Esophageal cancer (EC) ranks as the eighth most prevalent malignancy and the sixth leading cause of cancer-related mortality worldwide. It exhibits significant geographic heterogeneity, with East Asia—particularly China—bearing the highest burden of esophageal squamous cell carcinoma (ESCC) (1–4). In China, ESCC accounts for over half of global cases, characterized by late-stage diagnoses and poor prognoses, making it the seventh most prevalent and fifth most lethal cancer nationally (1–4). The survival of patients with esophageal cancer is influenced by numerous factors, including surgical details such as lymph node resection, which are crucial for determining long-term prognosis (5–9). Despite advancements in multimodal therapies that combine surgery, radiotherapy, chemotherapy, and immunotherapy, long-term survival for patients with ESCC remains suboptimal. This underscores the critical role of esophagectomy as the cornerstone of curative intent (9–13).

The surgical management of ESCC necessitates esophageal resection followed by reconstruction via anastomosis, with intrathoracic (IA) and cervical (CA) techniques representing the primary approaches. The choice between these methods remains contentious, as each carries distinct trade-offs. Proponents of IA emphasize its technical feasibility, robust vascular supply, and reduced anastomotic leakage risk, whereas CA advocates highlight its potential for wider resection margins and lower locoregional recurrence, possibly conferring survival advantages (14–19). However, conflicting evidence persists: recent studies associate IA with fewer complications like recurrent laryngeal nerve injury, while emerging propensity-adjusted analyses suggest CA may improve overall survival, even after confounder adjustment. These survival benefits, however, must be balanced against CA’s higher morbidity rates, including anastomotic stenosis and pneumonia (14–19).

This study evaluates long-term survival following IA versus CA in middle and lower thoracic ESCC through a retrospective analysis, focusing on survival evidence. Our findings aim to inform surgical decision-making.

This retrospective cohort study compared long-term survival outcomes between CA and IA in patients with middle and lower thoracic ESCC. The results demonstrated that CA was associated with superior OS and DFS compared to IA in the overall cohort. However, after PSM to adjust for baseline confounders, these survival advantages were attenuated and no longer statistically significant.

The observed OS and DFS benefits for CA in the pre-matched analysis align with prior studies suggesting that CA may facilitate wider resection margins and more extensive lymph node dissection, potentially reducing locoregional recurrence and improving oncologic outcomes (14–16). Our data revealed that the CA group underwent dissection of significantly more lymph node stations (>7) compared to IA, which may partially explain its initial survival advantage. Notably, pathological T3/T4 stages and lymphovascular invasion were independently associated with poorer OS and DFS. These findings emphasize that advanced tumor invasion and aggressive biological behavior outweigh the impact of anastomotic technique on survival.

The prognosis of ESCC is influenced by multiple, potentially overlapping factors (20–24). While the current standard of care prioritizes surgical resection following neoadjuvant therapy, surgical variables remain critical determinants of outcomes (25–28). Our study investigated the impact of anastomotic location on survival, with findings suggesting that the observed differences may fundamentally stem from variations in lymph node dissection. Notably, the CA group demonstrated a significantly higher proportion of lymph node stations resected (>7 stations) compared to the IA group—a finding consistent with prior studies (15, 17). Although PSM attenuated survival differences and CA showed no statistically significant survival advantage over IA in matched cohorts, a clinically relevant gap in OS and DFS persisted between the groups. This trend suggests that prolonged follow-up and larger patient cohorts may amplify these differences, potentially revealing clearer survival benefits for CA over time. The survival gap observed in the unmatched analysis aligns with the hypothesis that CA facilitates more extensive lymphadenectomy, thereby addressing occult micrometastases and reducing locoregional recurrence (15–19).

When contextualized with the meta-analysis by You et al., highlight a critical tension in surgical decision-making for ESCC: the balance between survival outcomes and postoperative morbidity. While our unmatched analysis initially suggested a survival advantage for CA, the attenuation of this benefit after PSM underscores the confounding influence of baseline factors such as lymph node dissection extent and tumor biology. However, IA’s lower complication rates—particularly for anastomotic leak (RR = 2.76 for CA in You et al.)—may mitigate its theoretical survival disadvantages by reducing delays in adjuvant therapy and preserving quality of life (29).

The integration of findings from our study similar to work by Li et al. provides a nuanced perspective on the relationship between anastomotic location, lymphadenectomy extent, and survival outcomes in ESCC. While both studies highlight the potential survival advantages of CA, the mechanisms underlying these benefits appear intrinsically tied to the extent of lymph node dissection, particularly in the upper mediastinal zone. Notably, Li et al. identified significantly more extensive dissection of recurrent laryngeal nerve (RLN)-associated LN stations (105, 106recL/R) in the CA group, which correlated with reduced upper mediastinal recurrence (7.1% vs. 15.7%, P<0.001) and LN recurrence (19.1% vs. 28.4%, P<0.001). These results suggest that CA facilitates superior en bloc resection of occult micrometastases in the RLN region, a common site for ESCC recurrence (15).

These collective findings highlight the critical interplay between surgical technique, anatomical access, and oncologic precision in ESCC management. The superior upper mediastinal lymph node dissection achieved through CA—particularly in high-risk zones like the RLN basins—likely disrupts metastatic pathways that IA approaches may incompletely address, given their technical constraints in visualizing and resecting supra-aortic arch structures. From an evolutionary surgical perspective, these results advocate for a paradigm shift from binary technique comparisons to quality-oriented metrics, particularly standardized lymphadenectomy templates incorporating bilateral RLN node dissection. Future trials should stratify by both anastomotic approach and achieved lymph node yield to disentangle their individual contributions while accounting for the learning curve effect, as CA’s technical demands may introduce performance bias in non-specialized centers.

It is important to explore whether IA combined with postoperative radiotherapy (PORT), specifically targeting the upper mediastinal nodes, could offer comparable local control to CA in patients who are unfit for extensive lymphadenectomy, such as elderly patients or those with poor pulmonary function. IA combined with PORT may provide an alternative approach to achieve better local control in patients who cannot undergo aggressive lymph node dissection due to contraindications such as compromised health or anatomical considerations. The use of PORT in such cases may help to target residual microscopic disease in the mediastinal region, potentially improving local control without the need for extensive lymphadenectomy. It is essential to consider the balance between potential benefits and the risks associated with radiation therapy in these vulnerable populations. Future studies could further clarify whether this combination approach offers comparable long-term outcomes in terms of local control, DFS, and OS, particularly for patients who are not candidates for comprehensive surgical interventions. By incorporating this consideration, personalized decision-making can be better guided, optimizing treatment strategies for patients based on their individual health status and disease characteristics.

This study has several limitations inherent to its retrospective design. First, unmeasured confounders, such as surgeon experience, precise tumor location, and adjuvant therapy adherence, may have influenced outcomes. Second, the single-center design and modest sample size post-PSM limit generalizability, potential bias from single-center surgical expertise homogeneity (e.g., whether CA/IA selection was influenced by surgeon experience) and its impact on generalizability. Third, the median follow-up of 54 months, while substantial, may not capture very late recurrences. Then, our study lacks a detailed classification of lymph nodes by individual stations. This limitation arises from the use of the NCCN guidelines for staging in the pathology reports, which typically categorize lymph node involvement into broader groups without distinguishing between individual stations. Finally, the absence of standardized criteria for selecting CA versus IA introduces residual selection bias, despite PSM adjustment.

This retrospective cohort study found middle and lower thoracic ESCC, CA was initially associated with improved OS and DFS compared to IA, but this advantage diminished after balancing baseline covariates.