Abstract Background Recent studies showed that lipoprotein(a) Lp(a) is a causal risk factor for cardiovascular disease (CVD). However, whether Lp(a) modifies clinical risk assessment was not established. Objectives This study was conducted to determine whether Lp(a) improves CVD risk prediction. Methods In 1995, Lp(a) was measured in 826 men and women (age range, 45 to 84 years) from the general community. Incidence of CVD was recorded over 15 years of follow-up. Results In models adjusted for Framingham Risk Score (FRS) and Reynolds Risk Score (RRS) variables, the hazard ratio (HR) for incident CVD was 1.37 per 1-SD higher Lp(a) level (SD = 32 mg/dl) and 2.37 when comparing the top fifth quintile with other quintiles. The addition of Lp(a) to the RRS increased the C-index by 0.016. Of the 502 subjects who remained free of CVD, 82 were correctly reclassified to a lower risk category and 49 were reclassified to a higher risk category (predicted 15-year categories: <7.5%, 7.5% to <15%, 15% to <30%, ≥30%) (p < 0.001). Of the 148 subjects who developed CVD, 18 were correctly reclassified to a higher risk category and 17 were reclassified to a lower risk category. In subjects at intermediate risk (15% to <30%), the net reclassification improvement afforded by Lp(a) was 22.5% for noncases, 17.1% for cases, and 39.6% overall. Allele-specific Lp(a) levels did not add to the predictive ability of the FRS or RRS or to Lp(a). Conclusions Elevated Lp(a) predicts 15-year CVD outcomes and improves CVD risk prediction. These findings suggest that Lp(a) levels may be used in risk assessment of subjects in the general community, particularly in intermediate-risk groups.