substantial heterogeneity, in which case a random-effects model (DerSimonian-Laird) was applied; otherwise, a fixed-effects model was used. When substantial heterogeneity was present, we performed leave-one-out (LOO) influence analyses as a diagnostic to assess whether any single study disproportionately affected the pooled estimate. These LOO results were treated as exploratory; the primary inference is based on the all-studies random-effects model. Exploratory subgroup analyses (based on time period, surgical approach, chemotherapy, and radiotherapy) were performed when the outcome was reported by at least five studies. timepoints was done for survival measures and mortality. Publication bias was not eligible in all analyses given the small sample size included (< 10 studies). A two-sided p-value < 0.05 was considered statistically significant.

Results

Study selection
A total of 138 records were identified through database searches (Fig. 1): 43 from PubMed, 68 from Scopus, and 27 from Web of Science. After removal of 62 duplicate records via EndNote software, 76 unique records remained for title and abstract screening. Of these, 35 were excluded for not meeting the eligibility criteria. Full texts were retrieved for 41 reports, with one report not retrievable. Among the 40 full-text articles assessed for eligibility, 13 were excluded for the following reasons: lack of outcome stratification by lymphadenectomy field method (n = 7), duplicate reporting (n = 3), non-thoracic ESCC population (n = 1), or reporting irrelevant outcomes (gastrointestinal dysfunction) (n = 1). Ultimately, 27 studies (from 28 reports) were included in the qualitative and quantitative synthesis [7–10, 19–42]. A supplementary manual search through Google Scholar and references of included studies yielded no additional eligible records.

Study characteristics
The summary of the baseline characteristics of included studies is provided in Table 1. Overall, 23 non-randomized studies (85.19%) were included, while only 4 RCTs were found (14.81%). Most evidence came from China (15 studies, 55.56%), followed by Japan (11 studies; 40.74%), and South Korea (1 study, 3.70%). A total of 10,039 patients were enrolled, of whom 6504 were in the 2FL lymphadenectomy group and 3389 cases in the 3FL group. Patients’ age and gender is provided in Table 1. Details on surgical approach, and perioperative therapy are presented in Table S2. In summary, NAC was reported by 23 studies, of which 16 had no NAC administration, two had all patients administered NAC, and five studies using NAC selectively. AC was reported by 16 studies, of which only one study had all patients receive AC while 11 studies used AC selectively. Preoperative radiotherapy was reported by 21 studies, of which 18 had no radiotherapy, two reported selective cases, and one study having all patients receive radiotherapy. As for postoperative radiotherapy, it was reported by 13 studies, of which eight studies reported selective cases and the remaining five did not have radiotherapy. As for the surgical approach, 14 studies used open approach, three used MI approaches, and three had mixed cases. Thirteen studies were done before 2010 and 12 studies were done after 2010.

Methodological quality
Out of the 4 included RCTs (5 reports), two had some concerns, one had high risk, and one had low risk of bias (Figure S1). Out of the 23 non-randomized studies, most studies had fair quality (16 studies, 69.57%), while the remaining 7 studies (30.43%) had good overall quality (Table S3).

Primary outcomes

Overall survival
A time-wise comparison showed no differences between 3FL and 2FL lymphadenectomy in regards to OS rate across all examined timepoints from 1 year (14 studies; OR = 0.98; 95% CI: 0.89, 1.07) to 8 years (2 studies; OR = 1.34; 95% CI: 0.70, 2.57) (Figure S2). All timepoints showed consistent results with no to minimal (insignificant) heterogeneity (Table 2).

Disease-free survival
A time-wise comparison showed no differences between 3FL and 2FL lymphadenectomy in regards to DFS rate across all examined timepoints from 1 year (14 studies; OR = 1.00; 95% CI: 0.84, 1.18) to 8 years (2 studies; OR = 1.59; 95% CI: 0.77, 3.30) (Figure S3). All timepoints showed consistent results with no to minimal (insignificant) heterogeneity (Table 2).
In terms of overall recurrence, no differences were noticed between groups (OR = 1.09; 95% CI: 0.89, 1.33; I2 = 0%, p = 0.78). The same was observed for locoregional recurrence (4 studies; OR = 1.02; 95% CI: 0.80, 1.29), systemic recurrence (2 studies; OR = 0.95; 95% CI: 0.28, 3.22), distant recurrence (4 studies; OR = 0.97, 95% CI: 0.73, 1.28), and cervical recurrence (5 studies; OR = 1.30; 95% CI: 0.15, 11.58).

Residual tumor status
No differences were seen in R0 (four studies, OR = 1.30; 95% CI: 0.58, 2.88; I2 = 0%, p = 0.67) between 3FL and 2FL lymphadenectomy (Figure S4). The same was observed with R1 (four studies; OR = 0.80, 95% CI: 0.35, 1.83; I2 = 0%, p = 0.65) and R2 (two studies; OR = 0.71; 95% CI: 0.07, 6.86; I2 = 0%, p = 0.69) status.

Postoperative nodal involvement
The difference in N + was marginally significant between 3FL and 2FL lymphadenectomy (seven studies; OR = 1.20; 95% CI: 0.99, 1.47; I2 = 13.81%, p = 0.54) (Figure S5). The same was observed with postoperative pathological stage T3 or T4 (four studies; OR = 0.98; 95% CI: 0.74, 1.30; I2 = 21.70%, p = 0.38) (Figure S6).

Secondary outcomes

Lymphadenectomy metrics
Three-field lymphadenectomy was associated with a greater number of dissected LNs compared to 2FL lymphadenectomy (ten studies; MD = 15.01; 95% CI: 7.74–22.28) (Fig. 2). Heterogeneity was significant (I2 = 99.19%, p < 0.001); however, no change was observed with the sensitivity analysis (Figure S7). That said, the funnel plot showed asymmetry (Figure S8) with significant risk of publication bias (Egger’s test p-value = 0.0274).

Surgical efficiency
No difference in operative time was observed between 3FL and 2FL lymphadenectomy (seven studies; MD = 13.74 min; 95% CI: -3.06, 30.53). Heterogeneity was high (I2 = 96.75%< p < 0.001); however, the sensitivity analysis showed a longer operative time with the 3FL technique after excluding the study of Song et al. [39] (MD = 20.65, 95% CI: 9.32, 31.99) (Figure S9).
No significant difference in blood loss was observed between both groups (five studies; MD = 127.99; 95% CI: -74.96, 330.93). Heterogeneity was high (I2 = 99.75%, p < 0.001); however, the sensitivity analysis showed that the 3FL approach was associated with a greater blood loss compared to the 2FL technique after excluding the study of Yamashita et al. [42] (MD = 31.83; 95% CI: 2.83, 60.84) (Figure S10).

Hospital course
No significant difference in length of hospital stay was observed between both groups (six studies; MD = -5.44; 95% CI: -20.43, 9.55). Heterogeneity was high (I2 = 99.84%, p < 0.001); however, the sensitivity analysis showed a significantly longer hospital stay in the 3FL approach after excluding the study of Yamashita et al. [42] (MD = 0.65; 95% CI: 0.20, 1.09) (Figure S11).
No differences in reoperation rate were also observed (four studies; OR = 1.69; 95% CI: 0.48, 5.97). Heterogeneity was high (I2 = 69.80%, p = 0.03); however, the sensitivity analysis showed a greater reoperation risk in the 3FL technique after excluding the study of Yamashita et al. [42] (OR = 2.88; 95% CI: 1.36, 6.12) (Figure S12).

Postoperative complications

Complications
No differences in the overall complications were observed between both groups (nine studies; OR = 1.85; 95% CI: 0.98, 3.47, I2 = 91.61%, p < 0.001). No differences were observed in most complications (Table S4); however, the 3FL approach was associated with a significantly greater risk of pulmonary complications (nine studies; OR = 1.67; 95% CI: 1.06, 2.63, I2 = 59.90%, p < 0.001) and recurrent laryngeal nerve palsy (10 studies; OR = 1.69; 95% CI: 1.06, 2.69; I2 = 60.06%, p = 0.01) compared to the 2FL technique.

Mortality
Overall, no significant differences in the risk of mortality were observed between both procedures (11 studies; OR = 0.79; 95% CI: 0.51, 1.21; I2 = 5.87%, p = 0.80). This was observed across 30- and 90-day mortality as well. However, the 3FL approach was associated with a significantly lower risk of in-hospital death compared to the 2FL technique (two studies; OR = 0.35; 95% CI: 0.13, 0.93, I2 = 0%, p = 0.51) (Figure S13).

Subgroup analysis
The findings of the predefined subgroup analyses are presented in Figures S14–S23 and summarized in Table 3. Treatment era (≥ 2010), surgical approach, and multimodal therapy exposure were associated with differences in selected perioperative parameters and lymph node yield. Specifically, three-field lymphadenectomy was associated with longer operative time in open surgery (MD = 14.54; 95% CI: 5.55–23.53), and with a higher number of dissected lymph nodes in contemporary cohorts (≥ 2010; MD = 12.97; 95% CI: 4.28–21.66), regardless of surgical approach. Neoadjuvant chemotherapy, adjuvant chemotherapy, and preoperative radiotherapy were each associated with increased lymph node yield in the three-field subgroup.

Importantly, none of the investigated factors—including treatment era, surgical technique, or multimodal therapy—consistently moderated nodal positivity, postoperative mortality, overall complication rates, disease-free survival, or recurrence.

Discussion

Principal findings
This systematic review and meta-analysis, comprising 27 studies and over 10,000 patients, represents one of the most comprehensive evaluations to date of the oncologic and surgical outcomes associated with 3FL versus 2FL in thoracic ESCC. Across all examined timepoints, 3FL did not improve overall or disease-free survival compared with 2FL. Although 3FL achieved a higher lymph-node yield (MD ≈ 15 nodes), this endpoint showed very high between-study heterogeneity and evidence of small-study effects/publication bias. Nodal positivity (N+) was only marginally higher with 3FL and did not reach statistical significance (OR 1.20, 95% CI 0.99–1.47). Conversely, 3FL was associated with higher risks of recurrent laryngeal nerve (RLN) palsy and pulmonary complications, while overall (composite) complications were not significantly different between approaches. Unlike earlier meta-analyses suggesting a survival advantage for three-field lymphadenectomy, our expanded and ESCC-specific dataset demonstrates that increased nodal yield does not translate into durable survival benefit when examined across multiple long-term timepoints (Table 4).

Subgroup analyses and interpretation
Over the ~ 30-year span of the included literature, esophageal cancer care has shifted toward minimally invasive/robotic esophagectomy, standardized enhanced-recovery pathways, and expanded use of neoadjuvant chemoradiotherapy and, more recently, immunotherapy. Our subgroup analyses explored the potential modifying effects of study period, surgical approach, and perioperative therapy on major outcomes. These subgroup findings indicate that while technical evolution and multimodal therapy influence lymph node yield and selected perioperative metrics, they do not alter the absence of a survival or recurrence benefit associated with three-field lymphadenectomy. Additionally, none of the examined factors significantly moderated nodal positivity, recurrence, disease-free survival, or mortality.
It is important to note that these subgroup findings should be interpreted cautiously. Several analyses were driven by one or two studies only, limiting their statistical reliability. Furthermore, the classification of studies by perioperative therapy followed pragmatic rules based on available reporting: “not reported,” “none” (0% of the cohort received the therapy), or “yes” (100% of patients received it). The latter category applied to only three studies, further constraining interpretability. Hence, while the subgroup findings generate valuable hypotheses, they cannot establish causality or confirm true interactions between surgical extent and multimodal therapy exposure.

Oncologic effectiveness: survival and nodal clearance
Contrary to earlier meta-analyses which suggested that 3FL conferred superior OS and DFS rates [12], our pooled estimates at 1–8 years show no survival advantage for 3FL. This discrepancy may be partially explained by our stricter inclusion of studies with outcome stratification by surgical approach and histology. Moreover, some of the previously observed survival advantage may reflect stage migration rather than a true therapeutic benefit.
Indeed, 3FL was associated with increased lymph node retrieval—an observation that aligns with prior findings demonstrating its capacity to improve staging accuracy and locoregional control [44]. However, the magnitude and certainty of this effect are tempered by very high heterogeneity (I² ~99%) and funnel-plot asymmetry suggestive of publication bias. Consistent with this, the higher nodal yield did not translate into survival gains, and N + differences were marginal and non-significant [11, 14, 45]. These patterns underscore the complexity of lymphatic dissemination in ESCC and challenge the notion that anatomical extent alone can dictate oncologic success.
Interestingly, while some earlier studies reported a significant survival benefit of 3FL in node-positive patients [11, 46], this subgroup effect was not clearly replicated in our meta-analysis. Thus, the role of 3FL as a therapeutic rather than merely a diagnostic maneuver remains uncertain. In the modern context—where neoadjuvant therapy can alter nodal burden and tumor biology—the incremental therapeutic return of wider lymphadenectomy may be further attenuated [45].

Evolving role of neoadjuvant therapy, immunotherapy, and sentinel-node mapping
The therapeutic landscape of ESCC has shifted markedly over the past decade, with neoadjuvant therapy becoming the standard of care in locally advanced disease. The landmark CROSS trial [47] demonstrated that preoperative chemoradiotherapy improved both R0 resection rates and overall survival compared with surgery alone. Similar benefits were corroborated by the NEOCRTEC5010 trial [48] in an exclusively squamous-cell population. These multimodal regimens achieve significant tumor downstaging and nodal sterilization, thereby reducing the apparent yield of positive nodes and potentially diminishing the incremental value of extended lymphadenectomy.
More recently, the integration of immunotherapy into perioperative regimens—notably nivolumab in the CheckMate 577 trial [49] and pembrolizumab in KEYNOTE-590 [50]—has transformed postoperative management. Adjuvant nivolumab after neoadjuvant chemoradiotherapy and R0 resection prolonged disease-free survival, suggesting that systemic immune modulation may now achieve the micrometastatic control once sought through more extensive surgical clearance. As immune checkpoint inhibitors, targeted agents, and combination regimens become entrenched in multimodal paradigms, the relative benefit of three-field dissection may continue to narrow, especially given its higher morbidity profile.
Parallel advances in sentinel-node navigation surgery are redefining lymphatic mapping in esophageal cancer. Prospective Japanese trials such as JCOG0303 and subsequent validation cohorts [51] have demonstrated the feasibility and accuracy of radio- or dye-guided sentinel-node detection in early-stage ESCC. When combined with minimally invasive or robotic techniques, this approach allows selective lymphadenectomy with preservation of vital structures such as the recurrent laryngeal nerves. In the future, sentinel-node–guided strategies may permit personalization of the dissection field, limiting three-field procedures to patients with proven or highly suspected cervical drainage involvement.

Postoperative complications and mortality
The increased risk of certain complications associated with 3FL is well-established and was reaffirmed by our analysis. The technique’s inclusion of cervical lymph node dissection places patients at elevated risk for RLN injury and pulmonary complications. Our findings echo those of previous studies, including the meta-analyses by Ma and Wang, which reported significantly higher rates of RLN palsy and anastomotic leak following 3FL compared to 2FL [11, 13]. In the present synthesis, RLN palsy and pulmonary complications were significantly higher with 3FL, whereas the composite “overall complications” endpoint was not significantly different between groups. Although standardized grading systems such as Clavien–Dindo are valuable, their retrospective application in meta-analyses without patient-level data may obscure true morbidity patterns; therefore, complication-specific analyses may offer greater transparency in this context.
Importantly, overall 30- and 90-day mortality did not differ between approaches, while in-hospital mortality was lower with 3FL based on two studies (OR 0.35, 95% CI 0.13–0.93); given the small evidence base and potential for context-specific practice effects, this signal should be interpreted cautiously. This paradoxical finding may reflect improved postoperative vigilance and centralized surgical care in centers that routinely perform 3FL. Alternatively, it may be a spurious finding given the small number of studies contributing to this analysis.
It is also notable that complications associated with 3FL can impair functional recovery and quality of life, a concern increasingly relevant in the era of multimodal therapy and survivorship focus. The literature emphasizes that the morbidity burden of extensive lymphadenectomy must be carefully weighed against its oncologic yield, especially in older or medically frail patients [44, 46].

Patient selection and the case for individualization
Our findings, combined with the heterogeneity observed in prior studies, underscore the need for a personalized approach to lymphadenectomy in ESCC. Several large-scale retrospective and autopsy studies have confirmed that the distribution of nodal metastases in ESCC is influenced by tumor location, histologic type, and T-stage [14, 45]. For instance, cervical lymph node involvement is more common in upper and middle thoracic tumors than in lower-third lesions. Hence, the application of 3FL may be more appropriate in proximal tumors or when preoperative imaging suggests cervical or supracarinal node involvement [11, 46].
Moreover, the absence of a clear survival benefit for 3FL across the entire population suggests that a one-size-fits-all approach may no longer be tenable. Rather, selective application of 3FL, potentially guided by advanced imaging modalities (e.g., PET-CT, EBUS) or intraoperative frozen section assessment of sentinel nodes, may optimize the therapeutic index of this more extensive procedure. Future prospective work should also prespecify standardized definitions for nodal endpoints and complication grading to reduce methodological heterogeneity.

Strengths and limitations
The strengths of this review lie in its large sample size, inclusion of both RCTs and high-quality observational studies, and extensive sensitivity analyses. However, several limitations merit discussion. First, heterogeneity in surgical technique, perioperative care, and outcome definitions was unavoidable. Second, although we harmonized terminology (e.g., “dissected” vs. “harvested” nodes; “morbidity” vs. “overall complications”), residual inconsistencies across primary reports may persist. Third, very high heterogeneity and small-study effects in the lymph-node yield analysis limit the certainty around the magnitude of this difference. Fourth, despite rigorous efforts, residual confounding and selection bias are likely, given that most included studies were retrospective; accordingly, the overall strength of evidence is limited and pooled estimates should be interpreted with caution. Fifth, leave-one-out findings that appear to change significance (e.g., operative time, blood loss, length of stay, reoperation) reflect influence diagnostics under high heterogeneity and do not supplant the all-studies estimates. Importantly, the evidence spans multiple countries and three decades, during which operative techniques, perioperative care, and systemic therapies evolved. Reporting of these factors was incomplete and inconsistent, precluding formal subgroup/meta-regression; therefore, residual confounding by era, approach, and multimodal therapy is likely despite low statistical heterogeneity.

Conclusion
In conclusion, three-field lymphadenectomy increases lymph-node yield but does not confer a consistent survival advantage over two-field dissection in thoracic ESCC. The approach carries higher risks of specific complications (RLN palsy and pulmonary events) while overall complication rates were not significantly different, and perioperative mortality was similar between groups (with a cautiously interpreted signal of lower in-hospital mortality from two studies). Given the predominance of retrospective evidence, very high heterogeneity in nodal yield, and evolving multimodal practice, these findings should be interpreted conservatively. Prospective, high-quality randomized studies in contemporary practice settings are needed to confirm these results. Future research should stratify by tumor location, baseline nodal status, and neoadjuvant therapy; standardize definitions of nodal endpoints and complications; and evaluate the role of minimally invasive/robotic techniques and image- or molecular-guided nodal assessment to refine patient selection for 3FL.

Supplementary Information