OBJECTIVES: This research presents the lessons learned of a 5-year research project aiming to increase trust and confidence in selecting and interpreting results related to commonly applied structural approaches used in health economic obesity models.

METHODS: A three-step process was followed. First a systematic review was performed to determine commonly applied health economic modelling approaches. Second a replication of high-quality health economic models was performed in TreeAge and the reproduction success was evaluated. Third, using the successfully replicated models, we performed an external validation and health economic result comparison using state of the art methods.

RESULTS: Of 87 included papers reflecting health economic obesity models, 69 (79%) applied unique modelling approaches, but only for a minority (14%) an external validation was performed. Most commonly simulated obesity associated events were coronary heart disease (83%), type 2 diabetes (74%) and stroke (65%). The most commonly event simulation approaches were either using specific risk equations (influenced by various risk factors), or simpler body-mass-index (BMI) based approaches, either using BMI-related relative risks (RR) or using risk function that includes BMI as central variable. We were able to successfully reproduce the results of three models reflecting those three main approaches. In the external event validation, the risk function approach showed the best results, followed by the BMI-related RR risk and the BMI-function approaches. Although a variation in the mean incremental cost effectiveness ratio was identified, there was no statistically significant difference identified between the approaches.

CONCLUSIONS: This research identified important aspects related to health economic model reporting, reproduction and validation in general, and key aspects specifically related to obesity associated event simulation. Our research could form a basis for evaluating the strengths and weaknesses of different structural event simulation approaches and identified important areas for further research to improve model reporting, reproducibility and credibility.