DNA methylation-based measures of biological age: meta-analysis predicting time to death

Abstract

Estimates of biological age based on DNA methylation patterns, often referred to as "epigenetic age", "DNAm age", have been shown to be robust biomarkers of age in humans. We previously demonstrated that independent of chronological age, epigenetic age assessed in blood predicted all-cause mortality in four human cohorts. Here, we expanded our original observation to 13 different cohorts for a total sample size of 13,089 individuals, including three racial/ethnic groups. In addition, we examined whether incorporating information on blood cell composition into the epigenetic age metrics improves their predictive power for mortality. All considered measures of epigenetic age acceleration were predictive of mortality (p≤8.2x10-9), independent of chronological age, even after adjusting for additional risk factors (p<5.4x10-4), and within the racial/ethnic groups that we examined (non-Hispanic whites, Hispanics, African Americans). Epigenetic age estimates that incorporated information on blood cell composition led to the smallest p-values for time to death (p=7.5x10-43). Overall, this study a) strengthens the evidence that epigenetic age predicts all-cause mortality above and beyond chronological age and traditional risk factors, and b) demonstrates that epigenetic age estimates that incorporate information on blood cell counts lead to highly significant associations with all-cause mortality.

Keywords: DNA methylation; all-cause mortality; epigenetic clock; epigenetics; lifespan; mortality.

Conflict of interest statement

The Regents of the University of California is the sole owner of a provisional patent application directed at the invention of measures of epigenetic age acceleration for which SH is a named inventor. The other authors declare no conflicts of interest.

Figures

Figure 1. Epigenetic age acceleration in blood versus that in breast or saliva

(A-D) Epigenetic age acceleration in healthy female breast tissue (y-axis) versus various measures of epigenetic age acceleration in blood: (A) universal measure of age acceleration in blood, (B) intrinsic epigenetic age acceleration based on the Horvath estimate of epigenetic age, (C) extrinsic epigenetic age acceleration, (D) intrinsic epigenetic age acceleration based on the Hannum estimate of epigenetic age. (E-H) analogous plots for epigenetic age acceleration in saliva (y-axis) and (E) AgeAccel, (F) IEAA based on Horvath, (G) EEAA, (H) IEAA based on the Hannum estimate. The y-axis of each plot represents the universal measure of age acceleration defined as the raw residual resulting from regressing epigenetic age (based on Horvath) on chronological age.

Figure 2. Univariate Cox regression meta-analysis of all-cause mortality

A univariate Cox regression model was used to relate the censored survival time (time to all-cause mortality) to (A) the universal measure of age acceleration (AgeAccel), (B) intrinsic epigenetic age acceleration (IEAA), (C) extrinsic epigenetic age acceleration (EEAA). The rows correspond to the different cohorts. Each row depicts the hazard ratio and a 95% confidence interval. The coefficient estimates from the respective studies were meta-analyzed using a fixed-effect model weighted by inverse variance (implemented in the metafor R package [34]). It is not appropriate to compare the hazard ratios and confidence intervals of the different measures directly because the measures have different scales/distributions. However, it is appropriate to compare the meta analysis p values (red sub-title of each plot). The p-value of the heterogeneity test (Cochran's Q-test) is significant if the cohort-specific estimates differed substantially.

Figure 3. Multivariate Cox regression meta-analysis adjusted for clinical covariates

A multivariate Cox regression model was used to relate the censored survival time (time to all-cause mortality) to (A) the universal measure of age acceleration (AgeAccel), (B) intrinsic epigenetic age acceleration (IEAA), (C) extrinsic epigenetic age acceleration (EEAA). The multivariate Cox regression model included the following additional covariates: chronological age, body mass index (category), educational level (category), alcohol intake, smoking pack years, prior history of diabetes, prior history of cancer, hypertension status, recreational physical activity (category). The rows correspond to the different cohorts. Each row depicts the hazard ratio and a 95% confidence interval. The coefficient estimates from the respective studies were meta-analyzed using a fixed-effect model weighted by inverse variance (implemented in the metafor R package [34]). The sub-title of each plot reports the meta-analysis p-value and a heterogeneity test p-value (Cochran's Q-test).

Figure 4. Hazard ratio of death versus cohort characteristics

Each circle corresponds to a cohort (data set). Circle sizes correspond to the square root of the number of observed deaths, because the statistical power of a Cox model is determined by the number of observed deaths. (A-C) The y-axis of each panel corresponds to the natural log of the hazard ratio (ln HR) of a univariate Cox regression model for all-cause mortality. Each panel corresponds to a different measure of epigenetic age acceleration (A) universal age acceleration, (B) intrinsic age acceleration, (C) extrinsic age acceleration. Panels (D-F) are analogous to those in A-C but the x-axis corresponds to the median age of the subjects at baseline (Table 1). The title of each panel reports the Wald test statistic (T) and corresponding p-value resulting from a weighted linear regression model (y regressed on x) where each point (data set) is weighted by the square root of the number of observed deaths. The dotted red line represents the regression line. The black solid line represents the line of identify (i.e., no association).