Simonetto et al. Respond to "Mechanistic Models in Epidemiology"

We thank Meza and Jeon for their encouraging commentary (1) on our study (2). We fully agree with their perspective that the potential of mechanistic models in epidemiologic research is currently not sufficiently utilized. From our point of view, this is for 2 main reasons: data availability and technical manageability. Historically, primary epidemiologic data sets are devised to meet the needs of standard statistical association analysis. Mechanistic models are left with reusing these data, which may not be designed to fully support their strengths. For example, in radiation research much effort is spent to improve dosimetry and to increase case numbers for the outcome of interest by pooling radio-epidemiologic cohorts or extending their follow-up. Knowledge of concomitant behavioral risk factors is scarce, and molecular footprints or subclinical data of disease development may not be available in suitable form. Against this backdrop Kaiser et al. (3) argued for considering the use of mechanistic models from the onset and organizing data acquisition correspondingly.