The study aimed to develop improved lung cancer prediction models by updating existing single models and aggregating multiple models using external individual patient data. Methods applied to update a single model did not create an improved version, as the updated models performed worse in external validation than the original. Aggregating models using Bayesian model averaging with an informative prior led to a marginally better performing model that weighted three existing models. However, calibration of the aggregated model remained poor. In conclusion, no method reliably improved lung cancer prediction through model updating or aggregation based on this data.