Metabolic-inflammatory index CTI identifies high-risk early-stage natural killer/T-cell lymphoma and informs a novel prognostic model Authors: Qiuhui Jiang, Peng Xu, Huijuan Zhong, Shuhui Fu, Liangjie Wang, Weili Zhao, Shu Cheng, Jie Zha, Bing Xu Published: 2026 ABSTRACT AbstractBackgroundThe prognostic heterogeneity of early-stage natural killer/T-cell lymphoma (NKTCL) remains inadequately defined. This study aimed to investigate the prognostic value and potential biological relevance of the metabolic-inflammatory biomarkers in early-stage NKTCL.MethodsWe conducted a secondary analysis on 166 patients with early-stage NKTCL from a multicenter clinical trial (NCT02631239). C-reactive protein-triglyceride-glucose Index (CTI) was calculated as: 0.412* Ln (CRP [mg/L]) + Ln (TG [mg/dL] × FPG [mg/dL])/2. Cox proportional hazards models and restricted cubic spline analyses were applied to assess the association and dose–response relationship between CTI and overall survival (OS). Mechanistic insights were explored through animal experiments and clinical lymphocyte-subset analyses. Independent prognostic factors were identified using multivariable Cox regression, and a composite prognostic model incorporating CTI, body mass index (BMI), and albumin (ALB) was subsequently developed and internally evaluated.ResultsHigh CTI was independently associated with inferior OS in patients with early-stage NKTCL, with a significant linear dose–response relationship. Patients with high CTI had a higher incidence of B symptoms and elevated lactate dehydrogenase and β2-microglobulin levels. A simplified CTI-BMI-ALB prognostic model demonstrated moderate predictive performance for 1-, 3-, and 5-year OS, with AUCs of 0.71, 0.71, and 0.69, respectively, and an overall C-index of 0.69 (95% CI: 0.59–0.79). In the present cohort, the CTI-BMI-ALB model showed improved discriminative performance compared with conventional prognostic indices and provided additional survival stratification among patients classified as low-risk or favorable-risk by these indices within this cohort. In vivo experiments showed that the Gubra-Amylin NASH diet-induced metabolic-inflammatory state with elevated CTI levels was associated with accelerated lymphoma progression, reduced intratumoral CD8+ T-cell infiltration, and increased proportions of Tregs and M2 macrophages. Consistently, clinical lymphocyte-subset analysis confirmed Treg enrichment in patients with high CTI and a positive association between CTI and the proportion of Tregs.ConclusionCTI is an independent prognostic biomarker in early-stage NKTCL, reflecting systemic metabolic-inflammatory and immune dysregulation. The CTI-BMI-ALB model, which integrates inflammatory, nutritional, and metabolic dimensions, showed potential for complementary risk stratification beyond conventional prognostic indices and warrants further validation before broader clinical application.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s40364-026-00996-y.Keywords: C-reactive protein-triglyceride-glucose index, Natural killer/T-cell lymphoma, Prognosis, Tumor microenvironment, Metabolic inflammation, Immune dysregulation Introduction Natural killer/T-cell lymphoma (NKTCL) is a highly aggressive subtype of non-Hodgkin lymphoma that is closely associated with Epstein–Barr virus (EBV) infection and shows a high prevalence in East Asia and Latin America [1, 2]. Early-stage disease, defined as Ann Arbor stage I–II, accounts for approximately 70%–90% of newly diagnosed cases [3]. Although chemoradiotherapy-based regimens have substantially improved survival outcomes, marked prognostic heterogeneity persists among patients with early-stage NKTCL [4]. Reported 5-year overall survival (OS) rates vary widely, ranging from 40% to 90% [5, 6], reflecting underlying disease heterogeneity as well as differences in treatment strategies. Existing prognostic scoring systems, such as the International Prognostic Index (IPI) [7] and the Prognostic Index for Natural Killer Lymphoma-Epstein–Barr Virus (PINK-E) [8], rely largely on conventional clinical parameters, including disease stage and performance status. However, these models may not adequately capture host-related inflammatory, metabolic, and immune abnormalities that contribute to outcome heterogeneity, particularly in patients with early-stage disease. Therefore, clinically accessible biomarkers that provide complementary biological information are needed to improve risk stratification. The triglyceride–glucose (TyG) index, calculated from fasting triglyceride and fasting glucose levels, is widely used as a simple surrogate marker of insulin resistance [9]. Metabolic dysfunction and systemic inflammation are closely interconnected processes that may jointly influence tumor progression, antitumor immunity, and cancer-related wasting [10–12]. C-reactive protein (CRP), a routinely available marker of systemic inflammation, has also been associated with adverse outcomes in several malignancies, including lymphoma [13, 14]. The C-reactive protein-triglyceride-glucose index (CTI), which integrates CRP with the TyG index, was developed to capture both systemic inflammation and insulin resistance [15]. Previous studies have demonstrated an association between elevated CTI and poor survival in patients with cancer and have further suggested its prognostic relevance in gastrointestinal malignancies and cancer cachexia [15–17]. However, whether CTI has prognostic and biological significance in lymphoma, particularly early-stage NKTCL, remains unknown. In this study, we hypothesized that CTI could serve as an integrated marker of systemic metabolic-inflammatory dysregulation and may provide prognostic information beyond conventional clinical indices in early-stage NKTCL. We therefore evaluated the association between CTI and survival, developed a simplified CTI-based prognostic model for clinical risk stratification, and further explored its potential biological relevance using an immunocompetent lymphoma model and clinical lymphocyte-subset analyses. Methods Study population This secondary analysis used data from a previously conducted multicenter, randomized, phase III non-inferiority trial (NCT02631239), which compared etoposide, dexamethasone, and pegaspargase (ESA) with methotrexate, etoposide, dexamethasone, and pegaspargase (MESA), both administered in combination with sandwich radiotherapy, in patients with early-stage extranodal nasal-type NKTCL [18, 19]. A total of 256 patients were initially identified from the trial database and 248 patients were in the modified intention-to-treat population. Of these, 82 were excluded because of missing baseline fasting plasma glucose (FPG), triglyceride (TG), or C-reactive protein (CRP) data. Consequently, 166 patients with complete baseline measurements were included in the final analysis. The comparisons of baseline characteristics between included and excluded patients were presented in Supplementary Table 1. The patient selection process is shown in Supplementary Fig. 1. The original trial was conducted in accordance with the Declaration of Helsinki, and written informed consent was obtained from all participants. CTI calculation and baseline variables Pre-treatment fasting plasma glucose, triglyceride, and CRP levels were measured in certified clinical laboratories at each participating center according to local standardized procedures. These baseline laboratory values were used to calculate CTI according to the following formula: CTI = 0.412×ln (CRP [mg/L]) + ln (TG [mg/dL]×FPG [mg/dL])/2. An exploratory optimal CTI cutoff for overall survival was determined using receiver operating characteristic curve analysis by maximizing the Youden index, defined as sensitivity + specificity-1. The resulting cutoff value was 9.3, and patients were accordingly categorized into high- and low-CTI groups. Other baseline variables collected from the trial database for subsequent analyses included age, sex, Ann Arbor stage, Eastern Cooperative Oncology Group (ECOG) performance status, B symptoms, Epstein-Barr virus (EBV)-DNA status, lactate dehydrogenase (LDH), β2-microglobulin (β2-MG), body mass index (BMI), and albumin (ALB). For immune-related analyses, baseline inflammatory cytokines and lymphocyte-subset data, including IL-6, TNF-α, IL-2, and the proportion of CD4+CD25+CD127low regulatory T cells (Tregs) were obtained from patients with available measurements. Exploratory external cohort An independent exploratory cohort was retrospectively identified from a previously published phase II clinical trial conducted at Ruijin Hospital, Shanghai, China, which evaluated MESA combined with sandwich radiotherapy in patients with newly diagnosed stage IE–IIE extranodal NKTCL, nasal type [20]. Among patients with available baseline clinical and laboratory data, 21 patients with complete CTI, BMI, and ALB measurements were included for exploratory evaluation of the CTI–BMI–ALB score. During follow-up, 4 OS events occurred. The score was calculated using the same predefined strategy as in the primary cohort. Given the limited sample size and number of outcome events, this analysis was considered exploratory rather than formal external validation. Animal experiments Model establishment Specific pathogen-free male C57BL/6 mice aged 6–7 weeks were randomly assigned to receive either standard chow or a Gubra-Amylin NASH (GAN) diet characterized by high fat and high sugar content (n = 6 per group). After 10 weeks of dietary intervention, blood samples were collected from the tail vein, and serum CRP, TG, and FPG levels were measured for calculation of CTI. After confirmation that CTI was significantly elevated in the GAN diet group, an immunocompetent syngeneic lymphoma model was established. EL4 is a murine lymphoma cell line derived from C57BL/6 mice and is characterized by mosaic expression of NK- and T-cell-associated molecules, making it suitable for establishing an immunocompetent syngeneic lymphoma model [21–23]. Log-phase EL4 cells were resuspended in sterile PBS at a concentration of 1 × 10⁷cells/mL and injected subcutaneously into the right dorsal flank at a volume of 0.2 mL per mouse. Measurements General monitoring: Body weight was recorded every 2 days. Tumor long (a) and short (b) diameters were measured every 2 days after inoculation, and tumor volume was calculated as V = 0.5 × a × b2; growth curves were plotted. General monitoring: Body weight was recorded every 2 days. Tumor long (a) and short (b) diameters were measured every 2 days after inoculation, and tumor volume was calculated as V = 0.5 × a × b2; growth curves were plotted. Immune cell profiling: Tumor and spleen tissues were harvested after cervical dislocation. Single-cell suspensions were prepared and analyzed by flow cytometry to quantify CD8+T cells, Tregs, exhausted CD8+ T cells (Tex), and M2 macrophages in the tumor microenvironment and spleen. Antibodies used are listed in Supplementary Table 2. Representative gating strategies are shown in Supplementary Fig. 2. Immune cell profiling: Tumor and spleen tissues were harvested after cervical dislocation. Single-cell suspensions were prepared and analyzed by flow cytometry to quantify CD8+T cells, Tregs, exhausted CD8+ T cells (Tex), and M2 macrophages in the tumor microenvironment and spleen. Antibodies used are listed in Supplementary Table 2. Representative gating strategies are shown in Supplementary Fig. 2. Statistical analysis All statistical analyses were performed using R version 4.3.0 and GraphPad Prism 9.0. Continuous variables were expressed as mean ± SD or median (P25, P75) according to distribution and compared using independent-samples t-tests or Mann-Whitney U tests. Categorical variables were expressed as n (%) and compared using χ² or Fisher’s exact tests. Kaplan-Meier survival curves were generated for overall survival (OS) and progression-free survival (PFS), and differences between groups were assessed using the log-rank test. Univariable and multivariable Cox proportional hazards regression analyses were performed to identify prognostic factors associated with OS. Variables with clinical relevance and/or statistical significance in univariable analyses were considered for multivariable analysis. The proportional hazards assumption was assessed using Schoenfeld residuals. Restricted cubic spline analysis with four knots placed at the 5th, 35th, 65th, and 95th percentiles of CTI was used to evaluate the dose–response relationship between CTI and mortality risk and assess potential nonlinearity. ROC curve analysis was used to determine the cohort-derived CTI cutoff value based on the Youden index and to assess the predictive performance of CTI and the CTI-BMI-ALB model. Model performance was evaluated using time-dependent ROC curves, Harrell’s C-index, calibration curves, and decision curve analysis. Internal validation was performed using bootstrap resampling with 1,000 repetitions. Likelihood ratio tests were used to evaluate the incremental prognostic value of the CTI-BMI-ALB score beyond established prognostic indices. Sensitivity analyses were performed using CTI as a continuous variable, an alternative median-based cutoff, and stabilized inverse probability of complete-case weighting. Exploratory immune analyses were performed using Spearman correlation analysis. Multivariable linear regression was used to evaluate associations between CTI and immune cell subsets after adjustment for relevant clinical covariates. Exploratory analyses in the independent cohort were performed using Kaplan-Meier and Cox regression analyses. Statistical significance was defined as two-sided P < 0.05. Results A total of 166 patients with early-stage NKTCL were included in the analysis, comprising 85 patients in the low-CTI group and 81 patients in the high-CTI group. Baseline demographic, staging, prognostic-score, and treatment characteristics were generally comparable between the two groups. No significant differences were observed in age, sex, Ann Arbor stage, ECOG performance status, IPI score, PINK-E score, NRI score [24], CA stage, EBV-DNA positivity, or treatment regimen (Table 1). However, patients in the high-CTI group exhibited several adverse clinical and laboratory features. B symptoms were more frequent in the high-CTI group than in the low-CTI group (50.62% vs. 29.41%, P = 0.005). In addition, elevated lactate dehydrogenase (49.38% vs. 31.76%, P = 0.021) and elevated β2-microglobulin levels (60.27% vs. 25.32%, P < 0.001), both of which are associated with greater disease activity or burden, were more common in patients with high CTI. Patients with high CTI also had lower albumin levels (36.87 ± 5.37 vs. 40.26 ± 5.09 g/L, P < 0.001), lower red blood cell counts (4.27 ± 0.65 vs. 4.55 ± 0.51 × 10¹²/L, P = 0.002), and lower hemoglobin levels (126.00 [114.00–137.00] vs. 137.00 [125.00–148.00] g/L, P < 0.001) than those with low CTI. Collectively, these findings indicate that high baseline CTI was associated with a less favorable clinical profile, including more frequent B symptoms, higher LDH and β2-microglobulin levels, and poorer nutritional and hematologic status. Table 1Clinical characteristics of enrolled patients stratified by CTI valueVariablesTotal (n = 166)Low CTI (n = 85)High CTI (n = 81)Statistic P Age > 60 years, n(%)χ²=0.930.334 No130 (78.31)64 (75.29)66 (81.48) Yes36 (21.69)21 (24.71)15 (18.52)Sex, n(%)χ²=0.120.728 Female43 (25.90)23 (27.06)20 (24.69) Male123 (74.10)62 (72.94)61 (75.31)B symptoms, n(%)χ²=7.79 0.005 No100 (60.24)60 (70.59)40 (49.38) Yes66 (39.76)25 (29.41)41 (50.62)Ann Arbor stage, n(%)χ²=1.860.173 I103 (62.05)57 (67.06)46 (56.79) II63 (37.95)28 (32.94)35 (43.21)ECOG score, n(%)χ²=1.960.161 058 (34.94)34 (40.00)24 (29.63) 1–2108 (65.06)51 (60.00)57 (70.37)IPI score, n(%)χ²=0.830.363 0–1151 (90.96)79 (92.94)72 (88.89) 2–315 (9.04)6 (7.06)9 (11.11)PINK-E score, n(%)χ²=1.050.332 074 (44.58)41 (48.24)33 (40.74) 1–292 (55.42)44 (51.76)48 (59.26)NRI score, n(%) 029 (17.47)18 (21.18)11 (13.58)χ²=6.570.087 164 (38.55)34 (40.00)30 (37.04) 251 (30.72)27 (31.76)24 (29.03) 3–522 (13.25)6 (7.06)16 (19.75)CA stage, n(%)χ²=2.880.237 I55 (33.13)33 (38.82)22 (27.16) II48 (28.92)24 (28.24)24 (29.63) III63 (37.95)28 (32.94)35 (43.21)EBV-DNA(+), n(%)78 (46.99)35 (41.18)43 (53.09)χ²=2.360.124High LDH, n(%)67 (40.36)27 (31.76)40 (49.38)χ²=5.35 0.021 High β2-MG, n(%)64 (42.11)20 (25.32)44 (60.27)χ²=19.02 < 0.001 WBC, Mean ± SD6.00 ± 2.296.09 ± 2.605.90 ± 1.93t = 0.540.592RBC, Mean ± SD4.42 ± 0.604.55 ± 0.514.27 ± 0.65t = 3.09 0.002 HGB, M (Q₁, Q₃)132.00 (121.00, 142.00)137.00 (125.00, 148.00)126.00 (114.00, 137.00)Z=-3.58 < 0.001 PLT, Mean ± SD219.78 ± 65.72222.52 ± 68.27216.90 ± 63.23t = 0.550.584TP, Mean ± SD67.77 ± 7.5068.40 ± 7.4867.11 ± 7.50t = 1.110.269ALB, Mean ± SD38.60 ± 5.4840.26 ± 5.0936.87 ± 5.37t = 4.17 < 0.001 BMI, Mean ± SD22.12 ± 2.8722.04 ± 2.7522.21 ± 3.00t=-0.380.702Treatment, n(%)χ²=0.590.442 MESA85 (51.20)46 (54.12)39 (48.15) ESA81 (48.80)39 (45.88)42 (51.85)t: t-test, Z: Mann-Whitney test, χ²: Chi-square test, -: Fisher exactAbbreviations: SD, standard deviation; M, Median; Q1, 1st Quartile; Q3, 3st Quartile; ECOG, Eastern Cooperative Oncology Group; IPI, International Prognostic Index; PINK-E, prognostic index of natural killer lymphoma-Epstein-Barr virus; NRI, nomogram-revised risk index; CA, Chinese Southwest Oncology Group and Asia Lymphoma Study Group; EBV, Epstein-Barr virus; LDH, lactic dehydrogenase; β2-MG, βeta-2-microglobulin; WBC, White blood cell; RBC, Red blood cell, ×10¹²/L; HGB, Hemoglobin; PLT, Platelet; TP, Total protein; ALB, Albumin; BMI, Body Mass Index; ESA, etoposide, dexamethasone, and pegaspargase; MESA, methotrexate, etoposide, dexamethasone, and pegaspargase Clinical characteristics of enrolled patients stratified by CTI value t: t-test, Z: Mann-Whitney test, χ²: Chi-square test, -: Fisher exact Abbreviations: SD, standard deviation; M, Median; Q1, 1st Quartile; Q3, 3st Quartile; ECOG, Eastern Cooperative Oncology Group; IPI, International Prognostic Index; PINK-E, prognostic index of natural killer lymphoma-Epstein-Barr virus; NRI, nomogram-revised risk index; CA, Chinese Southwest Oncology Group and Asia Lymphoma Study Group; EBV, Epstein-Barr virus; LDH, lactic dehydrogenase; β2-MG, βeta-2-microglobulin; WBC, White blood cell; RBC, Red blood cell, ×10¹²/L; HGB, Hemoglobin; PLT, Platelet; TP, Total protein; ALB, Albumin; BMI, Body Mass Index; ESA, etoposide, dexamethasone, and pegaspargase; MESA, methotrexate, etoposide, dexamethasone, and pegaspargase During a median follow-up of 71.8 months, 31 deaths and 44 progression events occurred in the CTI complete-case cohort. To evaluate the prognostic significance of CTI in early-stage NKTCL, survival outcomes were compared between the high- and low-CTI groups. The 1-, 3-, and 5-year OS rates were 83.95%, 75.31%, and 71.60%, respectively, in the high-CTI group, compared with 96.47%, 91.76%, and 90.59% in the low-CTI group (P = 0.008, P < 0.005, and P = 0.002, respectively; Fig. 1A). Kaplan-Meier analysis demonstrated significantly worse OS in the high CTI group than in the low CTI group, with an HR of 3.35 (95% CI: 1.50–7.50; log-rank P = 0.0017; Fig. 1B). Patients with high CTI also experienced poorer PFS during follow-up (Supplementary Figs. 3 A and 3B). Time-dependent ROC analysis demonstrated moderate predictive performance of CTI for 1-, 3-, and 5-year OS, with AUCs of 0.74 (95% CI: 0.60–0.88), 0.71 (95% CI: 0.59–0.82), and 0.70 (95% CI: 0.60–0.81), respectively (Fig. 1C and E). Restricted cubic spline analyses showed significant positive linear associations between CTI and the odds of death at 1, 3, and 5 years (overall P = 0.016, 0.011, and 0.011, respectively), with no evidence of nonlinear associations (P for nonlinearity = 0.978, 0.696, and 0.637, respectively; Fig. 1F and H). Collectively, these findings demonstrate that increasing CTI is consistently associated with poorer survival and support its value as a continuous prognostic biomarker in early-stage NKTCL. Prognostic value and predictive performance of CTI in patients with early-stage NKTCL. (A) Comparison of the 1-, 3-, and 5-year overall survival rates between the low- and high-CTI groups. (B) Kaplan–Meier curves for overall survival according to CTI group; differences were assessed using the log-rank test. (C–E) Time-dependent receiver operating characteristic curves evaluating the predictive performance of CTI for 1-year (C), 3-year (D), and 5-year (E) overall survival. (F–H) Restricted cubic spline analyses showing the dose–response relationship between CTI and the odds of death at 1 year (F), 3 years (G), and 5 years (H).Abbreviations: CTI, C-reactive protein–triglyceride–glucose index; NKTCL, natural killer/T-cell lymphoma; OS, overall survival; ROC, receiver operating characteristic; RCS, restricted cubic spline To assess the robustness of the ROC-derived CTI cutoff value, additional sensitivity analyses were performed using CTI as a continuous variable and an alternative cutoff based on the cohort median value. The association between CTI and OS remained consistent across different approaches. Continuous CTI was significantly associated with inferior OS (HR = 2.16, 95% CI: 1.42–3.29; P < 0.001). Similarly, dichotomization using the ROC-derived cutoff of 9.3 and the median cutoff of 9.29 showed comparable prognostic associations (HR = 3.36, 95% CI: 1.50–7.51; P = 0.003 and HR = 3.84, 95% CI: 1.65–8.91; P = 0.002, respectively; Supplementary Table 3). These findings indicate that the prognostic relevance of CTI was not solely dependent on the ROC-derived cutoff. Univariable Cox proportional hazards regression analysis was performed to identify factors associated with OS, and the results are presented in Fig. 2. High CTI was significantly associated with inferior OS (HR = 3.36, 95% CI: 1.50–7.51; P = 0.003). B symptoms (HR = 2.26, 95% CI: 1.11–4.61; P = 0.025), low albumin level (HR = 4.14, 95% CI: 2.04–8.38; P < 0.001), and BMI < 18.5 kg/m² (HR = 2.73, 95% CI: 1.10–6.78; P = 0.030) were also significantly associated with poorer OS. ECOG performance status of 1–2 showed a borderline association with OS but did not reach statistical significance (HR = 2.41, 95% CI: 0.99–5.87; P = 0.053). Age, sex, IPI, PINK-E, NRI, Ann Arbor stage, EBV-DNA status, LDH, β2-microglobulin, BMI > 25 kg/m², and treatment regimen were not significantly associated with OS in the univariable analyses (all remaining P values > 0.05). Forest plot of univariable Cox regression analyses for overall survival. Hazard ratios and 95% confidence intervals are shown for baseline clinical and laboratory variables. The vertical dashed line indicates a hazard ratio of 1.0. Abbreviations: OS, overall survival; HR, hazard ratio; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; IPI, International Prognostic Index; PINK-E, Prognostic Index for Natural Killer Lymphoma with Epstein–Barr virus DNA; NRI, nomogram-revised risk index; CTI, C-reactive protein–triglyceride–glucose index; ALB, albumin; EBV, Epstein–Barr virus; LDH, lactate dehydrogenase; β2-MG, beta-2-microglobulin; BMI, body mass index; ESA, etoposide, dexamethasone, and pegaspargase; MESA, methotrexate, etoposide, dexamethasone, and pegaspargase Given the limited number of OS events, a parsimonious multivariable model was used to reduce the risk of overfitting and unstable estimates. In the multivariable analysis, high CTI (HR = 2.55, 95% CI: 1.10–5.91; P = 0.028), low ALB level (HR = 2.81, 95% CI: 1.32–5.99; P = 0.007), and BMI < 18.5 kg/m² (HR = 3.06, 95% CI: 1.22–7.69; P = 0.017) remained independently associated with inferior OS, whereas B symptoms were not retained in the final model (Table 2). To further assess whether the prognostic association of CTI was independent of established clinical risk factors for NKTCL, we performed an extended multivariable Cox regression analysis incorporating EBV-DNA status, LDH, β2-microglobulin, B symptoms, albumin, BMI, and CTI. Both continuous CTI and CTI categorized using the ROC-derived cutoff of 9.3 remained independently associated with OS, supporting the robustness of the prognostic association of CTI (Supplementary Table 4). In a sensitivity analysis using stabilized inverse probability of complete-case weighting, high CTI remained significantly associated with inferior OS (HR = 3.74, 95% CI: 1.58–8.88; P = 0.003), consistent with the primary complete-case analysis. Table 2Multivariate Cox regression analysis of factors associated with overall survivalVariablesUnivariateMultivariateβS.EZ P HR (95%CI)βS.EZ P HR (95%CI)B-symptoms0.810.362.24 0.025 2.26 (1.11 ~ 4.61)High CTI1.210.412.95 0.003 3.36 (1.50 ~ 7.51)0.940.432.19 0.028 2.55 (1.10 ~ 5.91)Low ALB1.420.363.94 < 0.001 4.14 (2.04 ~ 8.38)1.030.392.68 0.007 2.81 (1.32 ~ 5.99)BMI stage 18.5–251.00 (Reference)1.00 (Reference) < 18.51.010.462.17 0.030 2.73 (1.10 ~ 6.78)1.120.472.38 0.017 3.06 (1.22 ~ 7.69) > 25-0.500.74-0.680.4980.61 (0.14 ~ 2.58)Abbreviations: HR, Hazard Ratio, CI, Confidence Interval; CTI, C-reactive protein–triglyceride–glucose index; ALB, Albumin; BMI, Body Mass Index; Multivariate Cox regression analysis of factors associated with overall survival Abbreviations: HR, Hazard Ratio, CI, Confidence Interval; CTI, C-reactive protein–triglyceride–glucose index; ALB, Albumin; BMI, Body Mass Index; Given the prognostic value of CTI in early-stage NKTCL, we developed a simplified points-based prognostic model incorporating high CTI, BMI < 18.5 kg/m², and low ALB. The model was developed in 158 patients with complete CTI, BMI, ALB, and OS data, including 29 deaths and 42 progression events. Each factor was assigned 1 point. Among the 158 patients with complete data for all three variables, 66, 62, and 30 patients were classified into the low-risk (0 points), intermediate-risk (1 point), and high-risk (2–3 points) groups, respectively (Fig. 3A). In the high-risk group, 12 deaths and 14 progression events were observed. Kaplan-Meier analysis showed a stepwise decrease in OS with increasing risk score, with significant differences among the three groups (log-rank P < 0.001; Fig. 3B). The model demonstrated moderate discrimination for 1-, 3-, and 5-year OS, with AUCs of 0.71 (95% CI: 0.56–0.87), 0.71 (95% CI: 0.59–0.82), and 0.69 (95% CI: 0.58–0.80), respectively (Fig. 3C and E). At each time point, the CTI-BMI-ALB model showed numerically higher AUCs than the IPI, PINK-E, and NRI within this cohort. Time-dependent AUC analysis indicated relatively stable discriminative performance during follow-up (Supplementary Fig. 4A), and the overall C-index was 0.69 (95% CI: 0.59–0.79). Bootstrap internal validation with 1,000 resampling iterations showed a similar optimism-corrected C-index (0.686), compared with the apparent C-index (0.687), suggesting limited optimism in the internal estimation of model performance. In comparison, the C-indices of the IPI, PINK-E, and NRI were 0.51, 0.54, and 0.50 respectively (Supplementary Table 5). The calibration curve showed acceptable agreement between the predicted and observed 5-year OS probabilities (Supplementary Fig. 4B). Decision curve analysis further indicated that the model showed higher estimated net benefit than the treat-all and treat-none strategies across a range of threshold probabilities (Supplementary Fig. 4C). The model also stratified PFS, with higher scores associated with poorer PFS (Supplementary Fig. 3C). Its AUCs for predicting 1-, 3-, and 5-year PFS were also higher than those of the IPI, PINK-E, and NRI (Supplementary Figs. 3D-3 F). To determine whether the CTI-BMI-ALB model provides additional prognostic stratification beyond conventional indices, we further compared Kaplan–Meier curves for IPI, PINK-E, NRI, and the CTI-BMI-ALB model in the unified 158-patient cohort. Although IPI, PINK-E, and NRI showed limited discrimination in this early-stage cohort, the CTI-BMI-ALB model further stratified OS among patients classified as low-risk or favorable-risk by these conventional indices (Supplementary Fig. 5). The addition of the CTI-BMI-ALB score significantly improved model fit compared with IPI, PINK-E, and NRI (all P < 0.001; Supplementary Table 6). Development and evaluation of a simplified CTI-BMI-ALB prognostic scoring model for OS. (A) Points-based scoring system incorporating high CTI, BMI < 18.5 kg/m², and low ALB. Each factor was assigned 1 point. (B) Kaplan–Meier curves for OS according to risk groups defined by the cumulative score. (C–E) ROC curves comparing the new model with IPI, PINK-E, and NRI for predicting 1-, 3-, and 5-year OS events. Abbreviations: CTI, C-reactive protein–triglyceride–glucose index; BMI, body mass index; ALB, albumin; OS, overall survival; ROC, receiver operating characteristic; AUC, area under the curve; CI, confidence interval; IPI, International Prognostic Index; PINK-E, Prognostic Index for Natural Killer Lymphoma with Epstein–Barr virus DNA; NRI, nomogram-revised risk index To further investigate the biological relevance of CTI and its potential association with lymphoma progression, we performed in vivo experiments using an immunocompetent syngeneic lymphoma model. C57BL/6 mice were randomly assigned to either a GAN diet, characterized by high sugar and high fat content, or a normal control diet. Following dietary intervention, mice in the GAN diet group exhibited greater weight gain over time and significantly higher CTI levels than control mice (P = 0.0055; Fig. 4A and B), indicating successful establishment of a systemic metabolic-inflammatory state associated with elevated CTI. After lymphoma cell inoculation, GAN diet-fed mice exhibited more rapid tumor growth and larger excised tumor volumes than control mice (Fig. 4C and E). Moreover, overall survival was significantly shorter in the GAN diet group (P = 0.0034; Fig. 4D). High CTI-related metabolic-inflammatory dysregulation is associated with T cell lymphoma progression and immunosuppressive microenvironment remodeling. C57BL/6 mice were fed either a standard control diet or a GAN diet to establish a systemic metabolic-inflammatory state, followed by subcutaneous inoculation of EL4 lymphoma cells. (A) Body weight changes during dietary intervention. (B) CTI levels in the control and GAN diet groups. (C) Tumor growth curves after lymphoma cell inoculation. (D) Kaplan–Meier survival curves of mice in the control and GAN diet groups. (E) Representative images of excised tumors at the experimental endpoint. (F–I) Flow-cytometric analysis of tumor-infiltrating immune cells, including CD8 + T cells (F), Tregs (G), Tex (H), and M2 macrophages (I). (J–M) Flow-cytometric analysis of splenic immune cell populations, including CD8 + T cells (J), Tregs (K), Tex (L), and M2 macrophages (M). Data are presented as mean ± SD unless otherwise indicated. P values were calculated using unpaired two-tailed Student’s t test or Mann–Whitney U test, as appropriate; survival differences were assessed using the log-rank test. Abbreviations: CTI, C-reactive protein–triglyceride–glucose index; GAN, Gubra-Amylin NASH; Tregs, regulatory T cells; Tex, exhausted CD8 + T cells Analysis of the tumor immune microenvironment revealed a substantial reduction in CD8+ T-cell infiltration (20.08% vs. 26.03%, P = 0.0086; Fig. 4F) and a concomitant increase in regulatory T cells (Tregs; 13.30% vs. 9.80%, P = 0.0078; Fig. 4G) and M2-type macrophages (50.25% vs. 27.13%, P < 0.0001; Fig. 4I). No significant difference in the proportion of tumor-infiltrating exhausted CD8+ T cells was observed between the two groups (P = 0.6570; Fig. 4H). Flow cytometric analysis of splenic immune populations showed no significant difference in the proportion of CD8+ T cells between the GAN and control groups (P = 0.4553; Fig. 4J). The proportion of splenic Tregs was numerically higher in the GAN diet group, although the difference did not reach statistical significance (6.85% vs. 5.54%, P = 0.0788; Fig. 4K). In contrast, significantly increased proportions of exhausted CD8+ T cells (Tex; 19.52% vs. 13.16%, P = 0.0014; Fig. 4L) and M2 macrophages (29.95% vs. 24.70%, P = 0.0411; Fig. 4M) were observed in the GAN diet group. Collectively, these findings indicate that the GAN diet-induced metabolic-inflammatory state characterized by elevated CTI was associated with both local and systemic immune alterations consistent with an immunosuppressive phenotype. Consistently, analysis of immune subsets and inflammatory biomarkers in the clinical cohort further supported the association between elevated CTI and an immunosuppressive inflammatory phenotype. Among patients with available CTI grouping and immune subset data, the high CTI group showed a significantly higher proportion of CD4+CD25+CD127low regulatory T cells than the low CTI group [1.00 (0.80–2.05) vs. 0.80 (0.30–1.37), P = 0.034; Supplementary Table 7]. In addition, inflammatory cytokines were significantly increased in the high CTI group, including IL-6 [8.49 (6.68–14.80) vs. 5.69 (3.77–10.38), P < 0.001], TNF-α [9.89 (7.57–11.85) vs. 7.84 (6.67–9.38), P = 0.009], and IL-2 [520.00 (385.00-720.00) vs. 419.00 (336.50–551.00), P = 0.030; Supplementary Table 7]. Moreover, CTI was positively correlated with CD4 + CD25+CD127low regulatory T cells (R = 0.251, P = 0.015; Supplementary Fig. 6). To determine whether this association was independent of clinical confounders, we performed multivariable linear regression analysis with Treg proportion as the dependent variable. After adjustment for age, B symptoms, albumin, and BMI, CTI remained independently associated with Treg proportion (P = 0.023; Supplementary Table 8). These clinical findings were concordant with the animal data, suggesting that a CTI-elevated metabolic-inflammatory state may be linked to immune remodeling toward a more immunosuppressive tumor-promoting milieu. Discussion In this study, we evaluated the prognostic value and biological relevance of CTI in patients with early-stage NKTCL and developed a simplified CTI–BMI–ALB prognostic model. High CTI was independently associated with inferior OS and showed a significant linear dose–response relationship with the odds of death. The CTI–BMI–ALB model demonstrated moderate predictive ability and showed improved discriminative performance compared with conventional prognostic indices in the present cohort. Notably, this model further stratified survival among patients classified as low-risk or favorable-risk by IPI, PINK-E, and NRI, suggesting that it may provide complementary risk information beyond conventional clinical parameters in early-stage cohort. CTI combines the TyG index, a surrogate marker of insulin resistance, with CRP, a widely used marker of systemic inflammation [16]. Previous studies have linked both insulin resistance and elevated CRP to adverse outcomes in several malignancies [25, 26]. In particular, diabetes mellitus, a clinical manifestation of systemic metabolic dysfunction and insulin resistance, has been associated with increased mortality in patients with cancer [27]. A systematic review and meta-analysis of 23 studies showed that cancer patients with preexisting diabetes had a significantly higher risk of all-cause mortality than those without diabetes [28]. Similarly, a meta-analysis of cohort studies in non-Hodgkin lymphoma reported that diabetes was associated with poorer OS, PFS, and lymphoma-specific survival [29]. By integrating these two components, CTI may provide a more comprehensive assessment of systemic metabolic-inflammatory dysregulation than either component alone. Insulin resistance and chronic inflammation are closely interconnected: impaired insulin signaling may contribute to abnormal glucose and lipid metabolism, oxidative stress, and inflammatory activation, whereas persistent inflammation may further aggravate insulin resistance [30]. This self-reinforcing interaction may influence tumor progression, antitumor immunity, and cancer-related wasting [31]. In line with this biological framework, restricted cubic spline analyses showed significant positive linear associations between CTI and the odds of death at 1, 3, and 5 years, with no evidence of nonlinearity. These findings suggest that the prognostic information carried by CTI is not restricted to a single cutoff and support its value as a continuous indicator of metabolic-inflammatory burden. CTI was selected because it integrates two biologically complementary dimensions, namely metabolic dysfunction/insulin resistance represented by the TyG index and systemic inflammation represented by CRP. To further evaluate its prognostic relevance, we compared CTI with other commonly used systemic biomarkers, including TyG, CRP, neutrophil-to-lymphocyte ratio (NLR), systemic immune-inflammation index (SII), and prognostic nutritional index (PNI). CTI showed the highest numerical C-index for OS prediction among these biomarkers in the current cohort, suggesting that integrating metabolic and inflammatory information may provide complementary prognostic information not fully captured by individual or predominantly single-domain biomarkers (Supplementary Table 9). Complementary experimental and clinical immune profiling provided biological support for these clinical observations. In the high-fat/high-sugar diet-induced mouse model, elevated CTI was associated with accelerated tumor growth and shorter survival. Immune profiling of tumor tissues revealed reduced CD8+ T-cell infiltration together with increased Tregs and M2 macrophages. Splenic immune profiling showed increased exhausted CD8+ T cells and M2 macrophages, suggesting systemic immune remodeling in the setting of diet-induced metabolic-inflammatory dysregulation accompanied by elevated CTI. Consistent with the animal findings, analysis of lymphocyte subsets in the clinical cohort showed a higher proportion of CD4+CD25+CD127low regulatory T cells in patients with high CTI. Moreover, CTI was positively associated with the proportion of this Treg population. High-CTI patients also exhibited elevated circulating IL-6, TNF-α, and IL-2 levels, consistent with an altered systemic cytokine milieu. Collectively, these observations are compatible with an “inflamed yet immunosuppressed” immune state in which systemic inflammatory activation coexists with enrichment of immunoregulatory cell populations. Although these findings do not establish a direct causal pathway, the concordance between the clinical cohort and the immunocompetent lymphoma model provides a plausible biological context for the association between elevated CTI and adverse outcomes in early-stage NKTCL. Several non-mutually exclusive mechanisms may contribute to this association. Persistent inflammatory cytokine signaling may impair antitumor T-cell activity and favor the expansion or polarization of immunoregulatory populations. Meanwhile, insulin resistance and dysregulated glucose and lipid metabolism may alter nutrient availability, oxidative stress, and the metabolic programming of immune cells. Tumor burden, cancer-related catabolism, and other host factors may also contribute to the observed immune phenotype. CTI should therefore be interpreted as an integrated marker of systemic metabolic-inflammatory dysregulation rather than as a direct biological effector. Further pathway-specific intervention and rescue experiments are required to determine the relative contributions and causal relationships of these mechanisms. A key contribution of this study is the development and evaluation of a simplified prognostic scoring model integrating CTI, BMI, and ALB. Low ALB is a well-recognized adverse prognostic marker in lymphoma and reflects more than nutritional impairment alone. Serum ALB is influenced by systemic inflammation, increased catabolic activity, disease burden, and reduced physiological reserve. In NKTCL, the prognostic relevance of ALB has been supported by several albumin-based indices. A multicenter study involving 1,022 patients showed that a low prognostic nutritional index, which incorporates ALB and lymphocyte count, independently predicted poorer survival and improved risk stratification when combined with established prognostic systems [32]. Other studies evaluating the prognostic nutritional index, albumin-to-globulin ratio, and inflammation-based scores have similarly demonstrated the prognostic significance of ALB-related status in NKTCL [33–35]. BMI provides complementary information regarding body size, energy reserve, and cancer-related wasting. In the present study, BMI < 18.5 kg/m², the conventional clinical threshold for underweight, was independently associated with inferior OS. This finding is consistent with previous studies in NKTCL showing that lower BMI was associated with poorer survival, whereas higher BMI at diagnosis was linked to more favorable outcomes [36, 37]. Some studies in other lymphoma populations have similarly described an apparent “obesity paradox,” in which overweight patients experienced better survival than patients with lower BMI [38, 39]. However, this association has not been consistent across lymphoma subtypes and clinical settings, and BMI > 25 kg/m² was not independently associated with OS in our cohort. Therefore, our findings primarily support the adverse prognostic significance of underweight status rather than a significant protective effect of overweight. Low BMI may reflect cancer-related wasting, reduced energy and muscle reserves, impaired immune competence, and diminished tolerance to disease burden and subsequent treatment [40, 41]. The adverse prognostic significance of low BMI is not inconsistent with the metabolic-inflammatory implications of high CTI. Insulin resistance can disrupt glucose and lipid metabolism, promote hyperglycemia and hypertriglyceridemia, and amplify systemic inflammation through increased circulating free fatty acids, oxidative stress, and activation of inflammatory signaling pathways [42]. In turn, persistent inflammation may further impair insulin signaling, creating a self-reinforcing metabolic-inflammatory cycle [42]. In patients with lymphoma, this metabolic-inflammatory state may coexist with tumor-related catabolism. Pro-inflammatory cytokines can suppress appetite and accelerate protein and lipid breakdown, resulting in weight loss and reduced physiological reserve [41, 43]. Consequently, patients with low BMI may still exhibit marked insulin resistance, dysregulated glucose and lipid metabolism, and elevated CRP, leading to a high CTI despite reduced body mass. Thus, BMI and CTI reflect distinct but interconnected biological dimensions: BMI < 18.5 kg/m² indicates reduced body reserve and cancer-related wasting, whereas CTI captures systemic metabolic-inflammatory dysfunction. By integrating CTI, ALB, and BMI, the proposed model captures systemic metabolic-inflammatory dysregulation together with host nutritional impairment and reduced physiological reserve. The CTI–BMI–ALB model demonstrated moderate and relatively stable predictive performance for 1-, 3-, and 5-year OS, with AUCs of 0.71, 0.71, and 0.69, respectively, and an overall C-index of 0.69 (95% CI: 0.59–0.79). Its prognostic association was generally consistent across the ESA and MESA treatment groups, with no significant interaction by treatment regimen (P for interaction = 0.862). Importantly, the model further stratified survival among patients categorized as low-risk or favorable-risk according to IPI, PINK-E, and NRI, suggesting that it provides complementary prognostic information beyond conventional clinical indices. Although conventional prognostic indices were not significantly associated with OS in the present cohort, this finding should not be interpreted as indicating that these established models lack prognostic value in NKTCL. Rather, their relatively limited performance in this analysis may be partly attributable to the moderate sample size, limited number of outcome events, restriction to early-stage disease, and the relatively homogeneous treatment context, as all patients received contemporary ESA- or MESA-based chemoradiotherapy. Differences between the present cohort and the populations in which these models were originally developed and validated may also have influenced their discriminatory performance. Collectively, these findings suggest that host metabolic-inflammatory and nutritional factors may refine risk stratification in early-stage NKTCL, particularly among patients who would otherwise be classified as low-risk by existing prognostic models. From a clinical perspective, the present findings highlight the potential value of incorporating host-related metabolic-inflammatory and nutritional status into risk assessment for early-stage NKTCL. Unlike conventional prognostic indices that primarily rely on disease- and treatment-related clinical parameters, the CTI-BMI-ALB model integrates readily available biomarkers reflecting systemic metabolic-inflammatory burden, nutritional impairment, and reduced physiological reserve. This approach may be particularly useful for refining risk stratification among patients who would otherwise be categorized as low-risk by existing models. However, the clinical applicability of this model requires further validation. To provide preliminary evidence regarding the reproducibility of our findings, we performed an exploratory evaluation in an independent cohort of 21 patients with available CTI, BMI, and albumin data. The baseline characteristics of this cohort are summarized in Supplementary Table 10. During follow-up, four OS events occurred. The CTI-BMI-ALB score showed a numerical separation toward different OS among risk groups (Supplementary Fig. 7). Exploratory Cox regression analyses demonstrated a consistent direction of association between higher CTI and inferior OS, although statistical significance was not reached, likely owing to the limited sample size and number of events (Supplementary Table 11). Therefore, these findings should be interpreted as exploratory and cannot substitute for formal external validation. Several limitations should be acknowledged. First, although the study cohort was derived from a prospective phase III clinical trial with standardized treatment protocols, the CTI–BMI–ALB model was developed and primarily evaluated within the same cohort. Bootstrap internal validation suggested limited optimism, and an independent exploratory cohort provided preliminary supportive evidence; however, the small sample size and limited number of events precluded formal external validation. Therefore, the proposed model should currently be considered exploratory and hypothesis-generating rather than an established clinical risk model. Second, the CTI cutoff of 9.3 was derived from the present cohort using the Youden index. Although analyses using continuous CTI and a median-based cutoff yielded consistent prognostic associations, the optimal threshold requires confirmation in independent populations. Third, approximately one-third of the modified intention-to-treat population was excluded because of missing CTI-related laboratory measurements. Although inverse probability weighting yielded results consistent with the primary analysis, residual selection bias due to unmeasured factors or non-random missingness cannot be excluded. Finally, the GAN diet-induced mouse model represents a complex systemic metabolic-inflammatory state rather than selective manipulation of CTI-related pathways; therefore, further mechanistic studies are needed to distinguish the independent and synergistic contributions of inflammatory and metabolic pathways. Larger prospective multicenter studies are warranted to further evaluate the reproducibility, generalizability, and potential clinical utility of CTI and the CTI-BMI-ALB model in early-stage NKTCL. Conclusion In summary, CTI is an independent and clinically accessible prognostic biomarker in early-stage NKTCL and reflects systemic metabolic-inflammatory and immune dysregulation. The CTI–BMI–ALB model showed potential for complementary risk stratification beyond conventional prognostic indices within this cohort by integrating metabolic-inflammatory burden, nutritional impairment, and host physiological reserve. Independent external validation is required before broader clinical application. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Supplementary Material 7 Supplementary Material 8 Acknowledgements The authors thank all patients and investigators who participated in the NHL-004 study. Author contributions Conception and design: QHJ, BX, JZ. Research performance: QHJ, JZ, BX. Collection and assembly of clinical data: WLZ, SC, HJZ, BX, QHJ. Animal experiments: PX, QHJ, SHF, LJW. Data analysis and interpretation: QHJ, HJZ, JZ, SC, BX. Manuscript writing: QHJ, HJZ, PX. Study supervision: SC, JZ, BX. Final approval of manuscript: all authors. Funding This study was supported by research funding from Noncommunicable Chronic Diseases-National Science and Technology Major Project (2026ZD0553900, 2023ZD0500700), and National Natural Science Foundation of China (82570231). Data availability The data that support the findings of this study are available from the corresponding author upon reasonable request, subject to approval by the NHL-004 study investigators and relevant institutional regulations. Declarations Ethics approval and consent to participate The original NHL-004 trial was conducted in accordance with the Declaration of Helsinki, and written informed consent was obtained from all participants. All animal experiments were approved by the Laboratory Animal Ethics and Management Committee of Xiamen University (Approval number: XMULAC20260117) and were performed in accordance with relevant guidelines and regulations. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Qiuhui Jiang, Peng Xu and Huijuan Zhong contributed equally to this work and share first authorship. Contributor Information Shu Cheng, Email: cs11123@rjh.com.cn. Jie Zha, Email: zhajie@xmu.edu.cn. Bing Xu, Email: xubingzhangjian@126.com. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Supplementary Material 7 Supplementary Material 8 Data Availability Statement The data that support the findings of this study are available from the corresponding author upon reasonable request, subject to approval by the NHL-004 study investigators and relevant institutional regulations.