Glucocorticoid exposure predicts survival in female baboons

Abstract

Are differences in hypothalamic-pituitary-adrenal (HPA) axis activation across the adult life span linked to differences in survival? This question has been the subject of considerable debate. We analyze the link between survival and fecal glucocorticoid (GC) measures in a wild primate population, leveraging an unusually extensive longitudinal dataset of 14,173 GC measurements from 242 adult female baboons over 1634 female years. We document a powerful link between GCs and survival: Females with relatively high current GCs or high lifelong cumulative GCs face an elevated risk of death. A hypothetical female who maintained GCs in the top 90% for her age across adulthood would be expected to lose 5.4 years of life relative to a female who maintained GCs in the bottom 10% for her age. Hence, differences among individuals in HPA axis activity provide valuable prognostic information about disparities in life span.

Fig. 1. Longitudinal sampling of fGCs during the adult lives (≥age 5) of 242 female baboons included in this study.

Each female study subject is represented by a horizontal line showing periods of continuous observation of that subject, and small black crosses show fGC samples for that subject (N = 14,173). Red dots that terminate lines indicate deaths, and lines without a terminal dot indicate censored records (that is, females who were alive at the end of the observation period). The lines are ordered by age at entry into the study. The data begin after age 5 for females who were already adults when fGC sample collection began in 1999. Small gaps in the timelines of certain females represent infrequent interruptions of data collection (e.g., during political disruptions).

Fig. 2. Summary of fixed parameters in the cumulative effect joint model using area under the log fGC trajectory as a time-varying predictor of survival for adult female baboons.

For each estimate, the 50% (inner), 75% (middle), and 90% (outer) CIs are shown. (A) Posterior fixed effect parameter estimates from the linear mixed-effects submodel for ln(fGC) are listed on the y axis and standardized to the same scale. Estimates for levels of the categorical predictors “Pregnant” and “Lactating” are relative to the reproductive state “Cycling.” The categorical predictors “Alpha rank” and “Wet season” are relative to the condition of not being the alpha female and of the dry season, respectively; the other predictors are continuous. (B) Posterior fixed parameter estimates from the time-to-event submodel of the effects of time-varying covariates on the log hazard of death in adult female baboons. Social bond strength reflects the mean strength of the female’s top three social bonds with females (dyadic sociality index to females; DSI_F) and males (DSI_M); see Materials and Methods. “Area under fGC profile” is the association parameter that quantifies the predictive relationship between the area under the individual-specific fGC trajectory and the log hazard of death at the current time.

Fig. 3. Individual-specific log fGC trajectories for simulated out-of-sample individuals and dynamic predictions of survival based on current value and cumulative effect of fGC.

(A) Simulated log fGC profiles for two hypothetical adult female baboons that consistently maintain high (90th percentile) or low (10th percentile) fGC concentrations during each year of life until right censoring. The vertical dashed lines show age at entry “5” and age at censoring “14.” These trajectories were generated from the current value joint model of fGC and survival. (B) Predicted survival based on the current value (top) and cumulative effect (bottom) of fGC for the hypothetical high- and low-fGC individuals, conditional on survival to age 14, which is indicated by the vertical dashed lines. Shaded regions show 90% CIs.