Genetic and phenotypic associations of frailty with cardiovascular indicators and behavioral characteristics

Abstract

Introduction: Frailty Index (FI) is a common measure of frailty, which has been advocated as a routine clinical test by many guidelines. The genetic and phenotypic relationships of FI with cardiovascular indicators (CIs) and behavioral characteristics (BCs) are unclear, which has hampered ability to monitor FI using easily collected data.

Objectives: This study is designed to investigate the genetic and phenotypic associations of frailty with CIs and BCs, and further to construct a model to predict FI.

Method: Genetic relationships of FI with 288 CIs and 90 BCs were assessed by the cross-trait LD score regression (LDSC) and Mendelian randomization (MR). The phenotypic data of these CIs and BCs were integrated with a machine-learning model to predict FI of individuals in UK-biobank. The relationships of the predicted FI with risks of type 2 diabetes (T2D) and neurodegenerative diseases were tested by the Kaplan-Meier estimator and Cox proportional hazards model.

Results: MR revealed putative causal effects of seven CIs and eight BCs on FI. These CIs and BCs were integrated to establish a model for predicting FI. The predicted FI is significantly correlated with the observed FI (Pearson correlation coefficient = 0.660, P-value = 4.96 × 10^-62). The prediction model indicated "usual walking pace" contributes the most to prediction. Patients who were predicted with high FI are in significantly higher risk of T2D (HR = 2.635, P < 2 × 10^-16) and neurodegenerative diseases (HR = 2.307, P = 1.62 × 10^-3) than other patients.

Conclusion: This study supports associations of FI with CIs and BCs from genetic and phenotypic perspectives. The model that is developed by integrating easily collected CIs and BCs data in predicting FI has the potential to monitor disease risk.

Keywords: Behavioural characteristics; Electrocardiography; Frailty Index; Genome-wide association study; Mendelian randomization.

Graphical abstract

Fig. 1

A brief overview of the study. a. The observed frailty index (Rockwood Index) was calculated for each sample from the UK-Biobank using the cumulative deficits method, which includes 49 items from 11 categories. b. By leveraging the GWAS summary data of the Rockwood-Index based frailty index (FI) and cardiovascular indicators, the heritability and genetic correlations of cardiovascular indicators and FI were estimated through LDSC. A regression was performed to estimate the correlation between polygenic risk scores (PRS) for cardiovascular indicators and the observed FI. Four MR methods were used to infer the causal relationship between FI and cardiovascular indicators. c. The heritability and genetic correlations of behavioral characteristics and the observed FI were estimated by LDSC. The correlation between PRS for the behavioral characteristics and FI was estimated by regression model. The causal relationship between FI and behavioral characteristics was estimated by Mendelian randomization (MR). d. The behavioral characteristics and cardiovascular indicators were integrated with the PRS for FI to build the XGBoost model for predicting FI. The performance of the model was evaluated by the Pearson Correlation Coefficient (PCC) between the predicted FI and the observed FI. The relationships of the predicted FI and the risk type 2 diabetes and the risk of neurodegenerative disease of the participants were analyzed by Cox model. FI: frailty index; LDSC: linkage disequilibrium score regression; GSMR: generalized summary-data-based Mendelian randomization; IVW: inverse-variance weighted; CAUSE: causal analysis using summary effect estimates; LCV: latent causal variable model. The specifics of the data sources for each component are detailed in Sup. Fig. 1.

Fig. 2

Using linkage disequilibrium score regression (LDSC) to estimate the heritability (a, b) of cardiovascular indicators, and their genetic correlations (c, d) with FI. The blue radial column plots show the heritability of 12-lead ECG (a) or circulating metabolites (b). The length of the bar represents the magnitude of the heritability of the trait, with darker bars representing smaller corrected p-values, and bars with P_adjust < 0.05 are marked with “*”. Orange column plots show the genetic correlation of 12-lead ECG (c) and circulating metabolites (d) with FI, respectively; only traits significantly correlated with FI are plotted. Error bars indicate 95 % confidence intervals (CIs) for the estimated genetic correlations. The abbreviated meanings of the cardiovascular indicators and the full results of the LDSC are given in the Supplementary Table S5–10. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Estimating genetic causal effects of behavioral characteristics (a) and cardiovascular indicators (b) on FI using by MR methods. Error bars indicate 95% confidence intervals (CIs) for the associated MR point estimates. nSNP, number of genetic variants used as instrumental variables. The horizontal axis of the forest plot is the odds ratio (OR). GSMR: generalized summary-data-based Mendelian randomization; IVW: inverse-variance weighted; CAUSE: causal analysis using summary effect estimates. Detailed data from the MR analyses are provided in Supplementary Table S11–12.

Fig. 3

Estimating genetic causal effects of behavioral characteristics (a) and cardiovascular indicators (b) on FI using by MR methods. Error bars indicate 95% confidence intervals (CIs) for the associated MR point estimates. nSNP, number of genetic variants used as instrumental variables. The horizontal axis of the forest plot is the odds ratio (OR). GSMR: generalized summary-data-based Mendelian randomization; IVW: inverse-variance weighted; CAUSE: causal analysis using summary effect estimates. Detailed data from the MR analyses are provided in Supplementary Table S11–12.

Fig. 4

The coefficients of lifestyles, physical measures, ECG characteristics, and circulating metabolites with Frailty Index (FI) for individuals aged 60–70 through linear regression analysis. It provides a visual representation of the associations between behavioral characteristics and FI (Frailty Index) across the different age groups. Error bars show 95% confidence intervals (CIs). The horizontal axis of the forest plot is the coefficients (β).

Fig. 5

Pearson correlation coefficient of electrocardiography and behavioral characteristics with FI. (a) Pearson's correlation of electrocardiography and behavioral characteristics with Frailty Index (FI) for individuals aged 60–70. The correlation between behavioral traits and FI (Frailty Index) for aged 60–70(a) and aged 40–60(b). The red color indicates a positive correlation and the blue color indicates a negative correlation, the exact values are marked in the upper half triangle. The study included participants aged 60–70. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 6

Kaplan-Meier survival curves for Frailty Index (FI) and predicted Frailty Index (FI-pred). a. KM curve to evaluate the relationship between the risk of type 2 diabetes occurrence and FI. b. KM curve to evaluate the relationship between the risk of type 2 diabetes occurrence and the predicted FI (FI-pred). c. KM curve to evaluate the relationship between the risk of neurodegenerative disease and FI. d. KM curve to evaluate the relationship between the risk of neurodegenerative disease and the predicted FI (FI-pred). HR, the ratio of hazards.