OBJECTIVES: The standard approach to derive transition probabilities for use in state-transition cost-effectiveness models based on available risk functions with different time horizon than the model cycle length is to use an exponential distribution (constant event rate assumption) to derive model cycle risk. However this does not remove any age effect on risk, resulting in potential overestimation of cycle-specific risks in the case age is included as a risk factor. We present an approach for annualization of event risk predictions that ensures the correct risk is predicted. METHODS: The proposed approach ensures that the starting age-specific event rate multiplied by the cumulative effect of increasing age during the time horizon of the risk function equals the original risk prediction. To illustrate and compare the standard and the proposed approach, 10-year risk is modelled and compared to the initial 10-year predicted risk using the latest Framingham risk function for primary cardiovascular (CVD) event risk (D’Agostino et al. 2008). Scenarios are investigated using different starting age and different initial predicted risk level. RESULTS: The standard risk annualization approach consistently overestimates the modelled risk over a 10-year horizon compared with the initial risk prediction for the same period. To illustrate, for a 50-year old male with initial 10-year CVD risk of 20%, 30% or 40% the corresponding modelled risk is 25%, 37% and 49%, respectively. The proposed method for risk annualization provides the exact 10-year initial risk. CONCLUSIONS: Careful application of approach for time period adjustment of existing risk functions for use in state-transition models is needed when age is included as a risk factor and the model cycle length differs from the time horizon of the risk function. The standard risk annualization approach will lead to an overestimation of modelled risk and a more favorable estimate of cost-effectiveness.