Sex-specific threshold levels of plasma B-type natriuretic peptide for prediction of cardiovascular event risk in a Japanese population initially free of cardiovascular disease.

Elevated plasma B-type natriuretic peptide (BNP) levels have been reported to be related to a high risk for cardiovascular (CV) disease in the general population. However, there has been no accurate determination of the threshold levels of plasma BNP that indicate an increased potential for the development of general CV events (i.e., heart failure, stroke, and myocardial infarction) and the validity of these levels for predicting CV events compared to classic risk markers. To establish gender-specific thresholds of plasma BNP levels associated with increased risk for CV disease in the general population, baseline BNP levels were determined in community-dwelling adults (n = 13,209, mean age 62 ± 10 years,) and CV events in the cohort were captured prospectively. The cohort was divided by deciles of plasma BNP level in each gender. A Cox proportional-hazards model was used to determine the relative hazard ratios among the deciles. In addition, to compare the utility of plasma BNP to the Framingham 10-year risk score for predicting general CV events, receiver-operating characteristic analysis was performed. During follow-up, CV events were identified in 429 patients in the cohort. Compared to the reference decile level (first to fourth), the hazard ratio was significantly increased from the ninth decile in men (greater than approximately 37 pg/ml) and the highest decile in women (greater than approximately 55 pg/ml). The area under the curve generated on receiver-operating characteristic analysis of plasma BNP testing was comparable to that for the Framingham risk scoring system (0.67 vs 0.68 in men, 0.63 vs 0.68 in women; p = NS for both). In conclusion, within a community-based general population with no CV history, plasma BNP levels higher than defined thresholds show increased risk for general CV events, and the predictive ability for CV events occurring within several years may be comparable to that of an established long-standing risk score.