Walking Pace Optimizes Conventional Cardiovascular Disease Risk Prediction Models Among Vulnerable Subpopulations: A Prospective Cohort Study

Figures

• 

  Figure 1. Flow chart of the study population for UKB (A) and ELSA (B). CVD indicates cardiovascular disease; ELSA, English Longitudinal Study of Ageing; and UKB, UK Biobank.

• 

  Figure 2. The association of grip strength and walking pace with incident CVD in the UKB. HRs for “continuous grip strength” are per 1‐kg increase. HRs for “low grip strength” are vs “normal grip strength.” Model 1 was adjusted for age and sex. Model 2 was adjusted for age, sex, race, smoking status, drinking status, systolic blood pressure, Townsend deprivation index, and family history of CVD. CVD indicates cardiovascular disease; HR, hazard ratio; and UKB, UK Biobank.

• 

  Figure 3. Comparison of C index from models predicting incident CVD according to sex, age, grip strength, and walking pace. The radar plot depicts C indexes for the 4 conventional CVD risk prediction models–Framingham CVD risk equation (black), Reynolds risk score (green), ASSIGN score (orange), and pooled cohort equations (blue)–in total population and 24 subpopulations. Each axis in the radar plot represents total population or 1 subpopulation. Each dashed concentric circle corresponds to a specific value of C index. C indexes from different subpopulations are plotted along each axis and connected to form a polygonal circular curve. Bold represents the subpopulation with inferior C index relative to its comparison subpopulation, for example, inferior C index using ASSIGN model in the older subpopulation (0.624 [95% CI, 0.614–0.634]) vs young subpopulation (0.701 [95% CI, 0.689–0.712]). ASSIGN indicates ; Low grip, subpopulation with low grip strength; Slow pace, subpopulation with slow walking pace; Old_Low grip, older population with low grip strength; ASSIGN, ASSessing cardiovascular risk using Scottish Intercollegiate Guidelines Network; and PCE, pooled cohort equation.

• 

  Figure 4. The calibrations of risk prediction models in UKB. A, Calibration plots for observed vs predicted risk of CVD as estimated using the Framingham, Reynolds, ASSIGN, and PCEs model in the overall population, respectively. The dashed line reflects the perfect agreement between observed and predicted risk. B, Mean values and 95% CIs of calibration slopes of the ASSIGN model. The red horizontal dashed line at 1 denotes the ideal calibration reference. C, Mean values and 95% CIs of calibration intercepts of the ASSIGN model. The red horizontal dashed line at 0 signifies no systematic bias (average predicted risks matching observed event rates). Error bars on the bar chart represent the 95% CIs. ASSIGN indicates ; CVD, cardiovascular disease; ASSIGN, ASSessing cardiovascular risk using Scottish Intercollegiate Guidelines Network; PCE, pooled cohort equation; and UKB, UK Biobank.

• 

  Figure 5. Heat maps displaying C index changes after adding grip strength and walking pace to the ASSIGN model in UKB. The values of C index changes across different subpopulations were scaled, with darker colors indicating larger values. n=206 371 in total. *P<0.05; **P<0.01; ***P<0.001. ASSIGN, ASSessing cardiovascular risk using Scottish Intercollegiate Guidelines Network; UKB, UK Biobank.

• 

  Figure 6. Heat maps displaying C index changes after adding grip strength and walking pace to the ASSIGN model in ELSA. The values of C index changes across different subpopulations were scaled, with darker colors indicating larger values. N=5506 in total. ASSIGN indicates ASSessing cardiovascular risk using Scottish Intercollegiate Guidelines Network; and ELSA, English Longitudinal Study of Ageing.