Background The Framingham Risk Score (FRS) and the Pooled Cohort Equations (PCE) are established tools for the prediction of cardiovascular disease (CVD) risk. In the Western world, decreases in incidence rates of CVD were observed over the last 30 years. Thus, we hypothesise that there are also temporal trends in the risk prediction performance of the FRS and PCE from 1990 to 2000. Methods We used data from n=7789 men and women aged 40-74 years from three prospective population-based cohort studies enrolled in Southern Germany in 1989/1990, 1994/1995 and 1999/2000. 10-year CVD risk was calculated by recalibrated equations of the FRS or PCE. Calibration was evaluated by percentage of overestimation and Hosmer-Lemeshow tests. Discrimination performance was assessed by receiver operating characteristic (ROC) curves and corresponding area under the curve (AUC). Results Across the three studies, we found significant temporal trends in risk factor distributions and predicted risks by both risk scores (men: 18.0%, 15.4%, 14.9%; women: 8.7%, 11.2%, 10.8%). Furthermore, also the discrimination performance evolved differently for men (AUC PCE: 76.4, 76.1, 72.8) and women (AUC PCE: 75.9, 79.5, 80.5). Both risk scores overestimated actual CVD risk. Conclusion There are temporal trends in the performance of the FRS and PCE. Although the overall performance remains adequate, sex-specific trends have to be taken into account for further refinement of risk prediction models.