A biological age based on common clinical markers predicts health trajectory and mortality risk in dogs

Abstract

Predicting aging trajectories through biomarkers of biological aging can guide interventions that optimize healthy lifespan in humans and companion animals. Differences in physiology, genetics, nutrition, and lifestyle limit the generalization of such biomarkers and may therefore require species-specific algorithms. Here, we compared correlations of standard clinical blood parameters with survival probability in humans with those of the two most common mammalian companion animals, cats and dogs, and highlighted universal and species-specific relationships. Based on this comparative analysis, we generated and validated an algorithm that predicts biological age in canines using a longitudinal dataset with health records, blood count, and clinical chemistry from 829 dogs spanning over 12 years. Positive deviations of biological from chronological age (AgeDev) measured by this composite score significantly correlated with a decreased survival probability (hazard ratio = 1.75 per 1 year of AgeDev, p = 3.7e - 06). Importantly, in nearly half of the dogs whose biological age was accelerated by more than 1 year, none or only a single individual marker scored outside its respective reference range, suggesting practical applications for the detection of unfavorable health trajectories. Analyzing samples from a unique 14-year life-long diet restriction study, we show that restricted caloric intake lowers biological age, an effect that can be quantified at midlife years before a difference in survival is observed. Thus, a biological age clock based on clinical blood tests predicts the health trajectories of dogs for use in research and veterinary practice.

Keywords: Biological age; Biomarkers of aging; Calorie restriction; Companion animals; Comparative biology; Veterinary care.

Fig. 1

Associations of clinical laboratory markers with survival in cats, dogs, and humans. a Comparative analysis of mortality risk and blood parameters in dogs, cats, and humans. The hazard ratio for each parameter was estimated using databases of dogs and cats as well as a publicly available human dataset from the NHANES survey. The hazard ratio for an increase of 1 standard deviation (SD) of each parameter is estimated individually with age as the time variable, and the estimate is shown with 95% confidence interval (CI). For dogs, breed size based on weight and sex was included as covariates, while for cats, the parameters were used as single predictors. For humans, sex was included in the model. b Heatmap representation of hazard ratios for all-cause mortality as shown in (a) with significance indicated by asterisks. Species and parameters are clustered by Euclidian distance. *p < 0.05; **p < 0.01; ***p < 0.001. WBC white blood cell count, MCV mean cell volume, RBC red blood cell count, MCH mean corpuscular hemoglobin, MCHC mean corpuscular hemoglobin concentration, HCT hematocrit, HGB hemoglobin, PLT platelet count, GGT gamma-glutamyltransferase, AST aspartate aminotransferase, ALT alanine aminotransferase

Fig. 2

Blood marker-based composite score predicts mortality in dogs. a Application of the biological age algorithm to the testing sets of dogs consisting of animals that have not been used to identify the coefficients of the model (training set). The colors of the points indicate different breeds as shown in the legend. b Cox PH model of increased AgeDev (defined as the difference between biological age and chronological age) for mortality in dogs. HR indicates a hazard ratio for a 1-year increase in AgeDev. c Cox PH model of AgeDev quintiles for mortality in dogs. In b and c, the estimate is shown with 95% CI. HR = hazard ratio per 1 year of AgeDev

Fig. 3

Age acceleration and differences in survival despite largely normal clinical markers. a AgeDev stratified by the number of parameters outside of clinically defined normal ranges. Data are represented as boxplots showing the median line and 1st and 3rd quartiles (box) and whiskers extending 1.5 times the interquartile range (IQR) above and below the box. Individual points are overlayed with the boxplot. b Representation of the mean value (triangle) and 1st and 3rd quartile (bands) for each parameter in the different quintiles of AgeDev. The grey area represents the normal range for the given parameter. c Representation of the number of out-of-range parameters as stacked bars in each quintile of AgeDev. d ROC curve for prediction of survival at the age of 10 based on parameters measured after the age of 5. AgeDev or the number of parameters out of range was used in a Cox PH model. e and f Survival probability of 7-year-old male small (e) or medium/large (f) dogs with different values of AgeDev were predicted using a Gompertz model. A line represents the survival probability with the indicated parameters, and colored bands represent the 95% CI of the prediction. CK creatine kinase, HCT hematocrit, HGG hemoglobin, MCH mean corpuscular hemoglobin, MCV mean cell volume, WBC whole blood cell count

Fig. 4

Calorie restriction lowers biological age in dogs. a Schematic representation of the Purina Life Span Study adapted from Lawler et al. [23] (created with BioRender.com). b Evolution of AgeDev plotted over chronological age. Data are represented as boxplots showing the median line and 1st and 3rd quartiles (box) and whiskers extending 1.5 times the interquartile range (IQR) above and below the box. Data points outside of the whiskers are shown as individual dots. Lines indicate fit from a generalized additive model of AgeDev over time stratified by diet group. c AgeDev in dogs older than 7 years. Each dot represents a measure with the line indicating the mean of each group. d Percentage of parameters out of range in control (C) or restricted (R) dogs. The colors indicate the number of parameters out of range. e Representation of each parameter contributing to biological age shown as mean (triangle) and 1st and 3rd quartiles (band) for each diet group. Grey areas represent clinical reference ranges. f Cox PH ratio for AgeDev and diet effects observed in a long-term calorie restriction intervention. The estimate is shown with a 95% CI. C control, R calorie restriction, HR hazard ratio