OBJECTIVES: Economic modeling is an established tool used for allocation of health care resources. Modeling was designed to demonstrate the influence of variables on defined outcomes (e.g. cost-effectiveness) in complex systems. Valid information for health care decisions can be obtained if five types of bias can be avoided: selection-, performance-, attrition-, detection-, and sampling-bias. In this study the validity of results derived from economic modeling is investigated addressing these five types of possible biases. METHODS: A published economic model of costs and benefits of drug treatment in mild-to-moderate Alzheimer’s disease (Guo et al., J Med Econ 2010;13:641-654) was used for this analysis. Nine questions were asked to confirm the validity of the obtained results. Internal validity was tested by checking the first four of the above types of bias, external validity by checking for a possible sampling bias. RESULTS: The presented model is flawed by absence of an explicit study question. Selection bias cannot be excluded as the patient data were obtained from pooled clinical trials and other sources. Performance bias is likely as the outcomes in patients extracted from pooled clinical trials differed considerably to the outcomes of patients treated outside of trials. A detection bias is likely as observed data were compared with extrapolated data. Also the external validity of the study is likely to be impaired as the patients profiles were not derived from real world conditions but from patients enrolled in two clinical trials. CONCLUSIONS: This appraisal shows that phrasing a study question is essential for selection of the appropriate study method. Economic modeling is useful to discuss models and to generate hypotheses but always implies a high risk of bias. Therefore, results from modeling should only be accepted when internal as well as external validity of the used method has been confirmed.