Walking pace optimizes conventional cardiovascular disease risk prediction models among vulnerable subpopulations: a prospective cohort study

BackgroundGrip strength and walking pace are increasingly recognized as potential indicators of adverse health outcomes for general population. However, their clinical utility in the primary prevention of cardiovascular disease (CVD) amoung various subpopulations remains uncertain. This study aimed to evaluate performances of four conventional CVD prediction models across subpopulations with varying grip strength or walking pace and determine added predictive value of grip strength and walking pace. MethodsA total of 206,371 individuals without CVD (aged 40-69 years) from UK Biobank (UKB) were included. Four conventional CVD prediction models (Framingham, Reynolds, ASSIGN, and Pooled Cohort Equations) were used to estimate 10-year CVD risk. Model performance was compared across diverse subpopulations defined by age, grip strength, or walking pace using Harrells concordance index (C index) and calibration plot in UKB. Added predictive value of grip strength and walking pace was evaluated using C index change and net reclassification improvement (NRI) in UKB and further validated in English Longitudinal Study of Ageing (ELSA). Results19,664 cases of incident CVD were registered during follow-up period (mean 12.85 years [standard deviation 2.74]). All four models performed inferior among vulnerable subpopulations characterized by advanced age, low grip strength or slow walking pace. C index for ASSIGN model was 0.659 (95%CI, 0.646-0.672) in low grip strength subpopulation vs 0.702 (95%CI, 0.699-0.706) in normal grip strength subpopulation, 0.646 (95%CI, 0.635-0.657) in slow walking pace subpopulation vs 0.701 (95%CI, 0.698-0.705) in normal walking pace subpopulation, and 0.624 (95%CI, 0.614-0.634) in old subpopulation vs 0.701 (95%CI, 0.689-0.712) in young subpopulation. Adding walking pace to ASSIGN model improved its discriminative ability in vulnerable subpopulations, i.e., those with low grip strength (NRI, 0.046 [95%CI, 0.009-0.100]) and old subpopulation (NRI, 0.035 [95%CI, 0.020-0.048]). However, grip strength did not show significant added predictive value for CVD in these subpopulations. The finding was similar in ELSA. ConclusionsConventional CVD prediction models underperformed in vulnerable subpopulations; Walking pace, but not grip strength, conferred substantial incremental information to these models, paricularly those with advanced age or low grip strength. This current finding offers novel insights into the role of physical function in the primary prevention of CVD. Clinical PerspectiveO_ST_ABSWhat Is New?C_ST_ABSO_LIConventional CVD risk models did not work well in vulnerable subpopulations characterized by advanced age, low grip strength or slow walking pace. C_LIO_LIWalking pace, but not grip strength, conferred substantial incremental information to the established CVD risk models in vulnerable subpopulations, especially those with advanced age or low grip strength. C_LI What Are the Clinical Implications?O_LIWalking pace, a simple, easy-to-implement and low-cost measures, has the potential to better predict CVD risk in vulnerable subpopulations, supporting more accurate identification of high-risk population, personalization of care plans, and equitable health policies. C_LI