Predictive value for the rural Chinese population of the Framingham hypertension risk model: results from Liaoning Province.

Background: A prediction model from the US Framingham Heart Study (FHS) population has been established to estimate an individual's risk of developing hypertension. However, this model has not been widely tested in other cohorts. In this study, we examined the predictive capability of the FHS prediction model in a rural Chinese population.
Methods: A total of 24,434 rural Chinese adults aged ≥35 years, without prevalent hypertension, diabetes mellitus, stroke, and coronary heart disease at baseline, were followed for the incidence of hypertension. Standard clinical examinations of blood pressure, weight and height, smoking status, and parental history of hypertension were observed biennially.
Results: The mean age was 47.9 (SD = 10.2) years, and 49.5% of subjects were women. During a median 4.8 years of follow-up, we recorded a total of 8,675 incident hypertension cases. The cumulative 2-year and 4-year hypertension incidence rates were 7.7% and 25.6%, respectively. The C statistics for the 2-year and 4-year incidences of hypertension were 0.537 (95% confidence interval (CI) = 0.524-0.550) and 0.610 (95% CI = 0.602-0.618) for the FHS model, respectively. The Hosmer-Lemeshow χ(2) test results for 2-year and 4-year incidence of hypertension were 2,287.7 (P < 0.0001) and 8,227.1 (P < 0.0001), respectively. Sensitivity analysis indicates that the FHS prediction model still has a poor performance, although the predictive ability was better than for the overall population.
Conclusions: The FHS hypertension prediction model is not a valid tool with which to estimate the risk of incidence of hypertension among the rural Chinese population. A new hypertension risk equation for the rural Chinese population is needed.