Search

Additional file 1: Table S1. T cell subset definitions measured in the Health and Retirement Study. Figure S1. Pearson correlation heatmap between individual T-cell subsets measured in the Health and Retirement Study. Figure S2. Association of individual T-cell subsets with chronological age. Age is used as the dependent variable in survey linear regression models. The beta estimates are estimated per one standard deviation unit of individual T-cell subsets. The models are adjusted for sex, race/ethnicity, and CMV status. The solid black line along 0 indicates no association. Figure S3. Scatterplot of ARIP measures with chronological age with a LOWESS curve. Figure S4. Association of individual T-cell subsets with biological age. Biological age is used as the dependent variable in survey linear regression models. The beta estimates are estimated per one standard deviation unit of individual T-cell subsets. The models are adjusted for chronological sex, race/ethnicity, and CMV status. The solid black line along 0 indicates no association. Figure S5. Scatterplot of ARIP measures with biological age with a LOWESS curve. Table S2. Beta estimates of the association between the additional ARIP measures with chronological age and biological age. Table S3. Beta estimates of the association between ARIP measures and biological age acceleration measure after adjustment for age, sex, race/ethnicity, and CMV status. Figure S6. Odds ratios and 95% CI of association of multimorbidity levels with individual T-cell subsets per one SD unit increase in ARIP marker. Adjusted for age, sex, race/ethnicity, CMV status, smoking status, and BMI. The solid black line along OR of 1 indicates no association. Table S4. Estimates of association of immunosenescence measures with age-related outcomes. Table S5. Associations between self-rated health and ARIP measures after adjustment for age, sex, race/ethnicity, CMV status, smoking status, and BMI.

Background: Compared to cancer-free persons, cancer survivors of the same chronological age (CA) have increased physiological dysfunction, i.e., higher biological age (BA), which may lead to higher morbidity and mortality. We estimated BA using eight aging metrics: BA computed by Klemera Doubal method (KDM-BA), phenotypic age (PhenoAge), five epigenetic clocks (ECs, Horvath, Hannum, Levine, GrimAge, and pace of aging (POA)), and subjective age (SA). We tested if aging constructs were associated with total cancer prevalence and all-cause mortality in cancer survivors and controls, i.e., cancer-free persons, in the Health and Retirement Study (HRS), a large population-based study., Methods: In 2016, data on BA-KDM, PhenoAge, and SA were available for 946 cancer survivors and 4,555 controls; data for the five ECs were available for 582 cancer survivors and 2,805 controls. Weighted logistic regression was used to estimate the association between each aging construct and cancer prevalence (odds ratio, OR, 95%CI). Weighted Cox proportional hazards regression was used to estimate the associations between each aging construct and cancer incidence as well as all-cause mortality (hazard ratio, HR, 95%CI). To study all BA metrics (except for POA) independent of CA, we estimated age acceleration as residuals of BA regressed on CA., Results: Age acceleration for each aging construct and POA were higher in cancer survivors than controls. In a multivariable-adjusted model, five aging constructs (age acceleration for Hannum, Horvath, Levine, GrimAge, and SA) were associated with cancer prevalence. Among all cancer survivors, age acceleration for PhenoAge and four ECs (Hannum, Horvath, Levine, and GrimAge), was associated with higher all-cause mortality over 4 years of follow-up. PhenoAge, Hannum, and GrimAge were also associated with all-cause mortality in controls. The highest HR was observed for GrimAge acceleration in cancer survivors: 2.03 (95% CI, 1.58-2.60). In contrast, acceleration for KDM-BA and POA was significantly associated with mortality in controls but not in cancer survivors. When all eight aging constructs were included in the same model, two of them (Levine and GrimAge) were significantly associated with mortality among cancers survivors. None of the aging constructs were associated with cancer incidence., Conclusion: Variations in the associations between aging constructs and mortality in cancer survivors and controls suggests that aging constructs may capture different aspects of aging and that cancer survivors may be experiencing age-related physiologic dysfunctions differently than controls. Future work should evaluate how these aging constructs predict mortality for specific cancer types.