Post-treatment triglyceride-glucose index as survival protective factors in patients with esophageal squamous cell carcinoma

Abstract

BACKGROUND

Patients with esophageal cancer exhibit marked variability in prognosis, highlighting the need for metabolism-related biomarkers. Although metabolic reprogramming is considered to be correlated with the prognosis of patients with esophageal squamous cell carcinoma (ESCC), its specific association mechanisms remain unclear.

AIM

To explore the triglyceride-glucose (TyG) index and its derivatives - indicators for insulin resistance, a core feature of metabolic syndrome. Given their controversial associations with cancer and limited research on their dynamic changes in ESCC patients post-treatment, this study aims to analyze these dynamic changes.

METHODS

The present retrospective study analyzed 360 East Asian patients with ESCC who received definitive chemoradiotherapy to explore the associations of TyG, TyG-body weight, and TyG-body mass index (both pre- and post-treatment) with overall survival.

RESULTS

Elevated levels of post-treatment TyG and its derivatives (postTyG, postTyG-body weight, postTyG-body mass index) were significantly associated with a reduced risk of death, showing a superior prognostic value compared to the baseline levels (preTyG). Restricted cubic spline analysis confirmed the presence of non-linear or monotonic associations, with more pronounced correlations observed in male patients.

CONCLUSION

Post-treatment TyG and its derivatives may serve as independent prognostic indicators for East Asian patients with ESCC, providing a basis for personalized diagnosis and treatment.

Key Words: Esophageal squamous cell carcinoma; Chemoradiotherapy; Insulin resistance; Triglyceride-glucose index; Metabolic syndrome

Core Tip: This study shows post-treatment triglyceride-glucose (TyG) index and its derivatives (TyG-body weight, TyG-body mass index) correlate with lower death risk in East Asian esophageal squamous cell carcinoma patients after definitive chemoradiotherapy, outperforming pre-treatment levels (especially in males) and acting as independent prognostic indicators, which may guide prognosis stratification, treatment monitoring, offer novel insights into personalized management, and serve as a robust basis for clinical decision optimization.

INTRODUCTION

Esophageal cancer (EC) ranks as the 11^th most common malignancy and the 7^th leading cause of cancer-related death globally. It exhibits distinct histopathological features across geographic regions, with esophageal squamous cell carcinoma (ESCC) predominating in East Asia, Southern Africa, and Eastern Africa, whereas esophageal adenocarcinoma (EAC) being more prevalent in Western countries[1,2]. This geographic divergence reflects etiological heterogeneity, with regional variations in genetic backgrounds, dietary patterns, and lifestyle factors possibly shaping the metabolic profiles of these subtypes[3-5]. However, the mechanisms linking metabolic reprogramming in esophageal tumors to clinical prognosis remain poorly elucidated.

Metabolic syndrome (MetS), characterized by obesity, hypertension, hyperglycemia, and dyslipidemia[6,7], is centered on insulin resistance (IR), a which has been identified as a key driver of carcinogenesis and a valid tool for assessing cancer risk[8-11]. The triglyceride-glucose (TyG) index, a robust surrogate for IR, not only is associated with various cardiovascular and metabolic diseases but also has been linked to prognosis in several cancer types, including lung, breast, and prostate cancers, in recent studies[10-12]. However, critical inconsistencies persist. A cohort study of 3.52 million individuals analyzed the TyG index-mortality association. It showed a significant inverse L-shaped nonlinear correlation between TyG and all-cause/cardiovascular mortality: Death risk rose sharply when TyG exceeded thresholds (9.75 for all-cause, 9.85 for cardiovascular mortality). In contrast, TyG had a linear negative association with cancer mortality, with a slight but significant downward trend in cancer death risk as TyG increased[13]. Additionally, TyG and its derivatives reportedly are correlated with an increased EAC risk and a decreased ESCC risk, highlighting a “metabolic paradox” that warrants further validation[14].

Notably, current research presents several critical limitations. First, studies on the association between TyG and cancer prognosis remain scarce, particularly in EC. Second, only one study to date has explored the relationship between TyG and EC risk, focusing exclusively on European populations, leaving a gap in data from Asian cohorts[14]. Finally, most studies have focused solely on the baseline TyG levels, neglecting the prognostic value of the dynamic changes in the TyG levels before and after treatment - especially metabolic adaptive remodeling following intensive anti-tumor therapy[11-13].

Thus, the present study, focusing on East Asian ESCC patients, aimed to analyze the prognostic value of the dynamic changes in the levels of TyG and its derivatives across different treatment stages, as well as their distribution patterns across clinical characteristics. The study utilized Cox proportional hazards and restricted cubic spline models to evaluate the strength and non-linear relationships between these indices (at different stages) and overall survival (OS). Ultimately, stratified analyses by sex and treatment modality (chemotherapy alone vs chemoradiotherapy) were performed to clarify the heterogeneity in their prognostic value. The core objective of the present study was to elucidate the independent prognostic significance of post-treatment TyG (postTyG) and its derivatives in East Asian ESCC patients, providing a basis for optimizing the risk stratification systems and developing metabolism-based personalized diagnostic and therapeutic strategies.

MATERIALS AND METHODS

Study population

The present retrospective study included 360 ESCC patients who underwent definitive chemoradiotherapy between 2012 and 2023. The eligibility criteria were as follows: (1) Histologically confirmed ESCC; (2) Unresectable tumor or patient refused surgery; (3) Karnofsky Performance Status (KPS) score of ≥ 70; (4) Radiotherapy (RT) dose of ≥ 50 Gy; (5) Availability of fasting blood test data before and after treatment; (6) No distant metastasis at baseline; and (7) Tumor-node-metastasis staging based on the 7^th edition of the American Joint Committee on Cancer classification.

Treatment protocol

All patients underwent RT. Spiral computed tomography (CT) simulation was performed for the tumor target volume and surrounding normal organ delineation. Concurrent chemotherapy was administered, with some patients receiving 1-6 cycles of platinum-based single-agent or combination chemotherapy, elderly patients being treated with oral S-1, and those intolerant to first-line chemotherapy receiving raltitrexed.

Data collection and index calculation

The patients’ clinical data, including age, sex, body weight, smoking and alcohol consumption status, body mass index (BMI), and chemoradiotherapy details, were extracted from their electronic medical records. For all participants, the serum biochemical parameters [triglycerides (TG), fasting glucose, total cholesterol (TC)] were collected before and after treatment, with the samples obtained after an overnight fast.

The following indices were calculated using the formulas below: TyG index (for male/female patients)[14]: TyG = ln[TG(mg/dL) × fasting glucose (mg/dL)/2]. TyG- body weight (BW) (for male/female patients): TyG-BW = TyG × BW. TyG-BMI (for male/female patients): TyG-BMI = TyG × BMI.

Follow-up and outcome measures

Follow-up evaluations were scheduled every 3 months in the first year, every 6 months in the next 2 years, and annually thereafter. The routine assessments included physical examination, blood tests, ultrasound examination, tumor marker assays, chest CT, and barium esophagography. All patients were followed up via outpatient visits or telephone calls.

Statistical analysis

This study performed statistical analyses and graphing using RStudio 4.3.1 software: For normally distributed continuous variables, Student’s t-test was applied for two-group comparisons while one-way analysis of variance was used for multi-group comparisons, and categorical variables were compared via the χ² test; for continuous covariates at different treatment stages, optimal cutoff values were determined by combining the log-rank test with a stratified approach - first, the log-rank test was used to analyze survival differences of continuous variables, initially identifying critical intervals that could distinguish survival outcomes, and subsequently candidate cutoff values were further validated with clinical stratification factors (e.g., tumor stage, treatment modality) to finally confirm the cutoff value that maximized intergroup survival differences; to avoid overfitting, Bootstrap resampling (1000 iterations) was conducted for internal validation of these cutoff values; accordingly, patients were stratified into low-level and high-level groups according to the cutoff values, survival analysis was performed using the Kaplan-Meier method to generate survival curves (with intergroup differences compared via the log-rank test), univariate and multivariate analyses were carried out using the Cox proportional hazards regression model, and Pearson correlation analysis was applied to investigate the associations between relevant indicators and clinical/metabolic parameters (correlation strength was defined as follows: Negligible correlation (r < ± 0.10), weak correlation (± 0.11, ± 0.39), moderate correlation (± 0.40, ± 0.69), strong correlation (± 0.70, ± 0.89), and very strong correlation (r > ± 0.90); a two-tailed P value < 0.05 was considered statistically significant.

RESULTS

Baseline characteristics and survival outcomes of the cohort

Altogether, 360 patients were enrolled; of these, 151 (41.9%) patients survived and 209 (58.1%) died at the last follow-up. The median OS for the entire cohort was 22.2 months [interquartile range (IQR): 12.0-38.8 months], with significantly longer OS in the survival group (38.6 months, IQR: 25.8-50.0 months) than in the deceased group (15.0 months, IQR: 8.5-23.2 months; P < 0.001). The cohort comprised 287 men (79.7%) and 73 women (20.3%), with a median age of 65 years (IQR: 58.8-70.0 years). No significant differences in smoking history (P = 0.924) or alcohol consumption (P = 0.409) were observed between the two groups. Most patients were diagnosed with stage III (57.2%) or IV (38.1%) disease; the survival group had a higher proportion of patients with stage I-II diseases, whereas the deceased group had a significantly higher proportion of patients with stage IV disease (P < 0.001). The median BMI was 21.6 kg/m² (IQR: 19.9-23.9 kg/m²), with the survival group showing a higher BMI (P = 0.018). The fasting glucose levels before treatment were lower in the survival group (P = 0.047), whereas the pre-treatment TC and TG levels showed no between-group differences. After treatment, the survival group had significantly higher level of TC (P = 0.002), TG (P < 0.001), postTyG index, postTyG-BW, and postTyG-BMI (all P < 0.01), with no significant difference in the pre-treatment TyG (preTyG) levels observed between groups (Table 1, Figure 1).

Figure 1 Distribution of clinical characteristics stratified by gender and survival status. Survival status (survival vs death); age (< 65 years vs ≥ 65 years); Karnofsky Performance Status (70, 80, 90, 100); smoking history (no vs yes); alcohol history (no vs yes); tumor location (cervical, upper thoracic, middle thoracic, lower thoracic, abdominal); clinical stage (I, II, III, IV); radiotherapy dose (< 60 vs ≥ 60); chemotherapy (no vs yes). RT: Radiotherapy; KPS: Karnofsky Performance Status.

Table 1 Main characteristics of the whole cohort and the groups with survival, n (%).

Characteristics	Overall	Survival	Death	P value
n	360	151 (41.9)	209 (58.1)	-
Survive time, median (IQR)	22.2 (12, 38.78)	38.57 (25.815, 50)	15 (8.47, 23.17)	< 0.001
Gender				0.527
Male	287 (79.7)	118 (32.8)	169 (46.9)
Female	73 (20.3)	33 (9.2)	40 (11.1)
Age, mean ± SD	65 (58.75, 70)	63.649 ± 7.5888	64.12 ± 9.497	0.602
KPS				0.006
70	6 (1.7)	0 (0)	6 (1.7)
80	144 (40)	50 (13.9)	94 (26.1)
90	208 (57.8)	101 (28.1)	107 (29.7)
100	2 (0.6)	0 (0)	2 (0.6)
Smoking history				0.924
No	142 (39.4)	60 (16.7)	82 (22.8)
Yes	218 (60.6)	91 (25.3)	127 (35.3)
Alcohol history				0.409
No	153 (42.5)	68 (18.9)	85 (23.6)
Yes	207 (57.5)	83 (23.1)	124 (34.4)
Tumor location, median (IQR)	2 (1, 3)	2 (1, 3)	2 (2, 3)	0.036
Clinical stage				< 0.001
I	1 (0.3)	1 (0.3)	0 (0)
II	16 (4.4)	12 (3.3)	4 (1.1)
III	206 (57.2)	106 (29.4)	100 (27.8)
IV	137 (38.1)	32 (8.9)	105 (29.2)
RT dose, median (IQR)	65 (60, 66)	60 (60, 66)	66 (60, 66)	0.001
Chemotherapy				0.029
No	39 (10.8)	10 (2.8)	29 (8.1)
Yes	321 (89.2)	141(39.2)	180 (50)
BMI, median (IQR)	21.6 (19.947, 23.885)	22 (20.33, 24.17)	21.11 (19.53, 23.44)	0.018
Pre-glucose, median (IQR)	5.13 (4.71, 5.8225)	5.03 (4.605, 5.78)	5.24 (4.77, 5.84)	0.047
Pre-TC, median (IQR)	4.725 (4.0475, 5.3625)	4.82 (4.06, 5.425)	4.69 (4.01, 5.3)	0.410
Pre-TG, median (IQR)	1.13 (0.8975, 1.45)	1.12 (0.9, 1.45)	1.13 (0.89, 1.46)	0.880
Post-BMI, mean ± SD	20.89 (19.325, 22.58)	21.537 ± 2.6983	20.904 ± 2.6611	0.072
Post-glucose, median (IQR)	5.31 (4.81, 5.94)	5.26 (4.825, 5.915)	5.33 (4.8, 5.94)	0.886
Post-TC, median (IQR)	4.48 (3.87, 5.29)	4.64 (4.025, 5.525)	4.26 (3.73, 5.17)	0.002
Post-TG, median (IQR)	1.3 (0.9875, 1.67)	1.43 (1.1, 1.775)	1.23 (0.92, 1.53)	< 0.001
PreTyG, median (IQR)	8.4781 (8.1915, 8.7593)	8.4357 (8.1754, 8.7729)	8.5087 (8.207, 8.748)	0.393
PostTyG, mean ± SD	9.8787 ± 0.31856	9.9299 ± 0.29319	9.8418 ± 0.33149	0.009
PostTyG-BW, median (IQR)	82.961 (73.243, 92.833)	85.855 (74.724, 96.954)	80.109 (72.236, 89.091)	0.004
PostTyG-BMI, median (IQR)	214.85 (195.92, 234.25)	218.76 (202.14, 237.33)	211.38 (192.76, 230.09)	0.006

Clinical stage was determined according to the 7^th edition of the American Joint Committee on Cancer classification. IQR: Interquartile range; KPS: Karnofsky Performance Status; RT: Radiotherapy; BMI: Body mass index; TC: Total cholesterol; TG: Triglycerides; TyG: Triglyceride-glucose; TyG-BW: Triglyceride-glucose-body weight; TyG-BMI: Triglyceride-glucose-body mass index.

Sex-specific differences in TyG-related indices across clinical subgroups

To explore the sex-specific effects on the TyG-related indices, we analyzed the distributions of preTyG, postTyG, postTyG-BW, and postTyG-BMI across the subgroups stratified by clinical features (age, KPS score, BMI, and TNM stage). Significant sex differences in preTyG levels were observed in the subgroups of age (< 65 years vs ≥ 65 years; P = 0.002), BMI (< 24 kg/m²vs ≥ 24 kg/m²; P = 0.045), and clinical stage (I-II vs III-IV; P = 0.011). PostTyG-BW showed significant sex differences across all subgroups (age, KPS score, BMI, and stage; all P < 0.05). For postTyG-BMI, significant sex differences were noted in age (< 65 years vs ≥ 65 years; P = 0.02), KPS score (70-80 vs 90-100; P = 0.02), and clinical stage (I-II: P = 0.047; III-IV: P = 0.006; Figure 2).

Figure 2 Distribution of triglyceride-glucose index stratified by clinical characteristics and gender. A: Pre-treatment triglyceride-glucose index levels across subgroups defined by age (< 65 years vs ≥ 65 years), Karnofsky Performance Status (70-80 vs 90-100), body mass index (< 24 kg/m²vs ≥ 24 kg/m²), and clinical stage (I-II vs III-IV), comparing female (blue) and male (orange) patients; B: Levels across subgroups defined by age (< 65 years vs ≥ 65 years), Karnofsky Performance Status (70-80 vs 90-100), body mass index (< 24 kg/m²vs ≥ 24 kg/m²), and clinical stage (I-II vs III-IV), comparing female (blue) and male (orange) patients. KPS: Karnofsky Performance Status; BMI: Body mass index; postTyG: Post-treatment triglyceride-glucose; preTyG: Pre-treatment triglyceride-glucose.

Restricted cubic spline analysis of TyG-related indices and prognosis

After adjusting for confounding factors (age, sex, KPS score, smoking/alcohol history, tumor location, TNM stage, and chemoradiotherapy), the restricted cubic spline plots revealed no significant dose-response relationship between preTyG and OS (overall P = 0.808; non-linear P = 0.698). Moreover, a reverse L-shaped non-linear association was found for postTyG (overall P = 0.001; non-linear P = 0.042), with postTyG of < 9.8 showing an association with an increased mortality risk [hazard risk (HR) > 1]. A monotonic decreasing association was noted for postTyG-BW (overall P < 0.001; non-linear P = 0.289), where the values of > 80 showed enhanced protective effects (decreasing HR with increasing postTyG-BW). A monotonic decreasing trend was observed for postTyG-BMI (overall P < 0.001; non-linear P = 0.078), with values of > 200 being identified as a protective factor for OS (Figure 3).

Figure 3 Restricted cubic spline analysis of the association between triglyceride-glucose-related indices and overall survival. A: Pre-treatment triglyceride-glucose (TyG) index; B: Post-treatment TyG index; C: Post-treatment TyG-body weight; D: Post-treatment TyG-body mass index. HR: Hazard ratio; CI: Confidence interval; preTyG: Pre-treatment triglyceride-glucose; postTyG: Post-treatment triglyceride-glucose; postTyG-BW: Post-treatment triglyceride-glucose-body weight; postTyG-BMI: Post-treatment triglyceride-glucose-body mass index.

Survival analysis of TyG-related indices

Kaplan-Meier analysis showed a median OS of 27.4 months for the entire cohort, with 1-, 3-, and 5-year OS rates of 75.2%, 43.7%, and 33.0%, respectively. For the subgroups stratified by pre-treatment indices, significant survival differences were observed for preTC (P = 0.072) and preTG (P = 0.042), but not for preGlucose (P = 0.124) or preTyG (P = 0.342). For the subgroups stratified by post-treatment indices, significant survival differences were noted for postTC (P = 0.001), postTG (P = 0.002), postTyG (P < 0.001), postTyG-BW (P < 0.001), and postTyG-BMI (P < 0.001), but not for postGlucose (P = 0.517). These findings highlight the superior prognostic value of post-treatment metabolic indices, particularly TyG derivatives (Figure 4).

Figure 4 Kaplan-Meier survival analysis of overall survival stratified by metabolic parameters before and after treatment. A-D: Survival curves stratified by pre-treatment glucose, pre-treatment total cholesterol, pre-treatment triglycerides, and pre-treatment triglyceride-glucose (TyG) index, respectively; E-H: Survival curves stratified by post-treatment glucose, post-treatment total cholesterol, post-treatment triglycerides, and post-treatment TyG index, respectively; I and J: Survival curves stratified by post-treatment TyG-body weight and post-treatment TyG-body mass index, respectively. preGlucose: Pre-treatment glucose; preTC: Pre-treatment total cholesterol; preTG: Pre-treatment triglycerides; preTyG: Pre-treatment triglyceride-glucose; postGlucose: Post-treatment glucose; postTC: Post-treatment total cholesterol; postTG: Post-treatment triglycerides; postTyG: Post-treatment triglyceride-glucose; postTyG-BW: Post-treatment triglyceride-glucose-body weight; postTyG-BMI: Post-treatment triglyceride-glucose-body mass index.

Correlation between TyG-related indices and clinical features

The Pearson correlation analysis showed that, in the total cohort, postTyG was significantly correlated with clinical stage [R = -0.133, P = 0.012, false discovery rate (FDR) P = 0.036] and survival status (R = -0.125, P = 0.017, FDR P = 0.045) after FDR correction, while its correlation with survival time (r = 0.119, P = 0.023) became non-significant after correction. PostTyG-BW showed correlations with multiple features (e.g., clinical stage, BMI, survival time; all P < 0.05) before correction, but none remained significant after FDR adjustment. PostTyG-BMI remained significantly correlated with clinical stage (R = -0.221, P < 0.001, FDR P = 0.001), chemotherapy (r = 0.164, P = 0.002, FDR P = 0.033), BMI (r = 0.971, P < 0.001, FDR P < 0.001), survival time (r = 0.210, P < 0.001, FDR P = 0.014), and survival status (R = -0.146, P = 0.005, FDR P = 0.033) even after FDR correction. In men, postTyG was significantly correlated with clinical stage (R = -0.190, P = 0.001, FDR P = 0.008) after correction. PostTyG-BW only retained a significant correlation with BMI (r = 0.821, P < 0.001, FDR P = 0.033) after FDR adjustment. PostTyG-BMI remained significantly correlated with clinical stage (R = -0.222, P < 0.001, FDR P = 0.002), chemotherapy (r = 0.182, P = 0.002, FDR P = 0.033), BMI (r = 0.969, P < 0.001, FDR P < 0.001), survival time (r = 0.269, P < 0.001, FDR P = 0.002), and survival status (R = -0.152, P = 0.010, FDR P = 0.042) post-correction. In women, postTyG had no significant correlations. PostTyG-BW showed a correlation with BMI (r = 0.726, P < 0.001) and near-significant association with RT dose (R = -0.212, P = 0.071), but neither was significant after FDR correction. PostTyG-BMI only remained significantly correlated with BMI (r = 0.964, P < 0.001, FDR P < 0.001) after correction (Table 2).

Table 2 Pearson correlation coefficients between post-treatment triglyceride-glucose-related indices and clinical characteristics.

Group	TyG index		Clinical stage	RT dose	Chemotherapy	BMI	Survival time	Survival status
Total patients	PostTyG	R	-0.133	-0.024	0.079	0.049	-0.125	0.119
		P value	0.012	0.645	0.136	0.351	0.017	0.023
		FDR P value	0.036	0.572	0.294	0.341	0.341	0.045
	PostTyG-BW	R	-0.218	0.052	0.133	0.787	-0.151	0.2
		P value	< 0.001	0.329	0.012	< 0.001	0.004	< 0.001
		FDR P value	0.739	0.572	0.739	0.163	0.501	0.402
	PostTyG-BMI	R	-0.221	0.059	0.164	0.971	-0.146	0.21
		P value	< 0.001	0.264	0.002	< 0.001	0.005	< 0.001
		FDR P value	0.001	0.46	0.033	< 0.001	0.014	0.033
Male	PostTyG	R	-0.19	-0.016	0.096	0.075	-0.115	0.142
		P value	0.001	0.79	0.104	0.204	0.051	0.016
		FDR P value	0.008	0.615	0.196	0.341	0.288	0.112
	PostTyG-BW	R	-0.245	0.072	0.174	0.821	-0.163	0.257
		P value	< 0.001	0.223	0.003	< 0.001	0.006	< 0.001
		FDR P value	0.957	0.458	0.779	0.033	0.572	0.485
	PostTyG-BMI	R	-0.222	0.121	0.182	0.969	-0.152	0.269
		P value	< 0.001	0.04	0.002	< 0.001	0.01	< 0.001
		FDR P value	0.002	0.288	0.033	< 0.001	0.002	0.042
Female	PostTyG	R	0.089	-0.074	0.023	-0.047	-0.156	0.029
		P value	0.453	0.535	0.849	0.691	0.187	0.807
		FDR P value	0.679	0.787	0.957	0.823	0.988	0.415
	PostTyG-BW	R	-0.158	-0.212	0.117	0.726	-0.124	-0.018
		P value	0.183	0.071	0.323	< 0.001	0.296	0.88
		FDR P value	0.453	0.779	0.79	0.558	0.739	0.572
	PostTyG-BMI	R	-0.195	-0.258	0.117	0.964	-0.106	-0.055
		P value	0.098	0.028	0.323	< 0.001	0.373	0.641
		FDR P value	0.458	0.447	0.558	< 0.001	0.82	0.541

RT: Radiotherapy; BMI: Body mass index; TC: Total cholesterol; TG: Triglycerides; TyG: Triglyceride-glucose; TyG-BW: Triglyceride-glucose-body weight; TyG-BMI: Triglyceride-glucose-body mass index; FDR: False discovery rate.

Correlation between TyG-related indices and clinical features

The Pearson correlation analysis showed that, in the total cohort, postTyG was significantly correlated with clinical stage (R = -0.133, P = 0.012), survival status (R = -0.125, P = 0.017), and survival time (r = 0.119, P = 0.023). Moreover, postTyG-BW was correlated with clinical stage (R = -0.218, P < 0.001), chemotherapy (r = 0.133, P = 0.012), BMI (r = 0.787, P < 0.001), survival status (R = -0.151, P = 0.004), and survival time (r = 0.200, P < 0.001). PostTyG-BMI was correlated with clinical stage (R = -0.221, P < 0.001), chemotherapy (r = 0.164, P = 0.002), BMI (r = 0.971, P < 0.001), survival status (R = -0.146, P = 0.005), and survival time (r = 0.210, P < 0.001).

In men, postTyG was correlated with clinical stage (R = -0.190, P = 0.001) and survival time (r = 0.142, P = 0.016). PostTyG-BW was correlated with clinical stage (R = -0.245, P < 0.001), chemotherapy (r = 0.174, P = 0.003), BMI (r = 0.821, P < 0.001), survival status (R = -0.163, P = 0.006), and survival time (r = 0.257, P < 0.001). PostTyG-BMI was correlated with clinical stage (R = -0.222, P < 0.001), RT dose (r = 0.121, P = 0.040), chemotherapy (r = 0.182, P = 0.002), BMI (r = 0.969, P < 0.001), survival status (R = -0.152, P = 0.010), and survival time (r = 0.269, P < 0.001).

In women, no significant correlations were observed for postTyG. PostTyG-BW was correlated with BMI (r = 0.726, P < 0.001) and showed a near-significant correlation with RT dose (R = -0.212, P = 0.071). PostTyG-BMI was correlated with BMI (r = 0.964, P < 0.001) and RT dose (R = -0.258, P = 0.028) (Table 2).

Adjusted associations between TyG-related indices and mortality risk

After adjusting for confounding factors, quartile-based analysis showed that, in the total cohort, postTyG Quartiles 2 [HR = 0.588, 95% confidence interval (CI): 0.397-0.872, P = 0.008] and 3 (HR = 0.654, 95%CI: 0.451-0.949, P = 0.025) were associated with a significantly reduced mortality risk, as compared to the reference quartile. Moreover, Quartile 4 showed a near-significant trend (HR = 0.694, 95%CI: 0.476-1.010, P = 0.056). postTyG-BW Quartile 4 (HR = 0.471, 95%CI: 0.312-0.713, P < 0.001) was associated with a significantly reduced mortality risk. postTyG-BMI Quartiles 3 (HR = 0.619, 95%CI: 0.422-0.908, P = 0.014) and 4 (HR = 0.535, 95%CI: 0.361-0.794, P = 0.002) showed a significantly reduced mortality risk (Figure 5).

Figure 5 Forest plot analysis of the prognostic value of post-treatment triglyceride-glucose-related indices across subgroups. Quartile 1 served as the reference, with hazard ratios, 95% confidence intervals, and P values presented. Analyses were adjusted for confounding factors including age, gender, Karnofsky Performance Status, smoking/alcohol history, tumor location, and chemoradiotherapy. A: Cox proportional hazards regression for quartiles in the total patient cohort; B: Female subgroup; C: Male subgroup. HR: Hazard ratio; CI: Confidence interval; postTyG: Post-treatment triglyceride-glucose; postTyG-BW: Post-treatment triglyceride-glucose-body weight; postTyG-BMI: Post-treatment triglyceride-glucose-body mass index.

DISCUSSION

The prognostic relevance of MetS across cancer types remains highly contentious, stemming primarily from the stage-specific effects of metabolic indices, showing distinct impacts on carcinogenesis vs the clinical outcomes, and their modulation by tumor subtype and population characteristics[15,16]. The present study addresses this gap by investigating the prognostic value of IR surrogates, specifically the TyG index and its derivatives (TyG-BW and TyG-BMI), before and after treatment in ESCC patients. We demonstrate that elevated postTyG-related indices are correlated with a significantly reduced mortality risk in ESCC patients, with this association remaining robust after adjusting for confounding factors. Critically, postTyG outperforms baseline TyG (preTyG) in prognostic prediction, highlighting that dynamic metabolic changes post-treatment more accurately demonstrate the host’s metabolic adaptability and anti-tumor responses. This finding identifies a novel time window for clinical prognostic assessment.

Existing evidence supports the stage-specificity of MetS and its components in ESCC. Regarding carcinogenesis, Lee et al[17] have reported that MetS is correlated with an increased risk of ESCC-dominant EC in a Korean cohort (adjusted HR = 1.11), with increased waist circumference and hypertension as key risk factors. Conversely, a meta-analysis by Zhang et al[18] found no significant association between MetS and ESCC risk (I² = 68.9%), although MetS was linked to an increased risk of developing EAC (pooled HR = 1.24). Such discrepancies likely stem from population heterogeneity; the high ESCC incidence in Korea is linked to fermented vegetable consumption and human papillomavirus infection, where MetS-driven inflammation may exacerbate this risk, whereas, in other regions, the MetS effects are overshadowed by confounding factors, including smoking and alcohol consumption.

Yang et al[14] have demonstrated that each 1-standard deviation increase in TyG index or METS-IR reduces the ESCC risk by 0.65-0.80-fold. Extending this, our data showed that elevated levels of postTyG derivatives are correlated with a reduced mortality risk, suggesting that the IR-related indices may exert protective effects throughout ESCC progression - mitigating both carcinogenesis and disease progression. Lindkvist et al[19] similarly reported that BMI is correlated with an increased EAC risk (relative risk = 7.34, 95%CI: 2.88-18.7) and a decreased ESCC risk (relative risk = 0.38, 95%CI: 0.23-0.62), aligning with our observation that post-treatment IR indices confer protection in ESCC patients. This is further supported by the negative correlation observed between postTyG and clinical stage (R = -0.133) and higher postTyG-BMI in patients with earlier-stage diseases.

Wen et al[20] have observed that ESCC patients with MetS exhibited superior survival and better tumor differentiation, suggesting that MetS may improve the outcomes by reducing tumor aggressiveness. Liu et al[21] have also identified diabetes as a protective factor in ESCC (HR = 0.668). Our findings extend these observations by highlighting the postTyG indices as superior prognostic markers; postTyG and postTyG-BMI were significantly higher in the survival group than in the deceased group (both P < 0.01), whereas preTyG showed no between-group differences. This “dynamic discrepancy” may arise from two mechanisms. The first mechanism is metabolic remodeling by treatment. ESCC chemoradiotherapy may upregulate TyG by enhancing insulin sensitivity, with elevated postTyG levels reflecting restored metabolic reserve and improved capacity to counteract cancer-related cachexia. This complements Liu et al’s finding[21] that weight loss increases the mortality risk (HR = 1.961), underscoring metabolic reserve as a critical prognostic determinant. The second mechanism is the stage-specific metabolic profile. PreTyG primarily reflects “inadequate intake + passive catabolism” during uncontrolled tumor growth, lacking protective effects; contrarily, the elevation of postTyG levels represents adaptive metabolic adjustments to support tissue repair and anti-tumor responses, explaining its prognostic utility.

Peng et al[22] have reported sex-specific effects of MetS on the mortality risk of ESCC patients, with significant associations found only in men (HR = 1.45); hyperglycemia and dyslipidemia emerged as the risk factors in men, whereas obesity predicted the poor outcomes in women. Our data confirmed the analogous sex specificity for postTyG and its derivatives, with inverse correlations with mortality restricted to men. This divergence may reflect the following: (1) Sex-based metabolic differences: The IR in male ESCC patients is more accurately quantifiable by measuring TyG (preTyG was significantly higher in men aged < 65 years with low BMI; P = 0.002), and post-treatment IR improvement may more potently enhance the immune function in men; and (2) Confounding in women: PostTyG-BMI in women is strongly correlated only with BMI (r = 0.964) and RT dose, suggesting that the outcomes women with ESCC are more dependent on radiation intensity, whereas those of their male counterparts are regulated by metabolic status.

Okamura et al[23] have shown that a high visceral-to-subcutaneous fat ratio predicts increased lymphovascular invasion, nodal metastasis, and poor response to neoadjuvant therapy among ESCC patients. Contrarily, our study found that postTyG is positively correlated with chemotherapy administration (r = 0.164, P = 0.002), with postTyG Quartiles 2-3 showing an association with a reduced mortality risk, suggesting that postTyG could monitor the treatment response. Clinically, dynamic postTyG monitoring may guide chemotherapy adjustments (e.g., enhanced nutritional support for patients with low postTyG levels). Additionally, the strong correlation between postTyG-BMI and BMI (r = 0.971) indicates that combining the metabolic and anthropometric indices (e.g., TyG-BMI) more accurately reflects the systemic status, outperforming single markers. This positions postTyG-BMI as a novel tool for ESCC risk stratification - particularly in men, where it acts as an independent prognostic factor.

In recent years, the integration of large-scale cohorts with radiomics/artificial intelligence (AI) models has further advanced the development of metabolic oncology and precision oncology. Among these advances, Wu et al[24] enrolled 589 postoperative patients with pancreatic ductal adenocarcinoma, including internal and external validation cohorts. They combined CT radiomics with metabolic indicators (e.g., visceral fat index) to predict early recurrence. Second, a large cohort study focusing on breast cancer patients used radiomics to link metabolic disorders to prognosis[25,26]. Additionally, radiomics and AI have advanced rapidly in predicting tumor prognosis and evaluating treatment responses, providing new tools for clinical assessment. For instance, Cao et al[27] constructed a magnetic resonance imaging-based radiomics model for non-small cell lung cancer, which can accurately predict molecular changes and treatment responses. Fan et al[28] developed a CT-based delta-radiomics model for EC to effectively evaluate treatment responses. Wen et al[29] applied AI to immunotherapy, confirming its value in predicting therapeutic efficacy and toxicity. In comparison, the TyG index and its derived indicators - the focus of this study - exhibit superior clinical accessibility. The TyG index is calculated using conventional biochemical markers. It requires no specialized equipment, can be obtained via standard blood tests, and offers advantages including low cost, ease of operation, and the ability for dynamic monitoring.

However, this study has several limitations. First, this study adopts a single-center retrospective design. Inherent selection bias in this design cannot be completely avoided. In the future, verifying the reliability of the conclusions will require a multi-center prospective cohort. Second, female patients accounted for approximately 20% of the study population. This proportion aligns with the epidemiological characteristics of EC. However, it reduces the reliability of sex-stratified conclusions to some extent. Thus, the sex-stratified conclusions remain referenceable. Future plans include collaborating with centers with a high incidence of female EC to expand the sample size for further verification. Finally, several metabolic indicators (e.g., homeostatic model assessment for insulin resistance, hypertension) were not included in this study. This omission hinders the comprehensive evaluation of IR’s impact on EC. Additionally, the mechanism underlying dynamic changes in metabolic indicators was not explored in depth. Meanwhile, the study was based on data from East Asian patients and lacked external validation. Further evidence should be obtained from multi-center, large-sample, and multi-population studies.

CONCLUSION

The postTyG, postTyG-BW, and postTyG-BMI may serve as prognostic indicators for ESCC, with elevated levels associated with a reduced risk of death; these IR surrogate markers may also hold potential as therapeutic response markers, providing a foundation for their integration into the personalized management of ESCC, and future studies will explore the mechanisms linking IR improvement to ESCC prognosis, including by manipulating insulin sensitivity at the ESCC cell level, simulating changes in TyG levels and metabolomic profiles in chemoradiotherapy animal models, and integrating these experimental approaches with dynamically measured TyG indices and immune markers in clinical settings.

ACKNOWLEDGEMENTS

We appreciate everyone who participated in the study.