Abstract
Chronological age is a recognized cancer risk factor but does not capture inter-individual differences in systemic aging. Biological age, reflecting cumulative functional decline, may better predict cancer outcomes. This study examined associations between biological age, disease characteristics, metastasis, and survival in gastrointestinal cancers. Methods: We retrospectively analyzed 3,074 patients. Biological age was estimated from clinical biomarkers, and ΔAge (biological–chronological age) was calculated. Patients were grouped by biological age quartiles. Logistic and Cox regression, Kaplan–Meier, restricted cubic splines, and subgroup/sensitivity analyses were conducted. A prognostic nomogram integrating biological age was developed and validated. Results: Higher biological age correlated with older chronological age, male sex, smoking, systemic inflammation, and lower nutrition. ΔAge was largest in patients < 50 years. Biological age was not associated with TNM stage but independently predicted metastasis (per 1-SD: OR = 1.02, 95% CI: 1.01–1.03). The highest quartile nearly doubled metastasis risk, most evident in patients aged 50–64 years. During a median 47.9-month follow-up, 951 deaths occurred. Elevated biological age predicted poorer survival (highest vs. second quartile: HR = 2.04, 95% CI: 1.58–2.62), with nonlinear dose–response associations. Prognostic value was stronger in middle-aged and older patients. The nomogram demonstrated good discrimination (C-index 0.687–0.732; AUCs up to 0.846).Conclusions: Biological age, independent of chronological age and TNM stage, is a strong predictor of metastasis and survival in gastrointestinal cancers. ΔAge is greatest in younger individuals, while prognostic impact is most pronounced in middle-aged and older patients. Incorporating biological age may improve risk stratification and individualized management.
Supplementary Information
The online version contains supplementary material available at 10.1038/s41598-026-59078-6.
Keywords: Biological age, Gastrointestinal cancer, Metastasis, Survival, Chronological age, ΔAge, Prognostic biomarker, Nomogram
Subject terms: Biomarkers, Cancer, Oncology, Risk factors
Introduction
Digestive tract cancers, including gastric and colorectal cancer, remain a major public health burden worldwide1. Despite advances in screening, surgical techniques, and systemic therapies, the overall prognosis of these malignancies remains suboptimal. Traditionally, gastrointestinal cancers have been considered diseases of the elderly; however, an emerging epidemiological trend in the past two decades is the increasing incidence of early-onset cases, particularly in patients younger than 50 years2. Large-scale population-based studies have documented that early-onset gastrointestinal cancer is not only becoming more prevalent but is often characterized by aggressive clinical features, advanced disease stage at diagnosis, and distinct molecular signatures when compared with late-onset counterparts3.
Conventionally, chronological age has been regarded as a key determinant of physiological reserve, comorbidity burden, and treatment tolerance. Younger patients are often assumed to have superior organ function, stronger resilience, and a more favorable prognosis. Nonetheless, clinical observations challenge this assumption. A subset of early-onset gastrointestinal cancer patients paradoxically experience poor outcomes despite their younger chronological age, raising the hypothesis that chronological age alone may be insufficient to capture the true biological condition of cancer patients4,5.
In recent years, the concept of biological age has gained increasing attention in oncology and aging research. Unlike chronological age, which simply reflects the passage of time, biological age integrates physiological, biochemical, and molecular indicators of aging, thereby providing a more comprehensive assessment of an individual’s health status6. Various approaches have been developed to estimate biological age, including DNA methylation-based “epigenetic clocks” (Horvath, Hannum, PhenoAge, GrimAge), systemic inflammation-based scores derived from C-reactive protein (CRP), interleukin-6 (IL-6), and composite indices incorporating clinical laboratory biomarkers such as albumin, glucose, creatinine, and neutrophil-to-lymphocyte ratio (NLR)7. Accelerated biological aging, defined as a higher biological age compared with chronological age (ΔAge > 0), has been shown to predict morbidity, mortality, and poor treatment outcomes across a variety of diseases, including cardiovascular disease, neurodegeneration, and cancer8.
Although accelerated aging has been explored in breast cancer, pancreatic cancer, and hematological malignancies, studies specifically focusing on gastrointestinal tumors are scarce. Moreover, whether early-onset patients are truly “biologically young” or whether they already harbor signs of premature aging remains largely unknown. Addressing this knowledge gap may have important clinical implications, as it could inform individualized treatment strategies, surveillance policies, and lifestyle interventions9–11.
To this end, the present study utilized a large multicenter cohort from the Investigation on Nutrition Status and Clinical Outcome of Common Cancers in China (INSCOC) project to comprehensively evaluate the biological age of patients with digestive tract cancers. We aimed to (1) investigate the distribution of biological age across different chronological age groups, (2) explore the relationship between biological age and disease severity, including TNM stage and metastatic status, and (3) assess the prognostic significance of biological age in predicting overall survival. We further applied sensitivity and subgroup analyses to validate the robustness of our findings. Through this work, we sought to elucidate whether early-onset gastrointestinal cancer patients exhibit premature biological aging and to clarify the clinical value of integrating biological age assessment into cancer care.
Methods
Study population
This multicenter, observational cohort study was conducted as part of the Nutritional Status and Clinical Outcome of Common Malignant Tumors (INSCOC) research project (registration number: ChiCTR1800020329, registered at chictr.org.cn). The INSCOC (Investigation on Nutrition Status and Clinical Outcome of Cancer) project is a large-scale, nationwide, hospital-based research platform initiated in 2013 under the leadership of Professor Shahanping Shi, involving 152 participating medical institutions across China. Fujian Cancer Hospital is one of the participating centers and is among the major contributing units in terms of clinical data entry within this collaborative network.
This retrospective cohort study was conducted using data from the INSCOC project, a large-scale, multicenter investigation initiated in 2013 to examine the nutritional status and clinical outcomes of common cancers in China. Between 2013 and 2024, a total of 10,081 patients diagnosed with digestive tract cancers (gastric cancer and colorectal cancer) were identified. Patients without available biological age data (n = 6,771) were excluded. Among the 3,310 patients with sufficient biological age information, those with missing data on TNM stage (n = 193), body mass index (BMI, n = 8), serum albumin (n = 9), or NLR (n = 26) were further excluded. The final analytic cohort consisted of 3,074 patients (Fig. 1).The study design was approved by the local ethics committee(s) of all participating hospitals. All procedures were conducted in accordance with the Declaration of Helsinki and its later amendments, as well as relevant national and institutional guidelines and regulations.The study protocol was reviewed and approved by the institutional ethics committee(s) of the participating centers (including Fujian Cancer Hospital; approval details available upon request or provided in the ethics documentation).Given the retrospective nature of the study and the use of fully anonymized and de-identified data, the requirement for written informed consent was waived by the institutional ethics committee(s), in accordance with applicable regulations and ethical standards.
Fig. 1.
Flowchart of the study.
Estimation of biological age
Biological age (BioAge) was calculated using a composite algorithm integrating routinely available laboratory and clinical biomarkers, including serum albumin, glucose, creatinine, and inflammatory indices such as NLR. This approach has been validated in prior studies as a feasible proxy for biological aging in large-scale cohorts where DNA methylation data are unavailable12.
For analytic purposes, BioAge was categorized into quartiles based on its distribution within the study cohort:
Quartile 1 (Q1): < 47.262
Quartile 2 (Q2): 47.262 – < 56.390
Quartile 3 (Q3): 56.390 – < 64.776
Quartile 4 (Q4): ≥ 64.776
Quartile 2 served as the reference category in regression models. For sensitivity analyses, biological age was alternatively classified into tertiles (< 50.920, 50.920 – < 61.731, ≥ 61.731) to assess the stability of the results.
The biological age acceleration (ΔAge) was defined as the difference between estimated biological age and chronological age:
 |
A positive ΔAge indicates accelerated biological aging, whereas a negative ΔAge reflects decelerated aging.
Data collection
The baseline for survival analysis was defined as the date of initial cancer diagnosis. To ensure the representation of patients’ baseline clinical status, all laboratory parameters were abstracted from clinical records within 48 h following the diagnosis.
Outcomes
The primary endpoint of this study was all-cause mortality. Overall survival (OS) was utilized as the time-to-event metric. OS duration was calculated as the time elapsed from the date of hospital admission to the date of death from any cause (the event) or the date of the last follow-up (censoring), whichever occurred first.
Statistical analysis
Baseline demographic and clinical characteristics were summarized using medians and interquartile ranges (IQRs) for continuous variables, and frequencies with percentages for categorical variables. Group comparisons were performed using the Mann–Whitney U test or Kruskal–Wallis test for continuous variables and the χ2 test for categorical variables.
A comparison of the baseline characteristics between the included complete-case sample and the sample excluded due to missing biomarkers (Supplementary Table 1) revealed that the missingness of biomarkers in this study was not missing completely at random (MCAR). This study adopted a complete case analysis approach, excluding subjects with missing data for any core biomarker and retaining only the complete data set for subsequent analysis.
To examine differences in ΔAge across chronological age groups, nonparametric tests were employed. Logistic regression analyses were performed to assess the associations between biological age and disease severity, including TNM stage and metastatic status.
Survival outcomes were analyzed using the Kaplan–Meier method, with differences across groups evaluated by the log-rank test. The Cox proportional hazards regression model was applied to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between biological age categories and overall survival. The proportional hazards assumption was verified prior to model application. Multivariable Cox models were adjusted for potential confounders, including chronological age, sex, BMI, family history of cancer, TNM stage, smoking status, and alcohol consumption.
To assess potential nonlinear associations between biological age (as a continuous variable) and mortality risk, restricted cubic spline (RCS) regression models were employed. Subgroup and interaction analyses were further conducted stratified by sex, chronological age group, BMI category, smoking status, and TNM stage to explore effect modification.
Model stability was assessed through sensitivity analyses, including (1) excluding patients who died within the first three months of follow-up to minimize immortal time bias and (2) repeating analyses using tertile-based categorization of BioAge. Variance inflation factor (VIF) diagnostics were applied to detect multicollinearity, with VIF values < 2 indicating no significant multicollinearity.
In this study, we first performed univariate Cox proportional hazards regression on clinical and laboratory variables in the training cohort to identify factors associated with overall survival (OS). Variables with P < 0.05 were then incorporated into a multivariate Cox regression model, and stepwise selection was applied to exclude non-significant factors, leading to the identification of independent prognostic indicators. Model performance was evaluated in several ways: Harrell’s concordance index (C-index) was used to assess discrimination, ROC curves were constructed to evaluate predictive accuracy for 1-, 3-, and 5-year survival, and calibration curves were generated to compare predicted probabilities with observed outcomes. In addition, decision curve analysis (DCA) was performed to estimate the clinical utility of the model by quantifying net benefit across a range of threshold probabilities. Finally, the model was validated in an independent external cohort from Fujian Cancer Hospital to confirm its generalizability and robustness.
All statistical analyses were performed using SPSS version 26.0 (IBM Corp., Armonk, NY, USA) and R version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided P value < 0.05 was considered statistically significant.
Results
Baseline characteristics
Figure 2 depicts the differences in biological age (ΔAge) across chronological age groups. Statistically significant differences were observed between groups (all P < 0.05). Notably, as chronological age increased, the magnitude of ΔAge gradually decreased. Patients younger than 50 years demonstrated substantially greater ΔAge values compared with those aged ≥ 65 years, suggesting that the discrepancy between biological and chronological age diminishes with advancing age.
Fig. 2.
Differences in biological age (ΔAge) across different chronological age groups.
Baseline clinical and demographic characteristics of the study cohort are summarized in Table 1. A total of 3,074 patients with gastrointestinal cancers were included, among whom 2,013 (65.5%) were male. The median chronological age at enrollment was 60 years (IQR = 15). When stratified by quartiles of biological age, distinct distributions of patient characteristics were observed. Compared with the second quartile, patients in the fourth quartile were more frequently male, older, and less likely to report a family history of cancer. They also exhibited lower serum albumin levels and higher neutrophil-to-lymphocyte ratio (NLR), both of which are markers of systemic inflammation and nutritional status. By contrast, patients in the first quartile were more often female, younger, non-smokers, and non-drinkers, and they tended to have higher albumin but lower body mass index (BMI) levels.
Table 1.
Baseline characteristics of patients.
| Characteristic | Overall (n = 3074) |
Bioage category | P value | |||
|---|---|---|---|---|---|---|
| Quartile 1 (n = 769) | Quartile 2 (n = 768) | Quartile 3(n = 768) | Quartile 4(n = 769) | |||
| Gender, male, n (%) | 2013(65.5) | 397(51.6) | 504(65.6) | 531(69.1) | 581(75.6) | < 0.001 |
| Age, years, median (IQR) | 60(15) | 47(10) | 56(7) | 64(5) | 72(8) | < 0.001 |
| Age group, n (%) | < 0.001 | |||||
| < 50 | 557(18.1) | 511(66.4) | 42(5.5) | 2(0.3) | 2(0.3) | |
| 50–64 | 1463(47.6) | 258(33.6) | 697(90.8) | 437(56.9) | 71(9.2) | |
| ≥ 65 | 1054(34.3) | 0 | 29(3.7) | 329(42.8) | 696(90.5) | |
| Smoke, yes, n (%) | 1320(42.9) | 254(33.0) | 343(44.7) | 367(47.8) | 356(46.3) | < 0.001 |
| Alcohol, yes, n (%) | 696(22.6) | 132(17.2) | 193(25.1) | 186(24.2) | 185(24.1) | < 0.001 |
| Family history, yes, n (%) | 496(16.1) | 150(19.5) | 138(18.0) | 122(15.9) | 86(11.2) | < 0.001 |
| TNM stage, n (%) | 0.504 | |||||
| Ⅰ | 210(6.8) | 54(7.0) | 46(6.0) | 57(7.4) | 53(6.9) | |
| Ⅱ | 607(19.7) | 151(19.6) | 158(20.6) | 153(19.9) | 145(18.9) | |
| Ⅲ | 1353(44.0) | 344(44.8) | 353(46.0) | 338(44.0) | 318(41.3) | |
| Ⅳ | 904(29.4) | 220(28.6) | 211(27.4) | 220(28.7) | 253(32.9) | |
| Tumor metastasis, yes, n (%) | 834(27.1) | 203(26.4) | 208(27.1) | 202(26.3) | 221(28.7) | 0.686 |
| Albumin, g/L, median (IQR) | 39.30(7.20) | 40.90(6.10) | 39.80(6.55) | 39.00(6.50) | 36.30(7.50) | < 0.001 |
| NLR, median (IQR) | 2.41(2.48) | 2.01(1.76) | 2.16(2.07) | 2.48(2.50) | 3.18(3.54) | < 0.001 |
| BMI, kg/m2, median (IQR) | 21.96(4.61) | 21.39(4.52) | 21.88(4.72) | 22.25(4.26) | 22.04(4.57) | < 0.001 |
| BMI category, n (%) | 0.002 | |||||
| Underweight (< 18.5 kg/m2) | 446(14.5) | 132(17.2) | 95(12.4) | 90(11.7) | 129(16.8) | |
| Normal (18.5–23.9 kg/m2) | 1791(58.3) | 457(59.4) | 455(59.2) | 438(57.0) | 441(57.3) | |
| Overweight (24–27.9 kg/m2) | 715(23.3) | 148(19.2) | 184(24.0) | 211(27.5) | 172(22.4) | |
| Obesity (≥ 28 kg/m2) | 122(3.9) | 32(4.2) | 34(4.4) | 29(3.8) | 27(3.5) | |
Data are represented as median (IQR) or number (percentage). IQR, interquartile range; BMI, body mass index; cumBMI, cumulative body mass index.
In summary, baseline analyses revealed that higher biological age was associated with older chronological age, adverse clinical profiles, and unfavorable host-related characteristics, highlighting its potential role as a biomarker of physiological vulnerability.
Biological age and disease severity
The relationship between biological age and disease severity is presented in Fig. 3.
Fig. 3.
Associations between biological age and disease severity. (A) TNM stage; (B) Tumor metastasis.
A significant difference in biological age was observed between patients with TNM stage III and stage IV disease (P < 0.05). However, no statistically significant differences were noted across the earlier TNM categories or between groups with and without metastatic disease (P > 0.05).
Further analyses using logistic regression models are shown in Tables 2 and 3. In both univariable and multivariable models, biological age was not significantly associated with TNM stage (all P > 0.05), indicating that biological age did not differentiate between early-stage and advanced-stage disease when using TNM classification as the outcome.
Table 2.
Associations between biological age and TNM stage (TNM stage was dichotomized: stage I–II vs. stage III–IV).
| Bioage | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | |
| As continuous (per SD) | 1.00(1.00–1.01) | 0.765 | 1.00(0.99–1.01) | 0.853 | 1.00(0.99–1.01) | 0.996 |
| Q1 | 0.95(0.71–1.27) | 0.706 | 0.94(0.70–1.26) | 0.693 | 0.92(0.68–1.23) | 0.566 |
| Q2 | Ref | Ref | Ref | |||
| Q3 | 0.99(0.77–1.27) | 0.913 | 0.99(0.77–1.27) | 0.921 | 1.00(0.78–1.29) | 0.988 |
| Q4 | 1.10(0.80–1.52) | 0.567 | 1.10(0.80–1.52) | 0.557 | 1.10(0.80–1.53) | 0.559 |
Model 1: Adjusted for chronological age.
Model 2: Adjusted for chronological age and sex.
Model 3: Adjusted for chronological age, sex, family history, smoking, alcohol consumption, and BMI.
Table 3.
Associations between biological age and tumor metastasis.
| Bioage | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | |
| As continuous (per SD) | 1.02(1.01–1.03) | < 0.001 | 1.02(1.01–1.03) | < 0.001 | 1.02(1.01–1.03) | < 0.001 |
| Q1 | 0.76(0.56–1.02) | 0.065 | 0.73(0.54–0.98) | 0.038 | 0.71(0.53–0.97) | 0.029 |
| Q2 | Ref | Ref | Ref | |||
| Q3 | 1.22(0.96–1.56) | 0.112 | 1.24(0.97–1.58) | 0.089 | 1.24(0.97–1.59) | 0.083 |
| Q4 | 1.87(1.35–2.58) | < 0.001 | 1.94(1.40–2.68) | < 0.001 | 1.95(1.41–2.69) | < 0.001 |
Model 1: Adjusted for chronological age.
Model 2: Adjusted for chronological age and sex.
Model 3: Adjusted for chronological age, sex, family history, smoking, alcohol consumption, and BMI.
By contrast, when tumor metastasis was considered, a more consistent pattern emerged. In the baseline model (Model 1), biological age showed a significant association with metastasis (P < 0.05). After adjusting for chronological age and sex (Model 2), biological age demonstrated a significant independent association with metastatic disease (P < 0.05). This association remained robust after further adjustment for additional clinical and lifestyle covariates (Model 3, P < 0.05). Specifically, each 1-SD increase in biological age was associated with a 2% increase in the odds of metastasis (OR = 1.02, 95% CI: 1.01–1.03). When examined by quartiles, patients in the first quartile had a 29% reduced risk of metastasis compared with those in the second quartile (OR = 0.71, 95% CI: 0.53–0.97), while those in the fourth quartile had a markedly elevated risk (OR = 1.95, 95% CI: 1.41–2.69).
Subgroup analyses further suggested that the association between biological age and metastasis varied according to chronological age (Table 4, P for interaction < 0.05). In the subgroup of patients aged 50–64 years, those in the highest quartile of biological age had more than a twofold higher risk of metastasis compared with those in the second quartile (OR = 2.04, 95% CI: 1.23–3.40). No significant associations were identified in patients younger than 50 years or those aged ≥ 65 years.
Table 4.
Associations between biological age and tumor metastasis stratified by chronological age groups.
| Subgroup | No. of event | OR (95% CI) | P value | P for interaction |
|---|---|---|---|---|
| Age | < 0.001 | |||
| < 50 | 163/557 | |||
| Q1 | 145/511 | 0.63(0.32–1.23) | 0.175 | |
| Q2 | 15/42 | 1 | Ref | |
| Q3 | 1/2 | 1.71(0.10–29.71) | 0.714 | |
| Q4 | 2/2 | 2,518,321,677(0-NA) | 0.999 | |
| 50–64 | 410/1463 | |||
| Q1 | 58/258 | 0.73(0.52–1.03) | 0.076 | |
| Q2 | 188/697 | 1 | Ref | |
| Q3 | 135/437 | 1.23(0.94–1.60) | 0.129 | |
| Q4 | 29/71 | 2.04(1.23–3.40) | 0.006 | |
| ≥ 65 | 261/1054 | |||
| Q1 | 0/0 | NA | NA | |
| Q2 | 5//29 | 1 | Ref | |
| Q3 | 66/329 | 1.17(0.43–3.22) | 0.754 | |
| Q4 | 190/696 | 1.79(0.67–4.79) | 0.245 |
Adjusted for sex, family history, smoking, alcohol consumption, and BMI.
Taken together, these findings indicate that while biological age is not related to TNM stage, it is independently associated with tumor metastasis, particularly among middle-aged patients, underscoring its value as a predictor of disease progression beyond chronological age.
Biological age and survival outcomes
Over a median follow-up duration of 47.87 months (IQR: 29.63–64.23), 951 deaths (30.94%) were recorded. Kaplan–Meier survival curves stratified by biological age quartiles revealed clear gradients in survival probabilities (Fig. 4). Patients in quartile 3 experienced noticeably worse survival than those in quartile 2, whereas patients in quartile 4 had the lowest survival probabilities throughout follow-up.
Fig. 4.
Kaplan–Meier survival curves illustrating the impact of different biological age groups on overall survival in patients with gastrointestinal cancer.
Cox proportional hazards models adjusting for potential confounders are presented in Table 5. Compared with patients in quartile 2, those in quartile 3 had a 32% higher risk of death (HR = 1.32, 95% CI: 1.08–1.62), while those in quartile 4 had a more than twofold increased risk (HR = 2.04, 95% CI: 1.58–2.62). In addition, each 1-SD increase in biological age corresponded to a 3% higher risk of mortality (95% CI: 1.02–1.03). Restricted cubic spline (RCS) analyses confirmed these findings, revealing significant overall and nonlinear associations between biological age and mortality (both P < 0.05; Fig. 5).
Table 5.
Associations between biological age and overall survival in patients with gastrointestinal cancer.
| Bioage | Model 1 | Model 2 | Model 3 | |||
|---|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value | |
| As continuous (per SD) | 1.03(1.02–1.03) | < 0.001 | 1.03(1.02–1.04) | < 0.001 | 1.03(1.02–1.03) | < 0.001 |
| Quartile 1(n = 768) | 0.83(0.64–1.07) | 0.153 | 0.83(0.64–1.07) | 0.153 | 0.83(0.64–1.07) | 0.150 |
| Quartile 2(n = 769) | Ref | Ref | Ref | |||
| Quartile 3(n = 768) | 1.34(1.10–1.64) | 0.004 | 1.34(1.10–1.65) | 0.004 | 1.32(1.08–1.62) | 0.007 |
| Quartile 4(n = 769) | 2.22(1.73–2.85) | < 0.001 | 2.22(1.73–2.86) | < 0.001 | 2.04(1.58–2.62) | < 0.001 |
Model 1: Adjusted for chronological age.
Model 2: Adjusted for chronological age and sex.
Model 3: Adjusted for chronological age, sex, family history, TNM stage, smoking, alcohol consumption, and BMI.
Fig. 5.
Dose–response relationship between biological age and overall survival in patients with gastrointestinal cancer, adjusted for chronological age, sex, family history, TNM stage, smoking, alcohol consumption, and BMI.
Subgroup analyses are shown in Supplementary Figs. 1 and 2. No significant effect modification by chronological age, sex, BMI, or TNM stage was detected (P for interaction > 0.05). Nevertheless, when stratified by chronological age, patients in quartile 4 had substantially higher mortality risks compared with quartile 2 among those aged 50–64 years (HR = 2.59, 95% CI: 1.79–3.75) and those aged ≥ 65 years (HR = 2.81, 95% CI: 1.16–6.80), whereas no significant association was observed among patients younger than 50 years. Consistent patterns were also observed across sex, BMI, and TNM categories: quartile 4 was uniformly associated with significantly poorer prognosis in all subgroups (P < 0.05). Moreover, quartile 3 was associated with worse survival in males, individuals with lower BMI, and patients with advanced TNM stage (P < 0.05).
Sensitivity analyses supported the robustness of these findings. Exclusion of patients who died within the first 3 months of follow-up did not materially change the results (Supplementary Table 2). Furthermore, when biological age was categorized into tertiles rather than quartiles, higher biological age remained strongly associated with worse overall survival. Importantly, this simplified grouping strategy highlighted even more clearly the potential protective effects of lower biological age (Supplementary Table 3).
In summary, elevated biological age was consistently associated with increased mortality risk, independent of chronological age and other prognostic factors. These findings reinforce the role of biological age as a robust predictor of long-term survival in patients with gastrointestinal cancers.
In this large cohort of patients with gastrointestinal cancers, we observed that biological age differed significantly across chronological age groups and was closely associated with several unfavorable baseline characteristics. Although no associations were found between biological age and TNM stage, higher biological age was independently related to an increased risk of tumor metastasis, particularly among patients aged 50–64 years. Moreover, elevated biological age consistently predicted poorer overall survival, with robust and dose–response associations confirmed across multiple statistical models, subgroup analyses, and sensitivity analyses. Collectively, these findings highlight biological age as a clinically relevant marker that not only reflects patient heterogeneity at baseline but also provides independent prognostic value for disease progression and long-term outcomes in gastrointestinal cancer.
Development and validation of the prognostic model
A total of 3,074 patients were included in this study, with 450 patients from Fujian Cancer Hospital serving as an independent validation cohort, and the remaining 2,624 patients randomly assigned to a training set (n = 1,837) and a test set (n = 787) at a 7:3 ratio. In the training cohort, univariate Cox regression analysis showed that age, TNM stage, tumor metastasis, albumin, BMI, biological age (BioAge), and C-reactive protein (CRP) were all significantly associated with overall survival (OS) (P < 0.05). Further multivariate Cox regression analysis indicated that TNM stage, tumor metastasis, albumin, BMI, and biological age were independent prognostic factors for OS (Table 6).
Table 6.
Results of univariate and multivariate Cox regression analyses.
| Characteristic(n = 1837) | Univariate Cox regression analysis | Multivariate Cox regression analysis | ||
|---|---|---|---|---|
| HR(95%CL) | P value | HR(95%CL) | P value | |
| Gender | ||||
| Male | Ref | Ref | ||
| Female | 1.120 (0.934–1.345) | 0.222 | ||
| Age | 1.016 (1.008–1.024) | < 0.001 | 0.990 (0.973–1.009) | 0.306 |
| Age group, n (%) | ||||
| < 50 | Ref | Ref | ||
| 50–64 | 0.887 (0.684–1.149) | 0.362 | ||
| ≥ 65 | 1.266 (1.051–1.526) | 0.013 | ||
| Smoke, yes, n (%) | 0.964 (0.810–1.147) | 0.676 | ||
| Alcohol, yes, n (%) | 0.904 (0.738–1.107) | 0.328 | ||
| Family history, yes, n (%) | 0.888 (0.701–1.124) | 0.323 | ||
| TNM stage, n (%) | ||||
| Ⅰ | Ref | Ref | ||
| Ⅱ | 2.027 (1.066–3.852) | 0.031 | 1.942 (1.021–3.694) | 0.043 |
| Ⅲ | 3.287 (1.790–6.035) | < 0.001 | 3.233 (1.760–5.941) | < 0.001 |
| Ⅳ | 10.208 (5.575–18.691) | < 0.001 | 6.336 (3.379–11.880) | < 0.001 |
| Tumor metastasis, yes, n (%) | 3.635 (3.049–4.335) | < 0.001 | 1.923 (1.527–2.421) | < 0.001 |
| Albumin | 0.964 (0.951–0.978) | < 0.001 | 0.984 (0.969–1.000) | 0.048 |
| NLR | 1.002 (0.995–1.010) | 0.549 | ||
| BMI, kg/m2, median (IQR) | 0.930 (0.905–0.955) | < 0.001 | 0.943 (0.918–0.968) | < 0.001 |
| BMI category, n (%) | ||||
| Underweight (< 18.5 kg/m2) | Ref | Ref | ||
| Normal (18.5–23.9 kg/m2) | 1.387 (1.105–1.742) | 0.005 | ||
| Overweight (24–27.9 kg/m2) | 0.855 (0.686–1.066) | 0.164 | ||
| Obesity (≥ 28 kg/m2) | 0.350 (0.186–0.658) | 0.001 | ||
| Bioage | 1.019 (1.012–1.026) | < 0.001 | 1.022 (1.006–1.039) | 0.007 |
| CRP | 1.033 (1.011–1.054) | 0.003 | 1.013 (0.986–1.042) | 0.34 |
Nomogram Construction and Validation.
Based on the independent prognostic factors identified above, we constructed a nomogram model for individualized survival prediction (Fig. 6). The model assigns a corresponding score to each variable, integrating patients’ clinical and laboratory characteristics into a composite risk score to intuitively estimate 1-, 3-, and 5-year overall survival probabilities.
Fig. 6.
Nomogram predicting 1-, 3-, and 5-year overall survival (OS) based on TNM stage, tumor metastasis, albumin, BMI, and biological age.
In the training cohort, the nomogram achieved a C-index of 0.732. Time-dependent ROC curve analysis showed AUCs of 0.801 (95% CI, 0.766–0.835) for 1-year, 0.807 (95% CI, 0.780–0.834) for 3-year, and 0.807 (95% CI, 0.778–0.835) for 5-year survival (Supplementary Fig. 3). Calibration curves demonstrated good agreement between predicted and observed survival probabilities at all time points (Supplementary Fig. 4a–c). Decision curve analysis (DCA) in the training cohort indicated that the model provided favorable discriminative ability and reliable absolute risk estimates, yielding net benefit across commonly used threshold probabilities (Supplementary Fig. 5a–c).
In the test cohort, the nomogram achieved a C-index of 0.723. Time-dependent ROC curves showed AUCs of 0.795 (95% CI, 0.750–0.841) for 1-year, 0.784 (95% CI, 0.744–0.825) for 3-year, and 0.784 (95% CI, 0.737–0.831) for 5-year survival (Fig. 2). Calibration curves indicated close alignment between predicted probabilities and observed survival (Supplementary Fig. 4d–f), and DCA curves confirmed comparable net benefit to the training cohort (Supplementary Fig. 5d–f). These results suggest that the model maintained similar discrimination, calibration, and clinical utility in the independent test dataset.
In the external validation cohort (n = 450), the nomogram achieved a C-index of 0.687. ROC analysis showed AUCs of 0.760 (95% CI, 0.707–0.813) for 1-year, 0.844 (95% CI, 0.792–0.896) for 3-year, and 0.846 (95% CI, 0.770–0.923) for 5-year survival (Supplementary Fig. 6a). Calibration curves demonstrated overall agreement between predicted probabilities and observed outcomes (Supplementary Fig. 6b–d). DCA curves indicated that the model provided net benefit across clinically relevant threshold probabilities (Supplementary Fig. 6e–g). These findings suggest that the nomogram maintained moderate to good discriminative ability, acceptable calibration, and clinical utility across institutions, indicating its potential generalizability and applicability in practice.
Discussion
In this large cohort of patients with gastrointestinal cancers, we found that biological age differed substantially across chronological age groups and was strongly associated with several adverse host characteristics. Although no associations were detected between biological age and TNM stage, higher biological age independently predicted a greater likelihood of tumor metastasis, particularly among patients aged 50–64 years. Furthermore, elevated biological age was consistently associated with increased mortality risk in a dose–response manner, with results remaining robust across multiple models, subgroup analyses, and sensitivity analyses. Together, these findings underscore the clinical relevance of biological age as a prognostic marker that extends beyond chronological age and conventional staging systems.
In our baseline analyses, higher biological age was associated with older chronological age, male sex, smoking, systemic inflammation (reflected by elevated NLR), and poorer nutritional status (lower albumin), collectively indicating a pattern of physiological vulnerability. These findings are consistent with conceptual frameworks describing the hallmarks of aging—such as chronic inflammation, metabolic stress, and immune dysfunction—as fundamental drivers of cancer risk and progression13,14. Notably, we observed that ΔAge values were largest among patients younger than 50 years, suggesting that biological age can capture early physiologic deviations not fully explained by chronological age. This novel observation, less emphasized in previous cancer studies, implies that accelerated biological aging may already be detectable in younger individuals and could represent an early marker of vulnerability, with potential implications for prevention and risk stratification.
Interestingly, biological age was not associated with TNM stage but demonstrated a robust and independent association with metastatic disease, particularly among patients aged 50–64 years. This apparent discrepancy highlights that TNM staging primarily reflects local tumor burden, whereas metastatic spread is profoundly influenced by host systemic factors such as immune senescence, extracellular matrix remodeling, and chronic low-grade inflammation (“inflammaging”)15,16. Previous research has emphasized that aging-related changes in the tumor microenvironment can promote metastatic dissemination16,17, and our findings provide empirical support for this concept. As noted in the review “Is cancer biology different in older patients?” (The Lancet Healthy Longevity), tumour subtype distributions, molecular alteration profiles, and immune microenvironment patterns are reported to shift with increasing patient age, indicating that cancer biology differs in older versus younger individuals18, our results suggest that biological age integrates these systemic aging processes, thereby offering additional prognostic value beyond conventional staging. The stronger association in middle-aged patients further indicates that this may represent a critical window where biological aging disproportionately influences metastatic risk.
With respect to long-term prognosis, elevated biological age was consistently associated with worse overall survival, independent of chronological age, TNM stage, BMI, and other confounders. Prior studies have established biological age as a predictor of all-cause and cancer-specific mortality19,20, yet our study extends this evidence by showing its prognostic relevance in gastrointestinal cancers specifically, and by confirming its predictive ability across multiple subgroups. Notably, the non-linear dose–response relationship and the persistence of associations in sensitivity analyses reinforce the robustness of these findings. Moreover, the age-stratified analyses revealed that biological age was particularly informative for middle-aged and older patients, while its prognostic value was less pronounced in younger individuals, possibly reflecting different tumor biology or genetic susceptibility in early-onset cancers21,22.
In this study, we constructed and validated a prognostic nomogram incorporating biological age and clinical variables to predict survival in gastric and colorectal cancer patients. The model achieved favorable discrimination and calibration across training, test, and external cohorts, with C-index values above 0.68 and consistently high AUCs, demonstrating robust predictive performance and clinical applicability.Our findings emphasize the prognostic value of biological age, which reflects individual physiological status beyond chronological age. By integrating this marker, the nomogram provided more precise risk stratification, offering clinicians a practical tool for individualized survival prediction and treatment planning. External validation further confirmed its generalizability, supporting its potential use in routine practice.
Collectively, our findings suggest that biological age should be considered not merely as an adjunct to chronological age but as an independent biomarker reflecting cumulative systemic vulnerability23–25. From a clinical perspective, incorporating biological age into prognostic assessment may refine risk stratification and guide personalized management. Patients with high biological age may benefit from intensified surveillance, early systemic therapy, or interventions aimed at modifying host-related risk factors such as chronic inflammation or nutritional status26–28. Given that biological age has been shown to be modifiable through lifestyle and pharmacological interventions, future studies should evaluate whether reducing biological age can translate into improved oncological outcomes.
Several limitations merit consideration. First, our Biological Age was derived from a specific set of routine clinical biomarkers. Because our cohort lacked the requisite multi-omics data (such as DNA methylation profiles for epigenetic clocks) and specific serological inputs, it was not feasible to cross-validate or compare our results with other well-established aging models. Second, despite adjustment for multiple covariates, residual confounding from unmeasured or incomplete variables—such as specific treatment regimens (e.g., anti-inflammatory therapies) or comorbidities—cannot be ruled out. Notably, due to the substantial missingness of data regarding anti-inflammatory treatments, we were unable to evaluate their potential impact on patient survival. Due to the lack of granular data on recurrence intervals in the INSCOC cohort, we focused on metastatic status (a binary outcome) rather than disease-free survival (DFS). Future studies incorporating longitudinal recurrence data are warranted. Finally, subgroup analyses may have been underpowered in the youngest and oldest patients, limiting the generalizability of age-specific associations.
This study demonstrates that biological age is an independent predictor of survival in gastrointestinal cancers, reflecting baseline host vulnerability and providing clinically meaningful prognostic information beyond chronological age. By integrating biological age into a nomogram, we developed a reliable tool for individualized risk stratification and clinical decision-making, offering potential to enhance precision management in gastric and colorectal cancer patients.
Supplementary Information
Acknowledgements
The authors would like to thank the INSCOC project members for their substantial work on data collection and patient follow-up.
Author contributions
Jingxian Zheng performed data generation, data analysis and interpretation, and manuscript preparation; Chunhua Song and Hongxia Xu provided intellectual contribution and critically appraised the manuscript; Zengqing Guo conceived the study, designed experiments, interpreted data and prepared the manuscript. The Investigation on Nutrition Status and Clinical Outcome of Common Cancers (INSCOC) Group performed data generation.
Funding
The Fujian Provincial Natural Science Foundation Projects 2024J08271; Joint Funds for the innovation of science and Technology, Fujian province ( Grant number : 2024Y9629) ,High-level Talent Project (Category C) of Fujian Cancer Hospital (Grant No. 2024YNG02).
Data availability
All data generated or analysed during this study are included in this published article and its supplementary information files.
Declarations
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.
A list of authors and their affiliations appears at the end of the paper.
Contributor Information
Zengqing Guo, Email: guozengqing660@163.com.
The Investigation on Nutrition Status and Clinical Outcome of Common Cancers (INSCOC) Group:
Chunhua Song, Hongxia Xu, Zengqing Guo, Wang Kunhua, Fu Zhenming, Wang Chang, Weng Min, Cao Jingjing, Zhou Fuxiang, Lin Yuan, Li Suyi, Ba Yi, Yuan Kaitao, Liu Ming, Hu Wen, Zhou Lan, Ma Hu, Yao Qinghua, Cong Minghua, Li Tao, Chen Zihua, Chen Gongyan, Zhao Qingchuan, Li Zengning, Feng Changyan, He Ying, Wu Jing, Yang Jiajun, Song Xinxia, Yu Yaying, Ma Wenjun, Luo Suxia, Zheng Jin, Chen Junqiang, Luo Qi, Wang Wei, Qiao Qiugé, Shi Yongmei, Qi Yumei, Feng Yongdong, Jiang Haiping, Qiu Hong, Guan Wenxian, Chen Jiaxin, Huang He, Yu Zhen, Fang Yu, Cui Jiuwei, Li Wei, and Shi Hanping
References
- 1.Singh, A. Global burden of five major types of gastrointestinal cancer. Prz. Gastroenterol.19(3), 236–254. 10.5114/pg.2024.141834 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Gong, Y. Q. et al. Epidemiological trends of early-onset gastrointestinal cancers from 1990 to 2021 and predictions for 2036: Analysis from the global burden of disease study 2021. Ann. Med.57(1), 2555518. 10.1080/07853890.2025.2555518 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Harrold, E., Latham, A., Pemmaraju, N. & Lieu, C. H. Early-Onset GI Cancers: Rising Trends, Genetic Risks, Novel Strategies, and Special Considerations. Am. Soc. Clin. Oncol. Educ. Book.43, e398068. 10.1200/EDBK_398068 (2023). [DOI] [PubMed] [Google Scholar]
- 4.Codipilly, D. C. et al. Epidemiology and outcomes of young-onset esophageal adenocarcinoma: An analysis from a population-based database. Cancer Epidemiol. Biomarkers Prev.30(1), 142–149. 10.1158/1055-9965.EPI-20-0944 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Zhong, W. et al. Effect of younger age on survival outcomes in T1N0M0 breast cancer: A propensity score matching analysis. J. Surg. Oncol.119(8), 1039–1046. 10.1002/jso.25457 (2019). [DOI] [PubMed] [Google Scholar]
- 6.Wang, S. et al. Aging measures and cancer in the Health and Retirement Study (HRS). Nat. Commun.16(1), 5916. 10.1038/s41467-025-60913-z (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Wassel, C. L., Barrett-Connor, E. & Laughlin, G. A. Association of circulating C-reactive protein and interleukin-6 with longevity into the 80s and 90s: The Rancho Bernardo study. J. Clin. Endocrinol. Metab.95(10), 4748–4755. 10.1210/jc.2010-0473 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Mendy, A. & Mersha, T. B. Epigenetic age acceleration and mortality risk prediction in US adults. Geroscience47(4), 6029–6038. 10.1007/s11357-025-01604-x (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Rentscher, K. E. et al. Epigenetic aging in older breast cancer survivors and noncancer controls: Preliminary findings from the Thinking and Living with Cancer study. Cancer129(17), 2741–2753. 10.1002/cncr.34818 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Chung, M. et al. DNA methylation ageing clocks and pancreatic cancer risk: Pooled analysis of three prospective nested case-control studies. Epigenetics16(12), 1306–1316. 10.1080/15592294.2020.1861401 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Loh, K. et al. Exercise and epigenetic ages in older adults with myeloid malignancies. Eur. J. Med. Res.28, 180. 10.1186/s40001-023-01145-z (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Hao, M. et al. Gompertz law-based biological age (GOLD BioAge): A simple and practical measurement of biological ageing to capture morbidity and mortality risks. Adv. Sci.12(32), e01765. 10.1002/advs.202501765 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Montégut, L., López-Otín, C. & Kroemer, G. Aging and cancer. Mol. Cancer23(1), 106. 10.1186/s12943-024-02020-z (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.López-Otín, C., Pietrocola, F., Roiz-Valle, D., Galluzzi, L. & Kroemer, G. Meta-hallmarks of aging and cancer. Cell Metab.35(1), 12–35. 10.1016/j.cmet.2022.11.001 (2023). [DOI] [PubMed] [Google Scholar]
- 15.Lian, J. et al. Immunosenescence: a key player in cancer development. Aging (Albany NY).12(20), 19925–19945. 10.18632/aging.104163 (2020). [Google Scholar]
- 16.Liu, Q. et al. Immunosenescence and cancer: Molecular hallmarks, tumor microenvironment remodeling, and age-specific immunotherapy challenges. J. Hematol. Oncol.18, 81. 10.1186/s13045-025-01735-w (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Zhao, H. et al. Inflammation and tumor progression: Signaling pathways and targeted intervention. Signal Transduct. Target. Ther.6, 263. 10.1038/s41392-021-00658-5 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Van Herck, Y. et al. Is cancer biology different in older patients?. Lancet Healthy Longev.2(10), e663–e677. 10.1016/S2666-7568(21)00179-3 (2021). [DOI] [PubMed] [Google Scholar]
- 19.Zou, Y. et al. Associations between five indicators of epigenetic age acceleration and all-cause and cause-specific mortality among US adults aged 50 years and older. Clin. Epigenet.17, 66. 10.1186/s13148-025-01872-6 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Liu, X. et al. Association of phenotypic age acceleration with all-cause and cancer-specific mortality in cancer survivors. Aging (Albany NY).17(4), 1–14. 10.18632/aging.203234 (2025).39808121 [Google Scholar]
- 21.Nong, J. et al. Association between pace of biological aging and cancer risk and mortality: a nationwide cohort study. Clin. Epigenet.17(1), 1–12. 10.1186/s13148-025-01912-1 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Jeong, C. U., Leiby, J. S., Kim, D. & Choe, E. K. Artificial intelligence-driven biological age prediction model using comprehensive health checkup data: Development and validation study. JMIR Aging8, e64473. 10.2196/64473 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Bortz, J. et al. Biological age estimation using circulating blood biomarkers. Commun. Biol.6, 1089. 10.1038/s42003-023-05456-z (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Oh HSH, Lehallier B, Mormino E, Wagner AD, Henderson VW, Longo FM, et al. Wyss-Coray T (senior/corresponding author). Organ aging signatures in the plasma proteome track health and disease. Nature. Apr;624:164–172. 10.1038/s41586-023-06802-1. (2023). [DOI] [PMC free article] [PubMed]
- 25.Min, M., Egli, C., Dulai, A. S. & Sivamani, R. K. Critical review of aging clocks and factors that may influence the pace of aging. Front. Aging5, 1487260. 10.3389/fragi.2024.1487260 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Yang, Y. et al. Clinical Significance of Biological Age in Patients Undergoing Percutaneous Coronary Intervention. Mayo Clin. Proc.98(7), 1426–1435. 10.1016/j.mayocp.2023.04.002 (2023). [DOI] [PubMed] [Google Scholar]
- 27.Mandelblatt, J. S. et al. Applying a life course biological age framework to improving the care of individuals with adult cancers: Review and research recommendations. JAMA Oncol.8(5), 675–682. 10.1001/jamaoncol.2021.7228 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Long, L. et al. Association of the dietary inflammatory index with multi-organ biological age acceleration. J. Health Popul. Nutr.44(1), 343. 10.1186/s41043-025-01080-1 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
All data generated or analysed during this study are included in this published article and its supplementary information files.