Risk factors for metachronous esophageal squamous cell carcinoma after endoscopic submucosal dissection: a multicenter cohort study integrating overall risk prediction and timing of recurrence.

BACKGROUND

Endoscopic submucosal dissection (ESD) has become the standard of care for superficial esophageal squamous cell carcinoma (ESCC). Nevertheless, the persistence of field cancerization leaves patients vulnerable to metachronous lesions, underscoring an urgent need for reliable post-ESD surveillance strategies. We reasoned that a refined risk stratification approach might optimize personalized follow-up.

AIM: This study therefore aimed to identify independent predictors of metachronous ESCC after curative ESD and, secondarily, to explore factors distinguishing early from late recurrence among those who relapse.

METHODS

We conducted a retrospective multicenter cohort study involving 1,646 patients with curative ESD for superficial ESCC across two Chinese tertiary centers between 2014 and 2024. For the primary endpoint, we employed Cox proportional hazards regression to identify variables associated with overall metachronous recurrence. In a separate exploratory analysis restricted to the 192 patients who actually developed recurrence, we applied logistic regression to examine determinants of early recurrence (defined as ≤36.7 months, the median time to recurrence) vs. late recurrence (>36.7 months).

RESULTS

During a median follow-up of 68.6 months, 192 patients (11.7%) experienced metachronous ESCC. In the primary Cox model based on a parsimonious set of 8 variables selected via LASSO regression, the strongest predictor was high-risk Lugol pattern (type C/D vs. A/B: HR = 18.868, 95% CI: 12.494-28.495), followed by elevated white blood cell count (per 1 × 109/L: HR = 1.487, 95% CI: 1.369-1.615), macrocytosis (MCV ≥106 fL, HR = 1.757, 95% CI: 1.144-2.700), and tumor size (per mm: HR = 1.021, 95% CI: 1.012-1.031). RBC (HR = 0.344, 95% CI: 0.241-0.492) and BMI (HR = 0.942, 95% CI: 0.903-0.982) were protective. The parsimonious model demonstrated excellent discrimination (Harrell's C-index = 0.924, 95% CI: 0.907-0.940), with an events-per-variable ratio of 24, confirming robustness against overfitting.

CONCLUSION

Our findings indicate that endoscopic features-particularly Lugol pattern-dominate the prediction of overall metachronous recurrence risk. The exploratory early vs. late recurrence analysis suggests that systemic inflammatory (WBC) and nutritional/metabolic (macrocytosis) parameters may modulate recurrence timing, but these findings are hypothesis-generating and require external validation before clinical application.