This retrospective study enrolled patients who received neoadjuvant therapy followed by radical esophagectomy at our institution between June 2016 and December 2021. The inclusion criteria were as follows: (1) age between 18 and 75 years; (2) pathological confirmation of esophageal squamous cell carcinoma (ESCC); and (3) clinical stage II–IVA according to the American Joint Committee on Cancer (AJCC) 8th edition TNM staging system. Patients were excluded based on the following criteria: (1) concurrent other malignancies or ongoing anticancer therapy for other cancers; (2) postoperative pathology indicating mixed histological types; (3) non-R0 resection; (4) incomplete pre-treatment laboratory data; (5) presence of distant metastasis; or (6) lack of postoperative follow-up information. A total of 216 patients were included in the final analysis. The study protocol was approved by the Ethics Committee of Fujian Medical University Union Hospital (Approval No. 2025KY081) and was conducted in accordance with the principles of the Declaration of Helsinki and its subsequent amendments. Given the retrospective nature of the study, a waiver of informed consent was granted.

We retrospectively collected clinical data from patient cases within one week prior to neoadjuvant therapy initiation. The data encompassed baseline characteristics—including sex, age, body mass index (BMI), comorbidities, smoking history, and alcohol consumption—as well as hematological parameters such as hemoglobin, lactate dehydrogenase, neutrophil-to-lymphocyte ratio(NLR), serum albumin, total lymphocyte count, and total cholesterol. Tumor-related information was also recorded, covering tumor location, neoadjuvant treatment regimen, clinical T and N stages, postoperative pathological T and N stages, and tumor regression score.

GRIm was calculated using the following three variables: LDH (within the normal range: 0 vs. > Upper limit of normal (ULN) for each center, +1 for 245 U/L in our hospital), ALB (≥ 35 g/L: 0 vs. < 35 g/L: +1), and NLR (≤ p 75: 0 vs. > p 75: +1). Patients were divided into two groups: a high group (2 or 3 points) and a low group (0 or 1 point) (reference). The CONUT score consisted of LYM, TC, and ALB. Patients were divided into two groups based on their CONUT score: a low group (< 3 points) and a high group (≥ 3 points). Based on this, the study subjects were further divided into three groups: a GRIm-CONUT score of 2 (high GRIm and high CONUT), a GRIm-CONUT score of 1 (high GRIm and low CONUT or low GRIm and high CONUT), and a GRIm-CONUT score of 0 (low GRIm and low CONUT). Specific definitions are shown in Figure 1. A score of 0 indicates essentially normal nutritional and inflammatory status. A score of 1 reflects a mild, single-system abnormality, which is clinically common and generally manageable. A score of 2 signifies a more severe derangement.

Neoadjuvant regimens included: neoadjuvant chemotherapy (2–4 cycles of paclitaxel or albumin-bound paclitaxel combined with platinum-based regimens, every 3 weeks), neoadjuvant chemoradiotherapy (same chemotherapy regimen, with a radiotherapy dose of 40–50 Gy/23F), and neoadjuvant chemoimmunotherapy (same chemotherapy regimen, combined with immunotherapy including pembrolizumab, camrelizumab, or sintilimab). All patients underwent radical surgery within 4–6 weeks after the completion of neoadjuvant therapy, either thoracoscopic or laparoscopic combined with da Vinci robotic-assisted McKeown or Ivor Lewis esophagectomy.

All patients included in the study underwent systematic and regular follow-up. Follow-up primarily included survival status, disease recurrence and metastasis, and treatment status. Patients were followed up every 3 months for the first 2 years post-surgery, every 6 months from years 3 to 5, and annually thereafter. Monitoring methods included hematological markers, CT scans of the neck, chest, and abdomen, PET/CT, and endoscopy. The primary endpoint was overall survival(OS), defined as the period from the date of surgery to the last follow-up or the occurrence of the endpoint event (death). The secondary endpoint was recurrence-free survival (RFS), defined as the time from surgery to the first documented recurrence, distant metastasis, death from any cause, or the last confirmed recurrence-free follow-up.

This study used R 4.5.0 and SPSS 20 for data analysis. Continuous variables between groups were analyzed using t-tests or Mann-Whitney U tests, and dichotomous variables between groups were identified using chi-square tests or Fisher’s exact tests. OS and RFS were assessed using Kaplan-Müller’s method (KM) analysis, and curve differences were analyzed using log-rank tests. Univariate and multivariate Cox regression analysis was used to identify risk factors predicting OS and RFS; variables with p < 0.05 were included in multivariate regression analysis. Models were constructed based on the multivariate regression coefficients to predict OS. A nomogram model was constructed using the rms package in R for visualization, and the area under the ROC curve (AUC), time-dependent ROC curves, calibration curves, and decision curve analysis (DCA) were constructed using the pROC, riskRegression, rms, and dcurves packages to evaluate the performance of the predictive model. No external validation was designed; internal validation was performed using Bootstrap with 1000 iterations.

A total of 216 patients with locally advanced ESCC who received neoadjuvant therapy followed by curative surgery were included in the analysis. Based on pretreatment laboratory indices, patients were categorized into three GRIm–CONUT groups: score 0 (n = 118, 54.63%), score 1 (n = 79, 36.57%), and score 2 (n = 19, 8.80%).Baseline demographic and clinical characteristics are summarized in Table 1. Significant differences among the three GRIm–CONUT groups were observed in sex distribution (P = 0.015). In contrast, age, BMI, smoking status, alcohol consumption, and comorbidities—including hypertension, diabetes, and heart disease—did not differ significantly across groups (all P > 0.05).

Regarding tumor-related variables, there was a notable difference in clinical N stage (cN) across the GRIm–CONUT categories (P = 0.019). The distribution of clinical T stage (cT) and ypT stage demonstrated observable variation across groups, although ypT showed statistical significance (P = 0.023*). Similarly, ypN stage significantly differed among the groups (P = 0.022). Tumor location and the choice of neoadjuvant regimen (chemotherapy, chemoradiotherapy, or chemoimmunotherapy) showed no significant differences (P > 0.05).

Overall, patients with higher GRIm–CONUT scores tended to present with more advanced nodal disease, while other baseline clinical features were generally comparable among the groups.

Correlation analysis was performed to evaluate the associations among inflammatory markers, nutritional indicators, clinicopathologic variables, and survival outcomes. As shown in Figure 2, both GRIm and CONUT scores demonstrated stronger correlations with overall survival (OS) and recurrence-free survival (RFS) compared with other inflammation- or nutrition-based indices. Notably, although GRIm and CONUT share serum albumin as a common component, their correlation with each other was only moderate, indicating that they capture distinct and complementary aspects of systemic immune–nutritional status.

Importantly, the combined GRIm–CONUT score exhibited a stronger relationship with survival outcomes than either score alone, supporting its value as an integrated prognostic marker. Among clinicopathologic variables, ypN stage showed a significant correlation with both OS and RFS, whereas other tumor characteristics displayed weaker or non-significant associations.

The predictive performance of the GRIm–CONUT score and its individual components was evaluated using ROC curve analysis. As shown in Figure 3, the GRIm–CONUT composite demonstrated higher AUC values for predicting OS compared with either the GRIm or CONUT score alone.(GRIm–CONUT vs GRIm or CONUT at 3years:0.692 vs 0.639 or 0.595, P<0.05;GRIm–CONUT vs GRIm or CONUT at 5years:0.785 vs 0.667 or 0.629, P<0.05). These findings indicate that integrating immune-inflammatory and nutritional markers provides a more comprehensive assessment of patient prognosis than using either score in isolation. Moreover, the GRIm–CONUT score outperformed traditional inflammatory markers across multiple follow-up periods, further supporting its robustness as a prognostic tool in the neoadjuvant setting.

During a median follow-up of 36.5 months, significant survival differences were observed among the three GRIm–CONUT groups. As illustrated in Figure 4, patients with a GRIm–CONUT score of 0 had markedly better OS and RFS than those with scores of 1 or 2. Kaplan–Meier analysis demonstrated a clear stepwise decline in both OS and RFS with increasing GRIm–CONUT categories (P = 0.002 for OS; P < 0.001 for RFS, log-rank test).

The 5-year OS rates for scores 0, 1, and 2 were 71.19%, 63.29%, and 36.84%, respectively. Corresponding 5-year RFS rates were 67.80%, 60.76%, and 31.58%, indicating substantially higher risks of recurrence and mortality in patients with elevated GRIm–CONUT scores.

These findings support the prognostic relevance of the GRIm–CONUT composite, demonstrating that impaired immune–nutritional status before neoadjuvant therapy is strongly associated with inferior long-term outcomes following curative esophagectomy.

Univariate and multivariate logistic regression analyses were performed to identify independent predictors of pathologic complete response (pCR). In the univariate analysis, GRIm–CONUT score, cN stage, and neoadjuvant therapy regimen were significantly associated with pCR.

After adjustment for potential confounders, all three variables—GRIm–CONUT score, cN stage, and neoadjuvant regimen—remained independent predictors in the multivariate model (P < 0.05 for all). Patients with a GRIm–CONUT score of 0, those presenting with cN0 disease, and those receiving neoadjuvant chemoimmunotherapy demonstrated notably higher pCR rates (Table 2).

These findings indicate that pretreatment immune–nutritional status, nodal burden, and treatment modality independently influence the likelihood of achieving pCR following neoadjuvant therapy.

Univariate Cox regression analysis identified several variables significantly associated with OS, including GRIm–CONUT score, ypT stage, ypN stage, and pCR (all P < 0.05). Variables with P < 0.10 were subsequently included in the multivariate model.

In the multivariate analysis, both GRIm–CONUT score and ypN stage remained independent prognostic factors for OS. Patients with higher GRIm–CONUT scores had a significantly increased risk of mortality (P = 0.025), indicating that poorer immune–nutritional status before neoadjuvant therapy is strongly associated with inferior long-term survival. Similarly, advanced ypN stage independently predicted worse OS (P < 0.001).

These findings indicate that the GRIm–CONUT composite serves as a robust prognostic indicator for postoperative survival, providing additive prognostic information beyond traditional pathological staging (Table 3, Figure 5A).

Univariate Cox regression analysis identified several variables significantly associated with RFS, including GRIm–CONUT score, ypT stage, ypN stage, and pCR (all P < 0.05). Variables meeting the threshold of P < 0.10 were subsequently incorporated into the multivariate model. In the multivariate analysis, GRIm–CONUT score, ypT stage, and ypN stage remained independent prognostic factors for RFS. Patients with higher GRIm–CONUT scores experienced a significantly higher risk of disease recurrence (P = 0.018), while advanced ypT and ypN stages were also independently associated with inferior RFS (P < 0.05 for both).

These findings confirm that the GRIm–CONUT composite is an independent predictor of recurrence after neoadjuvant therapy and surgery for ESCC. Incorporating this immune–nutritional index into postoperative prognostic models may improve recurrence risk stratification beyond conventional pathological staging (Table 4, Figure 5B).

Based on the multivariable Cox model, we constructed a prognostic nomogram (nomogram–GRIm–CONUT) that integrates ypTNM stage with the GRIm–CONUT score to estimate postoperative survival in neoadjuvant-treated ESCC(Figure 6A). Internal validation using 1,000 bootstrap resamples demonstrated good model stability, with calibration curves showing close agreement between predicted and observed 3- and 5-year survival (Figures 7A, B).

The nomogram achieved a higher C-index than the conventional ypTNM staging system (0.717, 95% CI 0.657–0.777 vs 0.659, 95% CI 0.592–0.726), indicating superior discriminative ability(Figure 6B). Consistently, time-dependent ROC analyses showed higher AUCs for the nomogram across follow-up, and decision-curve analysis (DCA) demonstrated greater net clinical benefit over a relevant threshold probability range (0.2–1.0) compared with ypTNM alone (Figure 7C). Collectively, these results support the added prognostic value of incorporating immune–nutritional status into postoperative risk stratification.

Prespecified subgroup analyses demonstrated a consistent association between higher GRIm–CONUT categories and increased risk of death from OS across most strata. The direction and magnitude of effect were broadly concordant, indicating that the prognostic impact of the GRIm–CONUT score was not confined to any single subgroup (Figure 8). These findings support the generalizability of the composite immune–nutritional index for postoperative risk stratification in neoadjuvant-treated ESCC.