OBJECTIVES: Cost-effectiveness models should always be amendable to updating once new data become available. However, no one optimal method of synthesizing data currently exists. METHODS: We compared three different methods to pool existing parameters of a model for a chronic disease with new data: fixed effects (FE) and random effects (RE) meta-analysis and Bayesian updating (BU). These methods were applied to obtain updated estimates of transition probabilities between stable disease states and event probabilities. RESULTS: Homogeneity between studies providing “old” and “new” information was confirmed using the Q-statistic. The three methods resulted in different estimates of probabilities and their standard errors (SE). The FE model produced the lowest means and SEs for the event probabilities (moderate disease: 0.0496 (0.0028); severe disease:  0.0698(0.0026); very severe disease: 0.0892(0.0044)) and the probabilities of the event being severe (0.1009 (0.0194); 0.1032(0.0122); 0.1761(0.0206), respectively). It also produced the lowest mean and SEs for transition probabilities. The RE model resulted in the highest probabilities of getting an event: at most 20% higher than FE in very severe disease. Furthermore, the RE model resulted in the highest SEs. The SE of the probability of getting an event was at most 276% higher than in the FE model. The SE of the probability of the event being severe was at most 112% higher. The SE for the transition probabilities was at most 218% higher than in the FE model. BU resulted in the highest probability of an event being severe: up to 20% higher than FE in the case of moderate disease. BU resulted in the highest transition probabilities: up to 29% higher than FE. CONCLUSIONS: As shown, the choice of method can affect resulting model parameter updates considerably. This can affect estimates of cost-effectiveness and the uncertainty around them.