Background: With few exceptions, tools used to estimate cardiovascular disease (CVD) risk in those without prior events are based mainly on data from middle-aged subjects. Given the ever increasing number of older people, many with hypertension, a risk score relevant to this group is warranted. Our aim was to develop a cardiovascular risk equation suitable for risk prediction in elderly, hypertensive populations.
Methods: We utilized cardiovascular end point data from 4.1 years median follow-up in 5,426 hypertensive subjects without previous CVD from the Second Australian National Blood Pressure Study (ANBP2). Our risk model, based on Cox regression, was developed using 75% of subjects without evident CVD (n = 4,072), randomly selected and stratified by age and gender, and internally validated using the remaining 25%. The model was also externally validated against the Dubbo Study dataset.
Results: The final model included sex, age, physical activity in the 2 weeks prior to entry into study, family history, use of anticoagulants, centrally acting antihypertensive agents or diabetes medication, and an interaction term for sex and diabetes medication. The C-statistic was 0.65 (0.62-0.67) for our predictive model on the model development dataset and 0.62 (0.57-0.67) on the internal validation dataset. The Dubbo Data C-statistic for CVD was 0.68 (95% CI 0.65-0.71).
Conclusions: All models performed similarly. Because of greater ease of implementation, we recommend that existing algorithms be extended into older age groups.