The association between phenotypic age acceleration and the risk of all-cause and cancer mortality among cancer survivors: NHANES 1999–2018

Abstract

Phenotypic age (PhenoAge) is a biological aging clock that estimates an individual’s biological age. However, the effect of PhenoAge acceleration (PhenoAgeAccel) on cancer is unclear. This study investigates the relationship between PhenoAgeAccel and cancer survivors. Data for this cohort study were sourced from the U.S. National Health and Nutrition Examination Survey (NHANES) spanning 1999–2018. The relationship between PhenoAgeAccel and cancer prevalence was evaluated using weighted multivariate logistic regression. Kaplan–Meier analyses and weighted multivariate-adjusted Cox analyses were conducted to examine the association between PhenoAgeAccel and all-cause as well as cancer-specific mortality in cancer survivors. Restricted cubic spline (RCS) analysis was utilized to assess nonlinear associations. Subgroup and sensitivity analyses were also performed to confirm the robustness of the findings. A total of 34,246 participants were included in our study, of which 3067 were cancer survivors (8.95% prevalence). With a median follow-up of 117 months (interquartile range: 50–155 months), there were 1161 deaths, including 351 from cancer. Weighted multivariate regression analysis revealed a significant positive association between higher PhenoAgeAccel and cancer prevalence (P for trend < 0.001). Multivariable-adjusted Cox regression analyses showed that elevated PhenoAgeAccel was significantly associated with increased all-cause and cancer-specific mortality among cancer survivors (P for trend < 0.001). RCS regression indicated no nonlinear relationship between PhenoAgeAccel and mortality outcomes (P for nonlinear relationship > 0.05). Kaplan-Meier analyses indicated a poorer prognosis with higher PhenoAgeAccel. Subgroup analyses based on tumor classification highlighted the differential prognostic impact of PhenoAgeAccel across various tumor types. Our findings reveal a significant linear correlation between PhenoAgeAccel and both all-cause and cancer-specific mortality in cancer survivors.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-30747-2.

Introduction

Cancer is the second leading cause of death in the U.S. and the number one cause of death for people under the age of 85, with an estimated 611,720 deaths in the U.S. in 2024, which equates to approximately 1,680 deaths per day¹. Despite a decline in cancer mortality rates from 1991 to 2021—thanks to reduced tobacco use, improved cancer detection, and advances in treatment—the global cancer burden is expected to rise due to an aging population and risk factors such as obesity, sedentary lifestyles, and shifts in fertility rates^1–6. Cancer incidence and progression are shaped by multiple factors, with age as a significant determinant. As people age, they experience increased vulnerability to cancer due to accumulated exposure to carcinogens, age-related declines in DNA repair, and weakened immune surveillance^7–10. This heightened susceptibility emphasizes the need to understand the complex relationship between aging and cancer development.

Aging is a complex, multifaceted process that involves dysregulation across multiple physiological systems¹¹. It is marked by the accumulation of molecular changes and the disruption of key aging hallmarks, including genomic instability, telomere attrition, and stem cell exhaustion¹². While chronological age is a critical risk factor for age-related mortality, individuals of the same chronological age can display varying susceptibilities to these conditions, highlighting differences in their biological aging processes. This distinction underscores the need to differentiate between chronological and biological aging¹³. To assess biological aging, various measures have been proposed, from single biomarkers such as telomere length to algorithms that integrate data from epigenetics, proteomics, metabolomics, and other molecular-level analyses^14,15. Of these, algorithms that incorporate standard clinical parameters have emerged as some of the most accurate for predicting morbidity and mortality^16,17. Phenotypic age (PhenoAge) is a biological aging clock that estimates an individual’s biological age using chronological age, clinical biomarkers, and blood cell parameters^18,19. Developed by Levine et al., PhenoAge effectively identifies individuals at higher risk for age-related diseases²⁰. While the role of age in cancer development and progression is well established, the impact of PhenoAgeAccel on cancer survivors remains unclear.

Therefore, this study leveraged data from the NHANES to examine the association between PhenoAgeAccel and morbidity and long-term mortality in cancer survivors. This research provides valuable insights for screening and managing high-risk groups of cancer survivors.

Materials and methods

Study design and population

NHANES, an annual survey administered by the National Center for Health Statistics (NCHS), under the umbrella of the Centers for Disease Control and Prevention (CDC), serves the purpose of gathering data from a nationally representative sample of non-institutionalized civilians in the United States. Prior to implementation, all NHANES surveys undergo rigorous scrutiny and approval by the Disclosure Review Board of the NCHS. Detailed information regarding the ethical approval and informed consent processes can be accessed through the National Center for Health Statistics²¹. Extensive documentation on the design, methodology, and weighting of NHANES has been previously published^22,23.

NHANES employs a sophisticated stratified, complex, multistage sampling methodology to select households from randomized subgroups. Within these households, a subset of adults is randomly chosen to participate in surveys pertaining to health status, healthcare utilization, lifestyle risk factors, prevalent diseases, and other pertinent health-related matters^22,23. Trained investigators conduct personal interviews to gather the requisite data. This study adopted a nationwide cross-sectional design, utilizing secondary analyses of publicly accessible and deidentified data from NHANES. Therefore, supplementary institutional review board approval or informed consent was not necessary. Further information can be accessed via http://www.cdc.gov/nchs/nhanes.

Exclusion criteria: We analyzed data from 101,316 participants across NHANES surveys spanning 1999 to 2018. Sequentially, we excluded 46,235 individuals under 20 years old, 17,498 with missing phenotypic age data, 31,136 lacking survival data, 1 without self-reported physician-diagnosed cancer, 75 pregnant women, and 175 with incomplete demographic details (including age, sex, race/ethnicity, and body mass index). This process resulted in the exclusion of 67,070 participants, leaving a final analytic sample of 34,246 individuals, including 3,067 cancer survivors (Fig. 1).

Fig. 1.

Flowchart for the selection of participants in this study.

Measurement of phenoage acceleration

PhenoAge, a measure of biological age, is determined through analysis of clinical laboratory blood components using the PhenoAge algorithm^19,20. Given the established validity and practicality of the PhenoAge algorithm within NHANES, we utilized it to evaluate biological age based on pertinent clinical laboratory blood metrics. The PhenoAge algorithm was developed from elastic net regressions of various biomarkers in NHANES III. Clinical biomarkers included in PhenoAge calculations are albumin, creatinine, glucose, leukocyte count, percentage of lymphocytes, erythrocyte distribution width, mean erythrocyte volume, and alkaline phosphatase. Phenotypic age is computed using the following formula described previously²⁰.

Where

.

And

.

We utilized the R package “BioAge” to develop and train the PhenoAge algorithm using NHANES III data and subsequently compute PhenoAge for NHANES IV cycles spanning 1999 to 2018²⁴. Due to the lack of C-reactive protein (CRP) in the NHANES data from 2011 to 2018, we omitted it as a clinical biomarker for PhenoAge calculation, aligning with prior research findings^25–27. Moreover, in exploring the impact of CRP on PhenoAge determinations, some studies had compared PhenoAge outcomes derived from biomarker sets both with and without CRP, revealing a robust correlation (correlation coefficients ranging from 0.967 to 0.99)^26,27.

PhenoAgeAccel represents the disparity between biological age (PhenoAge value) and chronological age, normalized to yield a mean of 0 and a standard deviation (SD) of 1. An elevation in PhenoAgeAccel signifies a heightened state of biological aging (accelerated age), potentially elevating the individual’s susceptibility to disease and mortality. Conversely, diminishing PhenoAgeAccel values denote a deceleration in biological aging.

Definition of cancer survivor

The self-reported physician diagnosis of cancer is obtained from the “Medical Conditions” section of NHANES, collected through professionally self-administered questionnaires²⁸. Cancer survivorship status is determined based on responses to the question: “Have doctors or other healthcare professionals informed you of a diagnosis of cancer or any type of malignant tumor?” Positive responses classify individuals as cancer survivors, while negative responses categorize them as non-cancer individuals.

Measurement of mortality

Mortality data were extracted from the NHANES Public Use Associated Mortality File, which remains valid through December 31, 2019. Cause of death was documented using ICD-10 (International Statistical Classification of Diseases, 10th edition) codes²⁹. Our analysis centered on both all-cause mortality and cancer mortality (ICD-10: C00-C97). The follow-up duration spanned from the date of initial diagnosis to the date of death or December 31, 2019, whichever occurred first.

Covariates

The study examined several covariates, encompassing demographic factors such as age (in years), gender (male or female), ethnicity (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, or Other), educational attainment (below high school, high school, or above high school), and marital status (married/cohabiting, widowed/divorced/separated, or never married). Income was evaluated using the Poverty Income Ratio (PIR), categorized as ≤ 1.0, 1.1-3.0, and > 3.0 according to guidelines from the US Department of Health and Human Services.

Additional covariates potentially influencing the analysis results included body mass index (BMI) category (< 18.5, 18.5–25.0, 25.0-29.9, or > 29.9 kg/m²)³⁰ and waist circumference. Non-smokers were defined as individuals who had smoked fewer than 100 cigarettes in their lifetime. Current smokers were those who had smoked more than 100 cigarettes and were presently smoking, while former smokers were individuals who had smoked more than 100 cigarettes but had since quit³¹. Alcohol consumption was categorized as non-drinking and drinking (≥ 12 drinks in a year). Physical activity was quantified as metabolic equivalent (MET) minutes of moderate to vigorous exercise per week according to World Health Organization guidelines³². Diabetes was determined based on self-report, glycosylated hemoglobin ≥ 6.5%, or fasting blood glucose ≥ 126 mg/dL (7.0 mmol/L). Hypertension was determined based on medication use or self-reported diagnosis. The database also captured parameters including complete blood count parameters, serum albumin levels, high-density lipoprotein cholesterol (HDL-C), and total cholesterol (TG) levels.

Statistical analyses

Considering the stratification and clustering caused by the complex sample designs and the selectivity of the weights in different cycles^33,34, the analyses of the 1999–2002 cycle in our study were conducted using the four-year cycle Mobile Examination Center (MEC) exam weights (wtmec4 year), whereas the 2003–2018 were conducted using the NHANES two-year cycle Mobile Examination Center (MEC) exam weights (wtmec2 year) to generate nationally representative estimates.

Categorical variables were presented as numbers and corresponding weighted proportions, with associations tested using the Rao-Scott χ² test. Continuous variables’ distributions (normal or non-normal) were evaluated using the Kolmogorov-Smirnov normality test. Non-normally distributed variables were reported as weighted median and interquartile range (IQR), while normally distributed continuous variables were expressed as weighted mean and its corresponding standard error (SE).

We undertook a study to explore the association between PhenoAgeAccel, categorized into three groups (Tertile 1, Tertile 2, and Tertile 3), and cancer survivorship. The reference group comprised individuals in Tertile 1. Weighted logistic regression models were utilized to calculate odds ratios (ORs) and 95% confidence intervals (CIs) to evaluate cancer prevalence. Additionally, weighted Cox regression models were employed to compute hazard ratios (HRs) and 95% CIs to assess long-term mortality among cancer survivors.

Both the weighted logistic regression model and the weighted Cox regression model comprised three iterations. The crude model remained unadjusted, while Model 1 was adjusted for age, MET, sex (male or female), and ethnicity (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, or Other). Model 2 included additional adjustments for education level (below high school, high school, or above high school), family poverty income ratio (≤ 1.0, 1.1-3.0, or > 3.0), drinking status (nondrinker, drinker), smoking status (never smoker, former smoker, or current smoker), BMI (< 18.5, 18.5–25.0, 25.0-29.9, or > 29.9), self-reported diabetes (yes or no), and self-reported hypertension (yes or no).

We evaluated the dose-response relationship between PhenoAgeAccel and both all-cause mortality and cancer mortality among cancer survivors using restricted cubic spline regression analyses. These analyses incorporated nodes positioned at the 5th, 35th, 65th, and 95th percentiles for each exposure variable. Additionally, Kaplan-Meier analysis was employed to examine the association between PhenoAgeAccel and long-term survival among cancer survivors.

A subgroup analysis was conducted to explore potential variations in the impact of PhenoAgeAccel on long-term mortality across different populations of cancer survivors. This analysis stratified the final analyzed sample by age (< 60 and > 60 years), sex (male and female), smoking status (smoker and non-smoker), body mass index (< 25, 25-29.9, and > 29.9), self-reported hypertension (yes and no), and self-reported diabetes mellitus (yes and no). Multiplicative interaction terms among the subgroups, PhenoAgeAccel, and long-term mortality were incorporated into the model to evaluate potential interaction effects. Additionally, sensitivity analyses were performed to enhance the robustness of the findings.

The data analyses were conducted using R software, version 4.2.2, obtained from the R Project for Statistical Computing. A significance threshold of p < 0.05 was applied, using a two-sided test.

Result

Population characteristics

The study included 34,246 participants, representing 143.07 million noninstitutionalized residents of the United States, with a mean (SE) age of 45.5 (16.8) years, and 52% were female (Table 1). Among them, 3,067 were cancer survivors, aged 20 to 85 years, with a mean age of 61.9 ± 15.1 years. The racial distribution was: Non-Hispanic whites: 87%, Non-Hispanic blacks: 5.1%, Mexican Americans: 2.0%, Other Hispanics: 2.2%, and Other Race: 3.5%. Cancer survivors were more likely to be older women, have higher education and income, be divorced, and have a history of smoking and alcohol abuse. They also exhibited higher waist circumferences, lower physical activity levels, and increased rates of comorbid hypertension or diabetes. Compared to non-cancer participants, cancer survivors had significantly lower hemoglobin, serum albumin, lymphocyte counts, and platelet counts, while HDL levels were higher. PhenoAgeAccel and PhenoAge differed significantly between cancer and non-cancer participants (P < 0.001).

Table 1.

Baseline characteristics of participants in the NHANES, 1999–2018.

Characteristics	Overall, N = 34,246 (100%)	Cancer survivors		P Value
		No, N = 31,179 (91%)	Yes, N = 3067 (9%)
Sex, %				< 0.001
Female	17,749 (52%)	16,124 (51%)	1,625 (59%)
Male	16,497 (48%)	15,055 (49%)	1,442 (41%)
Age, %				< 0.001
20–35 years	9,039 (28%)	8,897 (31%)	142 (5.4%)
35–60 years	13,667 (48%)	12,958 (49%)	709 (34%)
60 + years	11,540 (24%)	9,324 (20%)	2,216 (61%)
Race/ethnicity, %				< 0.001
Non-Hispanic White	16,161 (70%)	13,924 (68%)	2,237 (87%)
Non-Hispanic Black	6,479 (10%)	6,112 (11%)	367 (5.1%)
Mexican American	6,766 (7.9%)	6,541 (8.5%)	225 (2.0%)
Other Race - Including Multi-Racial	2,185 (6.3%)	2,082 (6.6%)	103 (3.5%)
Other Hispanic	2,655 (5.4%)	2,520 (5.7%)	135 (2.2%)
Education level, %				0.026
Below high school	9,819 (18%)	9,081 (18%)	738 (16%)
High school	8,096 (25%)	7,354 (25%)	742 (24%)
Above high school	16,330 (57%)	14,743 (57%)	1,587 (60%)
Marital status, %				< 0.001
Married/cohabiting	21,289 (65%)	19,371 (65%)	1,918 (67%)
Widowed/divorced/separated	7,383 (18%)	6,403 (17%)	980 (28%)
Never married	5,574 (17%)	5,405 (18%)	169 (5.2%)
Family PIR, %				< 0.001
≤ 1.0	6,040 (12%)	5,654 (13%)	386 (9.4%)
1.1–3.0	16,390 (41%)	14,916 (41%)	1,474 (39%)
> 3.0	11,816 (47%)	10,609 (46%)	1,207 (51%)
Smoking status, %				< 0.001
Never smoker	18,307 (53%)	16,961 (54%)	1,346 (44%)
Former smoker	8,677 (25%)	7,428 (24%)	1,249 (39%)
Current smoker	7,261 (22%)	6,789 (23%)	472 (17%)
Drinking status, %				< 0.001
Drinker	5,999 (17%)	5,359 (16%)	640 (21%)
Nondrinker	28,247 (83%)	25,820 (84%)	2,427 (79%)
Body mass index, %				0.4
Underweight, kg/m2	513 (1.7%)	463 (1.7%)	50 (1.9%)
Normal, kg/m2	9,430 (30%)	8,612 (31%)	818 (29%)
Overweight, kg/m2	11,674 (34%)	10,584 (34%)	1,090 (35%)
Obese, kg/m2	12,028 (34%)	11,001 (34%)	1,027 (35%)
Family PIR	3.03 (1.62)	3.01 (1.63)	3.21 (1.58)	< 0.001
BMI, kg/m2	29 (7)	29 (7)	29 (6)	0.3
Waist Circumference (cm)	98 (16)	97 (16)	100 (16)	< 0.001
Self-reported hypertension, %	11,323 (29%)	9,671 (27%)	1,652 (48%)	< 0.001
Self-reported diabetes, %	4,591 (9.6%)	3,981 (9.0%)	610 (16%)	< 0.001
MET minute/week	4,297 (13,301)	4,349 (13,441)	3,761 (11,739)	0.008
Hemoglobin (g/dL)	4.30 (0.34)	4.31 (0.34)	4.22 (0.32)	< 0.001
Albumin (g/dL)	14.38 (1.46)	14.40 (1.47)	14.13 (1.36)	< 0.001
Platelet count (109/L)	260 (66)	261 (66)	249 (69)	< 0.001
Lymphocyte count (109/L)	2.12 (0.70)	2.14 (0.69)	1.99 (0.80)	< 0.001
HDL-C, mg/dl	53 (17)	53 (16)	55 (17)	< 0.001
Total cholesterol, mg/dl	198 (41)	198 (41)	200 (43)	0.050
Age, years	46.5 (16.8)	45.0 (16.2)	61.9 (15.1)	< 0.001
PhenoAge, years	43 (18)	41 (17)	59 (17)	< 0.001
PhenoAgeAccel	-3.5 (4.7)	-3.6 (4.6)	-2.5 (5.4)	< 0.001
PhenoAgeAccel classifcation				< 0.001
Tertile 1	11,301 (36%)	10,490 (37%)	811 (30%)
Tertile 2	11,301 (34%)	10,361 (35%)	940 (32%)
Tertile 3	11,644 (30%)	10,328 (29%)	1,316 (37%)

Continuous variables are described as means [SD]. Categorical variables are presented as numbers (percentages). N reflect the study sample while percentages reflect the survey-weighted.

PIR, poverty income ratio; HDL-C, High-density lipoprotein cholesterol; PhenoAge, phenotypic age; PhenoAgeAccel, PhenoAge acceleration.

Associations of phenoage acceleration with cancer

We examined the association between PhenoAge acceleration and cancer prevalence using weighted logistic regression models. Before modeling to avoid the emergence of multicollinearity affecting the robustness of the model, we performed variance inflation factor and tolerance tests on all included variables, and the results showed that the VIF was less than 5 and the tolerance values were significantly greater than 0.1, suggesting that there was no multicollinearity. Table 2 presents results from all three logistic regression models, revealing a significant positive relationship between PhenoAge acceleration and cancer prevalence. Across the three models, compared to participants in the lowest tertile, those in the highest tertile exhibited adjusted ORs and 95% CIs of 1.65 (1.50, 1.81), 1.40 (1.27, 1.54), and 1.36 (1.23, 1.51), respectively. Moreover, linear trend tests indicated statistically significant associations, with all P-values for trend being less than 0.001.

Table 2.

Logistic regression analysis between phenoageaccel and prevalence of cancer among adults in NHANES 1999–2018.

	PhenoAgeAccel					P for trend
	Tertile 1	Tertile 2	P Value	Tertile 3	P Value
Range	< -5.55	-5.55 to -1.83		> -1.83
Crude	1.00 [Reference]	1.17 (1.06, 1.29)	0.001	1.65 (1.50, 1.81)	< 0.001	< 0.001
Model 1	1.00 [Reference]	1.16 (1.05, 1.28)	0.004	1.40 (1.27, 1.54)	< 0.001	< 0.001
Model 2	1.00 [Reference]	1.15 (1.04, 1.28)	0.007	1.36 (1.23, 1.51)	< 0.001	< 0.001

Data are presented as OR (95% CI); Model 1 was adjusted as age (continuous), MET (continuous), sex (male or female), and race/ethnicity (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black or Other); Model 2 was adjusted as model 1 plus education level (below high school, high school, or above high school), family poverty income ratio (≤ 1.0,1.1–3.0, or > 3.0), drinking status (nondrinker, drinker), smoking status (never smoker, former smoker, or current smoker), BMI (< 18.5, 18.5–25.0, 25.0-29.9, or > 29.9), self-reported diabetes (yes or no), and self-reported hypertension (yes or no).

Associations of phenoage acceleration with long-term mortality

Table 3 delineated the association between PhenoAge acceleration and all-cause mortality, as well as cancer mortality, among cancer survivors. Over a mean (SD) follow-up of 108.61 (61.35) months, 1,161 (37.85%) out of 3,067 cancer survivors succumbed to all-cause mortality, with 351 (11.44%) attributed to cancer. All three weighted Cox regression models confirmed a significant positive association between PhenoAge acceleration and both all-cause mortality and cancer mortality in cancer survivors (all p for trend < 0.001). Comparing with the lowest tertile, the multivariable-adjusted HRs and 95% CIs for all-cause mortality were 1.97 (1.54, 2.52) for individuals in the highest tertile and 1.62 (1.20, 2.21) for cancer mortality among those in the highest tertile of PhenoAge acceleration in cancer survivors.

Table 3.

Cox regression analysis between phenoageaccel and long-term mortality among cancer survivor in NHANES 1999–2018.

	PhenoAgeAccel					P for trend
	Tertile 1	Tertile 2	P Value	Tertile 3	P Value
All-cause mortality
No. deaths/total	269/1012	362/1012		530/1043
Crude	1.00 [Reference]	1.54 (1.27, 1.86)	< 0.001	2.85 (2.37, 3.44)	< 0.001	< 0.001
Model 1	1.00 [Reference]	1.33 (1.06, 1.66)	0.013	2.30 (1.84, 2.87)	< 0.001	< 0.001
Model 2	1.00 [Reference]	1.22 (1.01, 1.55)	0.045	1.97 (1.54, 2.52)	< 0.001	< 0.001
Cancer mortality
No. deaths/total	78/1012	112/1012		161/1043
Crude	1.00 [Reference]	1.49 (1.10, 2.02)	0.01	2.19 (1.65, 2.92)	< 0.001	< 0.001
Model 1	1.00 [Reference]	1.34 (1.02, 1.77)	0.037	1.71 (1.28, 2.30)	< 0.001	< 0.001
Model 2	1.00 [Reference]	1.22 (0.90, 1.67)	0.202	1.62 (1.20, 2.21)	0.006	0.006

Data are presented as HR (95% CI); Model 1 was adjusted as age (continuous), MET (continuous), sex (male or female), and race/ethnicity (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black or Other); Model 2 was adjusted as model 1 plus education level (below high school, high school, or above high school), family poverty income ratio (≤ 1.0,1.1–3.0, or > 3.0), drinking status (nondrinker, drinker), smoking status (never smoker, former smoker, or current smoker), BMI (< 18.5, 18.5–25.0, 25.0-29.9, or > 29.9), self-reported diabetes (yes or no), and self-reported hypertension (yes or no).

Furthermore, to deepen our understanding of the relationship between PhenoAge acceleration and the long-term prognosis of cancer survivors, we conducted a Kaplan-Meier analysis. As depicted in Fig. 2, PhenoAge acceleration exhibited a significant correlation with the prognosis of cancer survivors: greater PhenoAge acceleration corresponded to higher all-cause mortality (p < 0.0001) and higher cancer mortality (p < 0.0001), indicative of a poorer prognosis.

Fig. 2.

Kaplan–Meier survival analysis of the association between PhenoAgeAccel and all-cause mortality (A) and cancer mortality (B) in cancer survivors.

Dose-response analysis of phenoage acceleration with long-term mortality

Multivariate-adjusted restricted cubic spline analyses revealed no nonlinear relationship between PhenoAge acceleration and all-cause mortality in cancer survivors (P for nonlinear relationship = 0.4012; Fig. 3A). Similarly, there was no nonlinear relationship between PhenoAge acceleration and cancer mortality among cancer survivors (P for nonlinear relationship = 0.7395; Fig. 3B).

Fig. 3.

Restricted cubic spline analysis to assess the association between PhenoAgeAccel and all-cause mortality (A) and cancer mortality (B) among cancer survivors. Adjusted for age (continuous), MET (continuous), sex (male or female), ethnicity (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black or Other race), education level (below high school, high school, or above high school), family poverty income ratio (≤ 1.0, 1.1–3.0, or > 3.0), drinking status (nondrinker, drinker), smoking status (never smoker, former smoker, or current smoker), BMI (< 18.5, 18.5–25.0, 25.0-29.9, or > 29.9), self-reported diabetes (yes or no), and self-reported hypertension (yes or no).

Subgroup analyses

Figure 4 illustrates the association between PhenoAge acceleration and long-term mortality in cancer survivors through subgroup analysis, stratified by age, sex, body mass index, smoking status, hypertension, and diabetes status. The effect sizes of PhenoAge acceleration on all-cause mortality and cancer mortality among cancer survivors remained largely stable across all prespecified subgroups, with P values greater than 0.05 for all interactions.

Fig. 4.

Subgroup analyses of the association between PhenoAgeAccel and all-cause mortality and cancer mortality among cancer survivors stratified by age, gender, smoking, body-mass index, self-reported hypertension, and self-reported diabetes.

To further evaluate the influence of PhenoAge acceleration on various tumor types, subgroup analyses were conducted based on tumor classification (detailed in Table S1). As depicted in Table 4, PhenoAge acceleration exhibited a significant impact on long-term mortality in patients with thyroid and breast cancer, as well as in patients with urologic tumors and skin and soft tissue tumors.

Table 4.

Subgroup analysis by cancer of phenoageaccel and long-term mortality among cancer survivor in NHANES 1999–2018.

	PhenoAgeAccel			P for trend
	Tertile 1	Tertile 2	Tertile 3
All-cause mortality
Breast and Thyroid	1.00 [Reference]	1.36 (0.75, 2.50)	2.54 (1.33, 4.93)	0.010
Urinary system	1.00 [Reference]	1.63 (1.00, 2.71)	1.95 (1.18, 3.22)	0.030
Digestive system	1.00 [Reference]	0.62 (0.28, 1.34)	1.47 (0.67, 3.220	0.058
Reproductive system	1.00 [Reference]	0.77 (0.34, 1.710	1.17 (0.50, 2.70)	0.600
Skin and Soft tissue	1.00 [Reference]	1.10 (1.02, 1.78)	1.63 (1.09, 2.42)	0.040
Other	1.00 [Reference]	1.34 (0.63, 2.88)	3.52 (1.64, 2.88)	0.002
Cancer mortality
Breast and Thyroid	1.00 [Reference]	0.72 (0.36, 1.44)	0.74 (0.36, 1.52)	0.600
Urinary system	1.00 [Reference]	1.95 (1.01, 4.12)	2.66 (1.37, 5.54)	0.014
Digestive system	1.00 [Reference]	0.71 (0.28, 1.77)	0.98 (0.42, 2.39)	0.700
Reproductive system	1.00 [Reference]	0.62 (0.22, 1.62)	0.40 (0.11, 1.20)	0.300
Skin and Soft tissue	1.00 [Reference]	1.05 (0.87, 1.89)	1.47 (1.20, 2.67)	0.032
Other	1.00 [Reference]	2.44 (0.97, 6.74)	3.46 (1.38, 9.61)	0.027

Data are presented as HR (95% CI), which was adjusted as age (continuous), MET (continuous), sex (male or female), ethnicity (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black or Other), education level (below high school, high school, or above high school), family poverty income ratio (≤ 1.0,1.1-3.0, or > 3.0), drinking status (nondrinker, drinker), smoking status (never smoker, former smoker, or current smoker), BMI (< 18.5, 18.5–25.0, 25.0-29.9, or > 29.9), self-reported diabetes (yes or no), and self-reported hypertension (yes or no).

Sensitivity analyses

To assess the robustness of the association between PhenoAge acceleration and all-cause mortality and cancer mortality in cancer survivors, we performed sensitivity analyses. To minimize the effect of reverse causality, we excluded cancer survivors who died within the 2 years prior to follow-up, and found that the association between PhenoAge acceleration and all-cause mortality and cancer mortality in cancer survivors remained significant (Table S2).

Discussion

To elucidate the relationship between PhenoAgeAccel and cancer, we analyzed data from 34,246 individuals from NHANES. Adjusting for confounding variables, we identified a significant positive association between PhenoAgeAccel and cancer incidence. Among cancer survivors, PhenoAgeAccel showed a significant linear relationship with all-cause and cancer-specific mortality, with faster acceleration linked to higher mortality. Kaplan-Meier analyses confirmed that PhenoAge acceleration significantly influenced the prognosis of cancer survivors. Subgroup analyses by tumor classification revealed varying prognostic impacts of PhenoAgeAccel on different tumor types. In conclusion, PhenoAgeAccel serves as an important prognostic index for cancer survivors.

No one escapes the marks of aging, which represent the progressive breakdown of the system that sustains youthful health. Cancer incidence rises with age and reaches its peak at 85, underscoring aging as the most significant risk factor for cancer³⁵. Siegel et al. found that patients aged 60 and older are twice as likely to develop invasive cancer as younger patients³⁶. While aging (decreased function and growth) and cancer (increased growth and survival) seem to be polar opposites, research shows they share many characteristics⁷, suggesting this overlap may be a key factor in how aging promotes cancer progression. Specifically, the underlying mechanisms may include: First, aging leads to the accumulation of somatic mutations, gene copy-number variations, and chromosomal aneuploidies in nuclear DNA across humans and model organisms. These alterations can impact crucial genes, causing cellular changes, tissue abnormalities, and organismal defects that contribute to the onset of age-related conditions such as cancer^37,38; Second, species have developed intricate DNA repair networks to counteract genetic damage and uphold cellular homeostasis. Nevertheless, not all DNA damage is repaired, and repair efficiency diminishes with age. The resulting buildup of genomic damage heightens susceptibility to cancer and other age-related diseases^39,40; Third, aging exacerbates inflammation, resulting in a state known as “inflammaging,” which contributes to age-related diseases such as osteoarthritis, atherosclerosis, sarcopenia, and neuroinflammation^41–43. Likewise, inflammation is a key feature in cancer progression^44,45; Fourth, the gut’s diverse bacterial community develops primarily during early childhood and remains relatively stable in adulthood. However, aging brings ongoing shifts in the microbial ecosystem’s composition and activity, ultimately reducing diversity. This reduction is closely associated with the progression of various cancers^46,47; Finally, various epigenetic changes, including shifts in DNA methylation patterns, post-translational histone modifications, chromatin remodeling, and non-coding RNA function, influence gene expression and other crucial cellular processes. These changes contribute to the emergence of aging-associated human pathologies, including cancer^48,49.

Biological age refers to an individual’s physiological condition and may differ from chronological age, offering insight into their aging and health status⁵⁰. It provides a measure of how an individual’s body functions compared to their chronological age. Biological age is commonly assessed using biomarkers such as DNA methylation, telomere length, transcriptomics, proteomics, metabolomics, and composite biomarker panels^13,51. Among these, DNA methylation age (DNAmAge), or the epigenetic clock, stands out as a particularly promising metric^13,52. As awareness of the impact of biological age on health grows, numerous tools for measuring biological age have emerged.

PhenoAge, also known as DNAm PhenoAge, is an epigenetic clock introduced by Levine et al. in 2018 that integrates chronological age and multiple clinical biomarkers²⁰. Using clinical data from the Third National Health and Nutritional Survey (NHANES III) in the United States, the model incorporates 10 clinical measures: chronological age, albumin, creatinine, glucose, C-reactive protein, lymphocyte percentage, mean cell volume, red blood cell distribution width, alkaline phosphatase, and white blood cell count. The combination of these factors allows PhenoAge to effectively differentiate between the risks of morbidity and mortality in individuals of the same actual age²⁰. PhenoAge outperforms other epigenetic clocks in predicting 10- and 20-year mortality, showing stronger links with lifestyle behaviors⁵³.

PhenoAgeAccel, the difference between biological and chronological age, represents an individual’s functional aging and is strongly linked to various diseases. Roberts et al. found a significant association between PhenoAge acceleration and the onset of atrial fibrillation using UK Biobank data, indicating that biological aging is a key factor independent of chronological age⁵⁴. Moreover, Gao et al. discovered that individuals with older biological ages had a higher risk of developing depression or anxiety (11.3% increase per SD of PhenoAge acceleration, 95% CI: 9%-13%) in a study of 424,299 UK Biobank participants. This finding suggests that accelerated biological aging may be a risk factor for depression and anxiety in middle-aged and older adults and a potential target for risk assessment and intervention⁵⁵. Another study from the UK Biobank revealed that PhenoAge acceleration, as measured by clinical biomarkers, is linked to higher risks of any cancer, lung cancer, and colorectal cancer⁵⁶. While the relationship between PhenoAge acceleration and many diseases is becoming clearer, its association with long-term mortality in cancer survivors remains underexplored. Our study found a significant linear correlation between PhenoAgeAccel and both cancer incidence and long-term mortality in cancer survivors. Faster PhenoAgeAccel was associated with a greater risk of cancer and higher long-term mortality in these individuals. To our knowledge, this is the first study to establish the relationship between PhenoAgeAccel and both all-cause and cancer-specific mortality in cancer survivors. Our findings emphasize the critical role of PhenoAgeAccel in managing cancer survivors and offer a valuable reference for their prognostic management. Cancer survivors with higher PhenoAgeAccel need better clinical supervision and higher re-examination frequency to avoid poor prognosis.

PhenoAge integrates 10 clinical metrics that respond to an individual’s overall state in multiple dimensions. Many of the metrics that make up PhenoAge have now been shown to correlate significantly with tumor prognosis. For example, Extensive research has underscored the association between hypoalbuminemia and poor survival across various cancer types^57,58. Moreover, lymphocytes play a pivotal role in the host’s anticancer defense mechanisms. They secrete cytokines such as interferon-g and tumor necrosis factor-a (TNF-a), which enhance prognosis by inducing apoptosis and impeding cancer cell proliferation, invasion, and migration⁵⁹. Therefore, PhenoAge can be associated with a more comprehensive and integrated assessment of the prognosis of cancer survivors by integrating these key clinical indicators. Finally, compared to other biological age indicators, such as the epigenetic clock, the simplicity and low cost of the PhenoAgeAccel, which is calculated through a routine blood test, provides a unique advantage for the prognostic monitoring of cancer survivors.

Our study also investigated the association between PhenoAgeAccel and long-term mortality across different cancer subtypes. We found that PhenoAgeAccel significantly affected long-term mortality in patients with breast cancer, thyroid cancer, urologic tumors, and skin and soft-tissue tumors, indicating varied impacts across different tumor types. These findings align with previous research. Valencia et al. reported a statistically significant relationship between DNA methylation-based accelerated age, as measured by epigenetic clocks, and breast cancer risk⁶⁰. Similarly, Liu et al. found that DNAm age acceleration was linked to poorer outcomes in thyroid cancer⁶¹. PhenoAgeAccel contains multiple indicators of inflammation, such as white blood cell counts and lymphocyte percentages, which reflect the degree of inflammation in the tumor microenvironment. Inflammation, a hallmark of the tumor microenvironment, significantly influences disease progression and prognosis in cancer survivors^62,63. Numerous studies have investigated various systemic inflammatory biomarkers and have demonstrated their high predictive value for the prognosis of different cancer types^64–66. Therefore, whether the differential effect of PhenoAgeAccel on the prognosis of different tumors is achieved by responding to the inflammatory components of their microenvironment needs to be further investigated.

Our study presents several notable strengths. First, it leverages ten cycles of NHANES data from 1999 to 2018 to examine the impact of PhenoAgeAccel on long-term mortality in cancer survivors, offering robust results through a large sample size and an extensive time frame. Second, as the first cross-sectional study to assess the impact of PhenoAgeAccel on long-term mortality in cancer survivors, it provides valuable reference points for clinical management and prognostic evaluation, shedding new light on the treatment of this population. Third, we used a range of analytical methods, including restricted cubic spline nonlinear analysis, subgroup analysis, and Kaplan-Meier survival analysis, to ensure the robustness of our findings and emphasize the significance of PhenoAgeAccel’s impact on cancer survivors.

There are several limitations to our study that should be acknowledged. First, our data on cancer survivors were sourced from the medical questionnaire portion of the NHANES database, which depends on patient self-report and may introduce recall bias, which could potentially have an impact on the results. Second, while we accounted for known confounders, unrecognized confounding factors may still influence our results, such as cancer-specific prognostic factors, including cancer stage, specific treatment history, and so on. Third, as our study is a cross-sectional study based on the U.S. NHANES database, it cannot establish causality, and more research is needed to better understand the relationship between PhenoAgeAccel and cancer survivor prognosis, as well as larger multicenter studies to verify the generalizability of this relationship. Finally, we used COX regression modeling to assess the correlation between PhenoAgeAccel and cancer-specific mortality in cancer survivors, which may be affected by competing risk bias.

Conclusions

Through a comprehensive survey of cancer survivors across the United States, our study demonstrates a significant linear relationship between PhenoAgeAccel and both all-cause and cancer-specific mortality. We found that among cancer survivors, higher PhenoAgeAccel is associated with increased long-term mortality. This is the first study to evaluate the relationship between PhenoAgeAccel and prognosis in cancer survivors. Our findings suggest that PhenoAgeAccel could have great potential as a simple and low-cost prognostic tool in the clinical management of cancer survivors.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Luo Lei, Xin Sui contributed to the research design. Luo Lei, Jixin Fu, Xinjian Wang and Xin Sui contributed to data collection, data processing and graphing. Luo Lei, Xin Sui contributed to the writing of the manuscript. Xiao Bing and Jingchao Lu contributed to review and to edit. All authors have read and approved the final manuscript.

Data availability

Publicly available datasets were analyzed in this study. This data can be found here: The National Health and Nutrition Examination Survey dataset at https://www.cdc.gov/nchs/nhanes.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

The studies involving human participants were reviewed and approved by The National Center for Health Statistics (NCHS) Research Ethics Review Board. The patients/participants provided their written informed consent to participate in this study.