Abstract
Background
While conventional risk factors for liver cancer are well-established, the contribution of biological aging to hepatocarcinogenesis remains poorly understood, representing an important gap in its etiological framework.
Methods
We conducted a prospective cohort study using data from the UK Biobank (UKB). Biological age (BA) was assessed using the Klemera–Doubal method (KDM-BA) and PhenoAge algorithms. Multivariable Cox proportional hazards models were used to evaluate associations between BA acceleration and liver cancer incidence, with robustness assessed using restricted cubic splines, subgroup analyses, and sensitivity analyses. Mediation analyses evaluated whether alanine aminotransferase (ALT) and aspartate aminotransferase (AST) mediate the relationship between biological aging and liver cancer.
Results
The study included 274,966 participants (mean age 55.75 ± 8.10 years; 53.8% female). Over a median follow-up of 14.67 years, 325 incident liver cancer cases were documented. After multivariable adjustment, each 1-standard deviation (SD) increase in BA acceleration was associated with an elevated liver cancer risk (KDM-BA acceleration: HR = 1.28, 95%CI 1.16–1.41; PhenoAge acceleration: HR = 1.32, 95%CI 1.22–1.43). A nonlinear dose–response relationship was observed between KDM-BA acceleration and liver cancer risk (P < 0.001). Exploratory mediation analyses indicated that the effect of KDM-BA acceleration may be partially mediated by ALT and AST, whereas PhenoAge acceleration appeared to act primarily through direct mechanisms.
Conclusions
Accelerated biological aging, as quantified by KDM-BA and PhenoAge, appears to be independently associated with incident liver cancer. The underlying pathways may differ: KDM-BA acceleration is partially mediated by liver enzymes, while PhenoAge acceleration appears to operate largely through direct mechanisms.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s12876-026-05103-x.
Keywords: Biological aging, Liver cancer, Cohort study, Mediation analysis
Introduction
Liver cancer constitutes a critical global health burden, representing the third most common cause of cancer-related deaths globally [1]. It is histologically categorized into hepatocellular carcinoma (HCC), intrahepatic cholangiocarcinoma (iCCA), and combined hepatocellular-cholangiocarcinoma (cHCC-CCA), with HCC accounting for roughly 80% of primary liver malignancies [2]. Known risk factors include chronic hepatitis B or C virus infection, alcohol-associated liver disease, metabolic dysfunction, and genetic predispositions [3]. Nonetheless, approximately 20%−30% of cases arise in individuals lacking these conventional risk factors, suggesting the involvement of understudied pathogenic mechanisms [4]. This heterogeneity, combined with suboptimal early detection strategies, contributes to late diagnosis and dismal outcomes: the five-year survival rate drops from over 70% in early-stage cases to under 10% in advanced cancer [5]. These challenges highlight the urgency of identifying novel predictive biomarkers, particularly those capturing biological processes such as accelerated aging and subclinical hepatic injury, which could help clarify unexplained risk pathways and facilitate earlier intervention in an increasingly aging population.
Aging is increasingly recognized as a key contributor to oncogenesis [6–8]. Unlike chronological age (CA), which merely quantifies elapsed time, biological age (BA) measures functional decline at molecular and cellular levels, providing superior predictive value for physiological health and disease susceptibility [9, 10]. Widely used biomarkers of aging include telomere length, frailty indices, DNA methylation clocks, and clinical biomarker composites, each capturing distinct facets of the aging process [11]. Among them, the Klemera-Doubal method biological age (KDM-BA) [12] and PhenoAge [13] algorithms have demonstrated strong predictive validity for mortality, age-related disease incidence, and functional decline, independent of CA. Since various BA measures assess diverse biological pathways [14, 15], it is crucial to elucidate their differential associations with liver cancer risk to uncover the specific mechanisms that link aging to hepatocarcinogenesis. In particular, liver enzymes such as alanine aminotransferase (ALT) and aspartate aminotransferase (AST) are not only key indicators of hepatic injury but may also act as mediators between accelerated biological aging and liver cancer initiation. Elevated ALT and AST levels reflect ongoing hepatocellular damage [16], which can promote fibrosis and create a procarcinogenic microenvironment, thereby potentially bridging accelerated aging and liver cancer.
Although accelerated epigenetic aging has been linked to pre-malignant hepatic conditions, such as hepatic fibrosis and cirrhosis [17], direct epidemiological evidence connecting BA with incident liver cancer remains limited. Moreover, while BA has been implicated in the development of multiple cancers [18], large-scale, prospective studies specifically examining its relationship with liver cancer risk and the potential mediating roles of ALT and AST are still lacking.
To address these gaps, this study utilized the comprehensive, multidimensional data from the UK Biobank (UKB) to examine the association between two well-validated biological aging measures (KDM-BA and PhenoAge) and the incidence of liver cancer. We further conducted formal mediation analyses to assess whether ALT and AST mediate the relationship between BA and liver cancer risk. Our study aims to enhance understanding of the interplay between biological aging and hepatocarcinogenesis, with implications for refined risk prediction and targeted prevention strategies in an aging population.
Methods and materials
Data source and study population
This study utilized data from the UKB, a prospective population-based cohort comprising over 500,000 participants aged 37–73 years who were recruited between 2006 and 2010. All participants completed an electronic questionnaire, a verbal interview, physical examinations, and provided biological samples at one of 22 assessment centers across England, Scotland, and Wales [19]. The UKB study was approved by the North West Multi-centre Research Ethics Committee (MREC; REC reference: 21/NW/0157), and written informed consent was obtained from every participant at the time of enrollment.
From the initial cohort of 502,370 participants in the UKB, we excluded 200,116 individuals with missing data on BA algorithms, 20,658 with a pre-existing cancer diagnosis (excluding non-melanoma skin cancer), and 6630 with incomplete data on mediators or covariates, resulting in a final analytical sample of 274,966 participants. A detailed participant recruitment flowchart is provided in Supplemental Figure S1. A comparison of baseline characteristics between included (n = 274,966) and excluded (n = 227,404) participants is shown in Supplementary Table S1.
Assessment of biological age and biological age acceleration
Biological age (BA) was assessed using two validated algorithms: KDM-BA and PhenoAge, which were based on blood-chemistry-derived measures. This approach provided a more comprehensive evaluation of biological aging. The KDM-BA algorithm was derived from a linear regression model incorporating nine clinical biomarkers [20]. In contrast, the PhenoAge was constructed from a Cox proportional hazards model predicting mortality risk using nine clinical biomarkers plus CA [21, 22]. Both measures were computed according to their original published algorithms. BA acceleration was quantified as the residuals of KDM-BA or PhenoAge after adjusting for CA via restricted cubic splines with 3 degrees of freedom [18]. For analytical purposes, BA acceleration was dichotomized based on its distribution relative to CA. Individuals with a BA acceleration > 0 were classified as exhibiting accelerated biological aging, meaning their BA exceeded their CA. While those with BA acceleration ≤ 0 were classified as exhibiting non-accelerated biological aging [23]. Both BA and BA acceleration metrics were computed using the R package BioAge. The corresponding field codes for clinical biomarkers are provided in Supplementary Table S2.
Assessment of outcome
In this study, the primary outcome was the first primary incident of liver cancer. Liver cancer cases were ascertained through record linkage to the cancer registries in England, Wales, and Scotland, which provide complete coverage of cancer diagnoses and mortality. We identified cancer diagnoses using the International Classification of Diseases, 10th Revision (ICD-10) codes for malignant neoplasms of the liver and intrahepatic bile ducts (C22). This included the following specific codes: C22.0 (liver cell carcinoma), C22.1 (intrahepatic bile duct carcinoma), C22.3 (angiosarcoma of liver), C22.4 (other sarcomas of liver), C22.7 (other specified carcinomas of liver), and C22.9 (liver, unspecified). Participants were followed from baseline until the earliest occurrence of the following events: first liver cancer diagnosis, death, or the end of the study period.
Assessment of mediators
The mediating variables, specifically serum alanine aminotransferase (ALT; Field ID 30620) and aspartate aminotransferase (AST; Field ID 30650), were measured from blood samples collected at the baseline assessment of the UKB.
Measurements of covariates
We systematically adjusted for potential confounders across three domains: sociodemographic characteristics, lifestyle factors, and pre-existing chronic diseases. The following covariates were collected through a touchscreen questionnaire: age (year), sex (male or female), ethnicity (White, Asian, Black, Mixed background or other), education level (≥ college/university degree or < college/university degree), household income (< £18,000, £18,000-£51,999, ≥ £52,000, and unknown, pound/year), Townsend deprivation index (TDI), smoking status (never smoker, previous smoker, current smoker), drinking status (never drinker, previous drinker, current drinker), sun protection (never, sometimes, most of the time, always, other), physical activity (low, moderate, high), obesity (yes or no), hypertension (yes or no), diabetes (yes or no) and family history of cancer (yes or no). Hypertension and diabetes were self-reported physician diagnoses. The TDI, derived from unemployment, overcrowding, non-car ownership, and non-home ownership, serves as a composite measure of socioeconomic deprivation. Higher TDI values represent greater socioeconomic disadvantage. Family history of cancer was defined as having at least one first-degree relative (parent, sibling, or offspring) diagnosed with any type of cancer. Obesity was defined as a body mass index (BMI) ≥ 30 kg/m2, and BMI was calculated using the formula: weight (kg)/[height (m)]2. Physical activity was assessed using the International Physical Activity Questionnaire (IPAQ) and quantified as metabolic equivalent of task (MET, minutes/week). For analysis, participants were categorized into tertiles based on their MET values: low (< 230), moderate (230–840), and high (≥ 840). The field IDs for all covariates are provided in Supplemental Table S3.
Statistical analysis
Person-time accrual started at baseline assessment and continued until the earliest of: diagnosis of first primary liver cancer, death from any cause, or the end of the study period. Participant characteristics were stratified by sex and presented using descriptive statistics. Continuous variables were expressed as mean ± standard deviation (SD) for normally distributed data or median (P25, P75) for non-normally distributed variables. Categorical variables were summarized as frequency counts (percentages). Continuous variables were analyzed using Student's t-test or Mann–Whitney U test, as appropriate. Categorical variables were assessed using Chi-square tests. We systematically evaluated bivariate associations among CA, BA, and BA acceleration using Pearson correlation analysis.
We employed multivariable Cox proportional hazards models to evaluate the associations between BA acceleration and liver cancer risk. Two analytical approaches were implemented: 1) treating BA acceleration as a continuous variable (per 1-SD increase), and 2) dichotomizing BA acceleration into accelerated (> 0) versus non-accelerated (≤ 0) groups for categorical comparison. The results were presented as HRs with 95% confidence intervals (CIs). The proportional-hazard assumptions were verified using Schoenfeld's test (Supplemental Tables S4 and S5). Model 1 was a crude model. Model 2 adjusted for age, sex, ethnicity, and education level. Model 3 additionally adjusted for TDI, smoking status, drinking status, sun protection, physical activity, obesity, hypertension, diabetes, and family history of cancer. Prior to constructing the multivariable Cox proportional hazards models, we assessed multicollinearity among variables using generalized variance inflation factors (GVIF). All variables demonstrated GVIF values < 5 (range: 1.026–1.278), indicating no need to exclude any variables (Supplemental Table S6).
To explore potential nonlinear dose–response relationships between BA acceleration and liver cancer risk, we performed restricted cubic spline analyses (4 knots at the 5th, 35th, 65th, and 95th percentiles), with full adjustment for all covariates in Model 3. We formally compared the nonlinear spline model against a linear model using likelihood ratio tests and the Akaike Information Criterion (AIC).
Subgroup analyses were carried out, with stratification based on sex (male or female), age (< 60 or ≥ 60 years), ethnicity (White British or other), obesity (BMI ≥ 30 kg/m2 or < 30 kg/m2), smoking status (current smoker or other), drinking status (current drinker or other). To validate the robustness of our findings, we conducted four complementary sensitivity analyses. First, we analyzed the associations between continuous BA metrics and liver cancer risk in Cox regression models. Second, we excluded those diagnosed with liver cancer within the initial 2 years of follow-up to minimize the impact of reverse causality. Third, we removed individuals with a family history of cancer to reduce the potential influence of genetic factors. Fourth, we excluded extreme BA acceleration outliers (defined as ± 3 SD from the mean) to minimize potential distortion of the observed associations. Fifth, to address competing risk from non-liver cancer death, we conducted a sensitivity analysis using the Fine-Gray model. Sixth, to address potential confounding by liver disease, we conducted a sensitivity analysis further adjusting for a chronic liver disease (CLD) indicator (ICD-10 codes K70-K77) from UKB hospital records. The results were compared to the primary analysis (Model 3).
Additionally, to assess potential unmeasured confounding in the association between BA acceleration (binary) and liver cancer risk, we computed E-values as part of a sensitivity analysis.
To explore the potential mediating roles of ALT and AST in the associations between BA acceleration (binary) and incident liver cancer, we conducted exploratory mediation analyses using bootstrap methods (5000 resamples). Both mediators were simultaneously included in a single model to estimate mediator-specific indirect effects, direct effects, and total effects. All analyses were adjusted for the complete set of confounders specified in Model 3.
All analyses were conducted using R version 4.2.0 with two-tailed tests at an α = 0.05 significance level.
Results
Baseline characteristics and biological aging metrics
Among the 274,966 UKB participants included in the study, 127,016 (46.2%) were males, and 147,950 (53.8%) were females, with a mean baseline age of 55.75 years (SD = 8.10). Most baseline characteristics differed significantly between sexes. Males exhibited higher rates of current smoking, current alcohol consumption, diabetes, and hypertension, as well as a higher mean BMI, whereas females were more likely to report regular use of sun protection. Sex-based differences were also observed in measures of biological aging. During a median follow-up of 14.67 years (IQR: 14.06–15.40), 212 (0.17%) incident liver cancer cases were identified in males and 113 (0.08%) in females. Full baseline characteristics are presented in Table 1.
Table 1.
Characteristics of the study participants
| Characteristic | Total (n = 274,966) | Male (n = 127,016) | Female (n = 147,950) | P-value |
|---|---|---|---|---|
| Covariates | ||||
| Age, years, mean (SD) | 55.75 (8.10) | 55.81 (8.21) | 55.70 (8.00) | <.001 |
| Ethnicity, white, n (%) | 261,096 (95.0) | 120,419 (94.8) | 140,677 (95.1) | <.001 |
| College/university degree, n (%) | 93,976 (34.2) | 45,606 (35.9) | 48,370 (32.7) | <.001 |
| Current smoker, n (%) | 27,927 (10.2) | 15,532 (12.2) | 12,395 (8.4) | <.001 |
| Current drinker, n (%) | 255,795 (93.0) | 120,088 (94.5) | 135,707 (91.7) | <.001 |
| BMI, kg/m2, mean (SD) | 27.10 (4.54) | 27.55 (4.02) | 26.71 (4.92) | <.001 |
| Physical activity, high, n (%) | 116,646 (42.4) | 56,581 (44.5) | 60,065 (40.6) | <.001 |
| Sun protection, always, n (%) | 55,725 (20.3) | 16,627 (13.1) | 39,098 (26.4) | <.001 |
| TDI, median (P25, P75) | −2.25 (−3.70, 0.26) | −2.25 (−3.70, 0.33) | −2.25(−3.68, 0.21) | 0.046 |
| Diabetes, n (%) | 5403 (2.0) | 3445 (2.7) | 1958 (1.3) | <.001 |
| Hypertension, n (%) | 58,495 (21.3) | 31,732 (25.0) | 26,763 (18.1) | <.001 |
| Family history of cancer, n (%) | 79,606 (29.0) | 36,892 (29.0) | 42,714 (28.9) | 0.314 |
| Mediator variables | ||||
| ALT, U/L, mean (SD) | 23.24 (13.82) | 27.29 (14.87) | 19.77 (11.81) | <.001 |
| AST, U/L, mean (SD) | 25.96 (9.86) | 28.05 (10.65) | 24.18 (8.72) | <.001 |
| Biological aging metrics | ||||
| KDM-BA, years, mean (SD) | 55.73 (8.25) | 55.80 (8.24) | 55.68 (8.25) | <.001 |
| PhenoAge, years, mean (SD) | 49.66 (9.07) | 50.27 (9.46) | 49.13 (8.69) | <.001 |
| KDM-BA acceleration, years, mean (SD) | −0.01 (1.53) | −0.01 (0.62) | −0.02 (2.00) | <.001 |
| PhenoAge acceleration, years, mean (SD) | −0.04 (4.15) | 0.51 (4.00) | −0.51 (4.21) | <.001 |
| Components of biological age | ||||
| FEV1, L, mean (SD) | 2.87 (0.80) | 3.37 (0.76) | 2.43 (0.54) | <.001 |
| SBP, mm Hg, mean (SD) | 138.8 (19.41) | 142.28 (18.19) | 136.15 (19.97) | <.001 |
| Total Cholesterol, mg/dL, mean (SD) | 222.97 (42.82) | 216.65 (42.06) | 228.39 (42.72) | <.001 |
| Glycated hemoglobin, %, mean (SD) | 5.40 (0.53) | 5.42 (0.59) | 5.38 (0.47) | <.001 |
| Blood urea nitrogen, mg/dL, mean (SD) | 14.96 (3.61) | 15.51 (3.64) | 14.48 (3.51) | <.001 |
| Lymphocyte, %, mean (SD) | 29.12 (7.28) | 28.22 (7.26) | 29.90 (7.21) | 0.834 |
| Mean cell volume, fL, mean (SD) | 82.79 (5.24) | 82.63 (5.22) | 82.93 (5.26) | 0.033 |
| Serum glucose, mg/dL, mean (SD) | 90.94 (19.10) | 91.77 (21.58) | 90.22 (16.65) | <.001 |
| Red cell distribution width, %, mean (SD) | 13.44 (0.93) | 13.39 (0.79) | 13.49 (1.03) | <.001 |
| White blood cell, 1000 cells/uL, mean (SD) | 6.80 (1.80) | 6.81 (1.84) | 6.79 (1.76) | 0.118 |
| Albumin, g/dL, mean (SD) | 4.53 (0.26) | 4.57 (0.26) | 4.50 (0.26) | 0.385 |
| Creatinine, mg/dL, mean (SD) | 0.81 (0.17) | 0.92 (0.16) | 0.72 (0.12) | <.001 |
| C-reactive protein, mg/dL, mean (SD) | 0.24 (0.39) | 0.22 (0.38) | 0.25 (0.40) | <.001 |
| Alkaline phosphatase, U/L, mean (SD) | 82.26 (24.81) | 81.04 (22.82) | 83.31 (26.37) | <.001 |
| Outcomes | ||||
| Incident liver cancer, n (%) | 325 (0.12) | 212 (0.17) | 113 (0.08) | <.001 |
Data were presented as means (standard deviation, SD), median (P25, P75), and n (%)
KDM-BA, Klemera-Doubal method biological age. TDI, Townsend deprivation index. FEV1, forced expiratory volume in 1 s. ALT, Alanine aminotransferase. AST, Aspartate aminotransferase. SBP, systolic blood pressure
The correlations between the biological aging measures and CA in the UKB cohort are shown in Supplemental Figure S2. As expected, both KDM-BA and PhenoAge showed strong positive correlations with CA (KDM-BA: r = 0.98; PhenoAge: r = 0.89; both P < 0.001). A high correlation was also observed between KDM-BA and PhenoAge (r = 0.89, P < 0.001). After residualizing for CA, KDM-BA acceleration and PhenoAge acceleration showed a modest yet statistically significant correlation (r = 0.25, P < 0.001).
Biological aging and liver cancer incidence
After full adjustment for sociodemographic characteristics, lifestyle factors, and pre-existing chronic diseases in Model 3, accelerated biological aging was significantly associated with increased liver cancer risk across all analyses. Participants with accelerated biological aging exhibited significantly elevated risks of liver cancer compared to those without accelerated aging. Specifically, KDM-BA acceleration showed an HR of 1.48 (95% CI: 1.17–1.88), while PhenoAge acceleration showed an HR of 1.55 (95% CI: 1.23–1.96), as detailed in Table 2. Similarly, for each 1-SD increase in continuous BA acceleration measures, the risks remained significant (KDM-BA acceleration: HR = 1.28, 95% CI = 1.16–1.41; PhenoAge acceleration: HR = 1.32, 95% CI = 1.22–1.43), as presented in Table 3. Individual components of the biological age metric and their associations with liver cancer risk are shown in Supplementary Table S7.
Table 2.
Association between BA acceleration (binary) and liver cancer risk
| Liver cancer | KDM-BA acceleration | PhenoAge acceleration | ||
|---|---|---|---|---|
| HR (95% CI) | P-value | HR (95% CI) | P-value | |
| Model 1 | 1.93 (1.54–2.42) | <.001 | 2.06 (1.65–2.58) | <.001 |
| Model 2 | 1.83 (1.46–2.30) | <.001 | 1.84 (1.46–2.30) | <.001 |
| Model 3 | 1.48 (1.17–1.88) | 0.001 | 1.55 (1.23–1.96) | <.001 |
Model 1 was crude model
Model 2 adjusted for age, sex, ethnicity, education level
Model 3 additionally adjusted for TDI, smoking status, drinking status, sun protection, physical activity, obesity, hypertension, diabetes and family history of cancer
KDM-BA, Klemera–Doubal method-biological age. CI, Confidence Interval. Ref, reference. HR, Hazard Ratio
Table 3.
Association between BA acceleration (continuous) and liver cancer risk
| Liver cancer | KDM-BA acceleration | PhenoAge acceleration | ||
|---|---|---|---|---|
| HR per 1 SD increase (95% CI) | P-value | HR per 1 SD increase (95% CI) | P-value | |
| Model 1 | 1.31 (1.21–1.43) | <.001 | 1.42 (1.33–1.51) | <.001 |
| Model 2 | 1.37 (1.27–1.48) | <.001 | 1.40 (1.31–1.50) | <.001 |
| Model 3 | 1.28 (1.16–1.41) | <.001 | 1.32 (1.22–1.43) | <.001 |
Model 1 was crude model
Model 2 adjusted for age, sex, ethnicity, education level
Model 3 additionally adjusted for TDI, smoking status, drinking status, sun protection, physical activity, obesity, hypertension, diabetes and family history of cancer
KDM-BA, Klemera–Doubal method-biological age. CI, Confidence Interval. Ref, reference. HR, Hazard Ratio
Moreover, the analysis revealed distinct dose–response patterns between BA acceleration measures and liver cancer risk. For KDM-BA acceleration, the nonlinear restricted cubic spline model provided a significantly better fit than the linear model (likelihood ratio test: χ2 = 15.4, P < 0.001; ΔAIC = −11.4), supporting a nonlinear association (P for overall < 0.001, P for nonlinearity < 0.001). In contrast, PhenoAge acceleration demonstrated a significant linear association without significant nonlinear components (P for overall < 0.001, P for non-linearity = 0.429), as shown in fig 1.
Fig. 1.
Dose–response relationships between biological age acceleration and risk of liver cancer. Panels: HR for KDM-BA acceleration (A), and HR for PhenoAge acceleration (B). Data were fitted by Cox regression models based on restricted cubic splines (knots at the 5th, 35th, 65th, and 95th percentiles). They were adjusted for age, sex, ethnicity, education level, TDI, smoking status, drinking status, sun protection, physical activity, obesity, hypertension, diabetes and family history of cancer. KDM-BA, Klemera–Doubal method-biological age
Subgroup analysis of biological aging and liver cancer incidence
Subgroup analyses revealed a significant interaction between PhenoAge acceleration (binary) and sex (Pinteraction = 0.026). Despite this modifying effect, the BA acceleration measures remained consistently and significantly associated with increased risks of liver cancer in the majority of subgroups, as illustrated in figs 2 and 3.
Fig. 2.
Subgroup analyses for the associations between KDM-BA acceleration (binary) and risk of liver cancer. Adjusted for age, sex, ethnicity, education level, TDI, smoking status, drinking status, sun protection, physical activity, obesity, hypertension, diabetes and family history of cancer. KDM-BA, Klemera–Doubal method-biological age. CI, Confidence Interval. HR, Hazard Ratio
Fig. 3.
Subgroup analyses for the associations between PhenoAge acceleration (binary) and risk of liver cancer. Adjusted for age, sex, ethnicity, education level, TDI, smoking status, drinking status, sun protection, physical activity, obesity, hypertension, diabetes and family history of cancer. KDM-BA, Klemera–Doubal method-biological age. CI, Confidence Interval. HR, Hazard Ratio
Sensitivity analyses of biological aging and liver cancer incidence
Comprehensive sensitivity analyses confirmed the robustness of the association between biological aging and primary incident liver cancer. The findings remained consistent across all sensitivity analyses: (1) Continuous BA measures showed persistent risk associations (Supplemental Table S8); (2) After excluding incident cases within the initial 2-year follow-up period to mitigate reverse causality, the associations remained significant (n = 274,935, Supplemental Table S9); (3) After excluding participants with a family history of cancer to minimize genetic confounding, the results stayed robust (n = 195,347, Supplemental Table S10); (4) The exclusion of extreme BA outliers (defined as ± 3 SD from the mean) did not substantially alter the risk estimates, indicating robustness to potential outlier effects (n = 274,565, Supplemental Table S11). (5) Competing risk sensitivity analyses using Fine-Gray models did not substantially alter the risk estimates, indicating robustness to competing risks from non-liver cancer death (Supplemental Table S12). To assess robustness to confounding by chronic liver disease (CLD), we performed a sensitivity analysis further adjusting for a CLD indicator. The sample size was reduced to 228,632 because CLD status was derived from hospital records, which were not available for all participants. The associations were attenuated but remained statistically significant (n = 228,632, Supplemental Table S13). For the associations between KDM-BA acceleration and liver cancer risk, the E-value was 2.32 (95% CI lower bound: 1.61), while for PhenoAge acceleration it was 2.47 (95% CI lower bound: 1.76). Although these E-values exceed the threshold for confounding by typical lifestyle factors, they are substantially lower than the effect sizes reported for major liver cancer risk factors such as HBV/HCV infection or cirrhosis. Therefore, residual confounding by these major risk factors cannot be ruled out. The E-values should be interpreted cautiously, and the observed associations may still be influenced by unmeasured or imperfectly measured confounders.
Mediation analysis of ALT and AST
Correlations between BA acceleration, ALT, AST, and liver cancer are presented in Supplemental Table S14. All correlation coefficients were statistically significant (P < 0.01), consistent with our pre-specified hypotheses. Mediation analyses suggested that accelerated KDM-BA was associated with a 37.5% higher risk of liver cancer, with 5.3% of this total association jointly accounted for by ALT and AST. For accelerated PhenoAge, the association with liver cancer risk was 39.5%. In these exploratory analyses, a negative indirect effect via lower ALT (−1.3%) and a positive indirect effect via higher AST (0.4%) were observed, resulting in a net negligible mediation. These exploratory findings should be interpreted with caution, as the cross-sectional measurement of mediators at baseline precludes strong causal inference. Complete mediation results are shown in fig 4 and Supplemental Table S15.
Fig. 4.
Mediation models for KDM acceleration, PhenoAge acceleration, ALT, AST, and liver cancer. Panels: Mediation models for KDM acceleration, ALT, AST and liver cancer (A), and Mediation models for PhenoAge acceleration, ALT, AST, and liver cancer (B). Note: *p < 0.01 (two-tailed). KDM-BA, Klemera–Doubal method-biological age. ALT, Alanine aminotransferase. AST, Aspartate aminotransferase
Discussion
To our knowledge, this study provides the first prospective evidence that accelerated biological aging, quantified by clinical biomarker-based algorithms (KDM-BA and PhenoAge), appears to be associated with an elevated risk of incident liver cancer, independent of CA, sociodemographic characteristics, lifestyle factors, and pre-existing chronic conditions. The observed associations remained robust across all subgroup and sensitivity analyses. Furthermore, a nonlinear dose–response relationship was observed between KDM-BA acceleration and liver cancer risk. Notably, mediation analyses further revealed distinct pathological pathways: KDM-BA acceleration partially exerted its effect through liver injury markers (ALT and AST), whereas PhenoAge acceleration influenced liver cancer risk predominantly via direct, extra-hepatic mechanisms.
These findings should be interpreted within the context of an aging global population and the established biological link between aging and chronic disease risk. Although aging is a well-established risk factor for hepatic impairment and cancer [24, 25], and previous studies have connected epigenetic aging markers with pre-malignant conditions such as liver fibrosis and cirrhosis [17], large-scale prospective evidence directly linking biological aging to incident liver cancer has been lacking. Furthermore, while accelerated BA measures have been linked to all-cause and cancer-related mortality [22, 26], as well as elevated risks of overall cancer, lung cancer, and colorectal cancer [18], the specific pathways of different BA metrics in hepatocarcinogenesis have not been systematically assessed. Currently, the absence of a gold standard for assessing BA remains a challenge in aging research. Existing BA metrics each capture distinct aspects of the aging process [14, 15]. Given these considerations, we simultaneously employed two well-established yet independent metrics, KDM-BA and PhenoAge, to capture divergent yet complementary dimensions of the aging process, thereby providing a more holistic view of its role in liver cancer development.
Our findings are consistent with and extend previous epidemiological evidence, confirming that accelerated biological aging is significantly associated with an elevated risk of liver cancer [27]. At the mechanistic level, inflammation and metabolism-related aging biomarkers, such as the Senescence-Associated Secretory Phenotype (SASP), have been demonstrated to promote malignant transformation of hepatocytes [28]. Clinically, cellular senescence has been found to be significantly elevated in patients who developed HCC [29], underscoring its potential role as a driver in hepatocarcinogenesis. Furthermore, a previous study has shown that advanced BA quantified by clinical biomarkers was associated with increased risks of multiple cancers [18]. These findings are consistent with our results, further supporting the role of biological aging in oncogenesis.
Accelerated biological aging may significantly increase the risk of liver cancer incidence through complex and multifactorial mechanisms [30, 31]. Firstly, a central pathway involves cellular senescence, which exerts a dual role in carcinogenesis. While initially serving as a tumour-suppressive mechanism by arresting damaged cells, the persistent accumulation of senescent cells leads to continuous secretion of the SASP [32, 33]. SASP comprises a diverse array of cytokines, growth factors, and proteases, which not only induce chronic inflammation but also remodel the hepatic microenvironment, thereby fostering tumor initiation and progression [32]. Secondly, with advancing age, elevated levels of chronic inflammation and diminished immune surveillance collectively result in a decline in the body's capacity to eliminate aberrant cells [30, 31]. Dysfunction of immune cells, such as Kupffer cells, further exacerbates hepatic inflammation, thereby fueling a favorable environment for tumor cell growth [34]. Furthermore, aging is characterised by a progressive increase in genomic instability, mitochondrial dysfunction, and metabolic disturbances. These factors act in concert to accelerate the progression from non-alcoholic fatty liver disease (NAFLD) to non-alcoholic steatohepatitis (NASH) and ultimately to HCC [35]. The specificity of BA algorithms: KDM-BA and PhenoAge may specifically capture pathways of metabolic dysregulation and chronic inflammation, respectively, which aligns closely with the established mechanisms of HCC pathogenesis.
Our analysis revealed distinct dose–response patterns between two key BA acceleration measures and liver cancer risk, reflecting the distinct aging dimensions captured by different BA measures, as well as the varying impacts of these dimensions on hepatocarcinogenesis. Specifically, the strictly linear association observed for PhenoAge acceleration implies a constant, proportional increase in risk with each unit increase in acceleration. In contrast, KDM-BA acceleration exhibited a nonlinear association, suggesting the existence of a potential threshold beyond which the risk escalates precipitously. Biologically, this threshold likely may represent a stage where accelerated aging contributes to liver cancer through mechanisms such as genomic instability, telomere shortening, epigenetic alterations, increased inflammation, and impaired immune function [25, 32, 36].
The algorithm-specific mediation patterns provide exploratory, hypothesis-generating insights into potential pathways linking biological aging to liver cancer. For KDM-BA acceleration, the small proportion mediated by ALT and AST may suggest a role of subclinical hepatic injury, but an alternative interpretation is equally plausible: ALT and AST may simply be markers of pre-existing subclinical liver disease rather than true causal intermediates. For PhenoAge acceleration, the net indirect effect was close to zero, indicating that the association is almost entirely direct, possibly reflecting systemic aging processes independent of liver damage biomarkers. Given the cross-sectional measurement of mediators, these findings should be interpreted with caution and considered hypothesis-generating rather than causal. Future longitudinal studies with direct liver health assessments are needed to validate these exploratory observations.
From a public health perspective, incorporating biological age measures could improve risk stratification and early detection strategies for liver cancer. Specifically, individuals exhibiting accelerated aging may benefit from enhanced surveillance, even in the absence of traditional risk factors. Furthermore, our study suggests that biological aging may represent a target for preventive interventions, should future studies demonstrate that slowing aging processes reduces cancer risk.
Notwithstanding its strengths, such as the large sample size, reliable assessment of biological aging, and extensive availability of covariates, several limitations of this study must be considered. First, the observational design of this study limits our ability to establish causality between biological aging and liver cancer. Second, specific HBV/HCV serology data were unavailable. As a proxy, we adjusted for chronic liver disease (CLD) using ICD-10 codes K70–K77 derived from hospital inpatient records. However, this approach does not fully capture viral hepatitis, and carriers without coded cirrhosis or chronic liver disease would be missed. Given these limitations, residual confounding by viral hepatitis cannot be excluded. Third, the broad C22 outcome definition encompasses heterogeneous liver cancer subtypes, which may have distinct etiologies and risk factor profiles. Due to the limited number of liver cancer events, we were unable to perform subtype-specific analyses to explore potential heterogeneous associations across different liver cancer subtypes. This modest event count also restricts statistical power for subgroup and interaction analyses and increases the risk of overfitting in secondary analyses. Fourth, as UKB participants are predominantly of European ancestry, the generalizability of our findings to other ethnic and racial populations is limited. Fifth, selection bias is inevitable due to data availability constraints, as biological aging data were available for only a subset of the UKB. Sixth, time-varying changes in exposures and covariates were not accounted for, potentially leading to underestimated associations. Despite these limitations, our study strengthens the epidemiological evidence linking accelerated biological aging to an increased risk of liver cancer. Future research will be needed to validate these findings in diverse cohorts, particularly in high-risk populations such as individuals with liver cirrhosis. Furthermore, integrating multi-omics data is essential to elucidate the precise molecular mechanisms through which accelerated biological aging contributes to liver cancer development.
Conclusions
In conclusion, accelerated biological aging, as measured by KDM-BA and PhenoAge, may be independently associated with an increased risk of liver cancer. The distinct mediation pathways may suggest that KDM-BA and PhenoAge capture complementary mechanisms of hepatocarcinogenesis, potentially involving both hepatic and extra-hepatic aging processes. These findings collectively suggest the potential of biological aging metrics to refine liver cancer risk prediction and inform future prevention strategies.
Supplementary Information
Acknowledgements
This study is conducted under application number 98124 for UK Biobank Resource. The authors gratefully thank all the participants and professionals contributing to the UK Biobank.
Abbreviations
- UKB
UK Biobank
- HCC
Hepatocellular carcinoma
- iCCA
Intrahepatic cholangiocarcinoma
- cHCC-CCA
Combined hepatocellular-cholangiocarcinoma
- HBV
Hepatitis B virus
- HCV
Hepatitis C virus
- BA
Biological age
- CA
Chronological age
- KDM-BA
Klemera-Doubal method biological age
- ALT
Alanine aminotransferase
- AST
Aspartate aminotransferase
- IPAQ
International Physical Activity Questionnaire
- SD
Standard deviation
- HR
Hazard ratio
- CI
Confidence interval
- MREC
Multi-centre Research Ethics Committee
- TDI
Townsend deprivation index
- MET
Metabolic equivalent task
- FEV1
Forced expiratory volume in 1 s
- SBP
Systolic blood pressure
- CLD
Chronic liver disease
- SASP
Senescence-associated secretory phenotype
- NASH
Non-alcoholic steatohepatitis
- NAFLD
Non-alcoholic fatty liver disease
Authors’ contributions
Conceptualization: Xiaojuan Liu and Jiangping Li. Software: Shuangshuang Yang and Conghui Zhang. Formal analysis: Xiaojuan Liu and Shuangshuang Yang. Writing—Original Draft: Xiaojuan Liu. Proposed critical revisions to the manuscript: Jiangping Li. Funding acquisition: Xiaojuan Liu. Supervision: Jiangping Li and Ping Chen. All authors contributed to the interpretation of the results, review, and approval of the final manuscript.
Funding
This work was supported by the Natural Science Foundation of Ningxia Province of China (2022AAC03174).
Data availability
The UK Biobank is an open-access resource, available at https://www.ukbiobank.ac.uk/researchers/, and can be obtained by submitting a data request proposal. The data that support the findings of this study were used under license for the current study (Application 98,124), and thus are not publicly available.
Declarations
Ethics approval and consent to participate
UK Biobank data has approval from the North West Multi-centre Research Ethics Committee (MREC) (REC reference: 21/NW/0157). All data collection and research use is in accordance with the Declaration of Helsinki.
All participants provided written informed consent at the time of recruitment to the UK Biobank.
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.
Contributor Information
Ping Chen, Email: chenping991232@163.com.
Jiangping Li, Email: lijp@nxmu.edu.cn.
References
- 1.Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2018;68(6):394–424. [DOI] [PubMed] [Google Scholar]
- 2.Rumgay H, Ferlay J, de Martel C, Georges D, Ibrahim AS, Zheng R, et al. Global, regional and national burden of primary liver cancer by subtype. Eur J Cancer. 2022;161:108–18. [DOI] [PubMed] [Google Scholar]
- 3.Rinella ME, Lazarus JV, Ratziu V, Francque SM, Sanyal AJ, Kanwal F, et al. A multisociety Delphi consensus statement on new fatty liver disease nomenclature. Ann Hepatol. 2024;29(1):101133. [DOI] [PubMed] [Google Scholar]
- 4.Llovet JM, Kelley RK, Villanueva A, Singal AG, Pikarsky E, Roayaie S, et al. Hepatocellular carcinoma. Nat Rev Dis Primers. 2021;7(1):6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Villanueva A. Hepatocellular carcinoma. N Engl J Med. 2019;380(15):1450–62. [DOI] [PubMed] [Google Scholar]
- 6.Aunan JR, Cho WC, Soreide K. The biology of aging and cancer: a brief overview of shared and divergent molecular hallmarks. Aging Dis. 2017;8(5):628–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Berben L, Floris G, Wildiers H, Hatse S. Cancer and aging: two tightly interconnected biological processes. Cancers (Basel). 2021;13(6):1400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Campisi J. Aging, cellular senescence, and cancer. Annu Rev Physiol. 2013;75:685–705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Hamczyk MR, Nevado RM, Barettino A, Fuster V, Andres V. Biological versus chronological aging: JACC focus seminar. J Am Coll Cardiol. 2020;75(8):919–30. [DOI] [PubMed] [Google Scholar]
- 10.Hagg S, Belsky DW, Cohen AA. Developments in molecular epidemiology of aging. Emerg Top Life Sci. 2019;3(4):411–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ferrucci L, Gonzalez-Freire M, Fabbri E, Simonsick E, Tanaka T, Moore Z, et al. Measuring biological aging in humans: a quest. Aging Cell. 2020;19(2):e13080. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Kwon D, Belsky DW. A toolkit for quantification of biological age from blood chemistry and organ function test data: BioAge. Geroscience. 2021;43(6):2795–808. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Levine ME, Lu AT, Quach A, Chen BH, Assimes TL, Bandinelli S, et al. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY). 2018;10(4):573–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Li X, Ploner A, Wang Y, Magnusson PK, Reynolds C, Finkel D, et al. Longitudinal trajectories, correlations and mortality associations of nine biological ages across 20-years follow-up. Elife. 2020;9:e51507. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Belsky DW, Moffitt TE, Cohen AA, Corcoran DL, Levine ME, Prinz JA, et al. Eleven telomere, epigenetic clock, and biomarker-composite quantifications of biological aging: do they measure the same thing? Am J Epidemiol. 2018;187(6):1220–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Jiao M, Yan S, Shi Q, Liu Y, Li Y, Lv J, et al. Alcohol-related elevation of liver transaminase is associated with gut microbiota in male. Front Med (Lausanne). 2022;9:823898. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Loomba R, Gindin Y, Jiang Z, Lawitz E, Caldwell S, Djedjos CS, et al. DNA methylation signatures reflect aging in patients with nonalcoholic steatohepatitis. JCI Insight. 2018;3(2):e96685. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Mak J, McMurran CE, Kuja-Halkola R, Hall P, Czene K, Jylhava J, et al. Clinical biomarker-based biological aging and risk of cancer in the UK Biobank. Br J Cancer. 2023;129(1):94–103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Sudlow C, Gallacher J, Allen N, Beral V, Burton P, Danesh J, et al. UK biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med. 2015;12(3):e1001779. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Klemera P, Doubal S. A new approach to the concept and computation of biological age. Mech Ageing Dev. 2006;127(3):240–8. [DOI] [PubMed] [Google Scholar]
- 21.Levine ME. Modeling the rate of senescence: can estimated biological age predict mortality more accurately than chronological age? J Gerontol A Biol Sci Med Sci. 2013;68(6):667–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Kuiper LM, Polinder-Bos HA, Bizzarri D, Vojinovic D, Vallerga CL, Beekman M, et al. Epigenetic and metabolomic biomarkers for biological age: a comparative analysis of mortality and frailty risk. J Gerontol A Biol Sci Med Sci. 2023;78(10):1753–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Yang G, Cao X, Li X, Zhang J, Ma C, Zhang N, et al. Association of unhealthy lifestyle and childhood adversity with acceleration of aging among UK biobank participants. JAMA Netw Open. 2022;5(9):e2230690. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Lopez-Otin C, Blasco MA, Partridge L, Serrano M, Kroemer G. Hallmarks of aging: an expanding universe. Cell. 2023;186(2):243–78. [DOI] [PubMed] [Google Scholar]
- 25.Mutz J, Roscoe CJ, Lewis CM. Exploring health in the UK Biobank: associations with sociodemographic characteristics, psychosocial factors, lifestyle and environmental exposures. BMC Med. 2021;19(1):240. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Liu Z, Kuo PL, Horvath S, Crimmins E, Ferrucci L, Levine M. A new aging measure captures morbidity and mortality risk across diverse subpopulations from NHANES IV: a cohort study. PLoS Med. 2018;15(12):e1002718. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Sanfeliu-Redondo D, Gibert-Ramos A, Gracia-Sancho J. Cell senescence in liver diseases: pathological mechanism and theranostic opportunity. Nat Rev Gastroenterol Hepatol. 2024;21(7):477–92. [DOI] [PubMed] [Google Scholar]
- 28.Li F, Huangyang P, Burrows M, Guo K, Simon MC. FBP1 loss disrupts liver metabolism and promotes tumorigenesis through a hepatic stellate cell senescence secretome. Nat Cell Biol. 2020;22(6):728–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Rey S, Quintavalle C, Burmeister K, Calabrese D, Schlageter M, Quagliata L, et al. Liver damage and senescence increases in patients developing hepatocellular carcinoma. J Gastroenterol Hepatol. 2017;32(8):1480–6. [DOI] [PubMed] [Google Scholar]
- 30.Ganguly S, Sakane S, Hokutan K, Zhang V, Miciano C, Wang A, et al. Aging and aging-related senescence in liver. Semin Liver Dis. 2025;45(4):549–66. [DOI] [PubMed] [Google Scholar]
- 31.He Y, Su Y, Duan C, Wang S, He W, Zhang Y, et al. Emerging role of aging in the progression of NAFLD to HCC. Ageing Res Rev. 2023;84:101833. [DOI] [PubMed] [Google Scholar]
- 32.Anastasopoulos NA, Charchanti AV, Barbouti A, Mastoridou EM, Goussia AC, Karampa AD, et al. The role of oxidative stress and cellular senescence in the pathogenesis of metabolic associated fatty liver disease and related hepatocellular carcinoma. Antioxidants (Basel). 2023;12(6):1269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Yang Y, Cai Q, Zhu M, Rong J, Feng X, Wang K. Exploring the double-edged role of cellular senescence in chronic liver disease for new treatment approaches. Life Sci. 2025;373:123678. [DOI] [PubMed] [Google Scholar]
- 34.Lowe KO, Tanase CE, Maghami S, Fisher LE, Ghaemmaghami AM. Inflammatory network of liver fibrosis and how it can be targeted therapeutically. Immuno (Basel). 2023;3(4):38. [Google Scholar]
- 35.Schädel P, Wichmann-Costaganna M, Czapka A, Gebert N, Ori A, Werz O. Short-term caloric restriction and subsequent re-feeding compromise liver health and associated lipid mediator signaling in aged mice. Nutrients. 2023;15(16):16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Macias R, Monte MJ, Serrano MA, Gonzalez-Santiago JM, Martin-Arribas I, Simao AL, et al. Impact of aging on primary liver cancer: epidemiology, pathogenesis and therapeutics. Aging (Albany NY). 2021;13(19):23416–34. [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
The UK Biobank is an open-access resource, available at https://www.ukbiobank.ac.uk/researchers/, and can be obtained by submitting a data request proposal. The data that support the findings of this study were used under license for the current study (Application 98,124), and thus are not publicly available.