Objectives Chronic diseases are usually slow-developing condition and their risk may result from both long-term exposure and successive exposure increments, hence calling for models accounting for dynamics of exposure and disease progression.

Method Discrete compartmental models are defined by a set of ordered states (compartments) reflecting the health status, and can be fully characterised by the set of transition probabilities between each compartment. When defined at the individual level, each participant contributes to the likelihood of the model at each year from the time of entering the initial stage (e.g. birth) to the moment they reach an absorbing state (e.g. death or clinical onset). Model estimation aims at quantifying the transitions ensuring the best reconstruction of the pathological trajectories in each subject, hence adding to the classification problem (discriminating healthy and diseased subjects) a dynamic component (estimating the time of onset).

Individual exposure histories can be summarised through cumulative exposure functions and subsequently plugged into the compartmental framework as parameters of transition probabilities.

Results While these models were initially developed to accommodate data from longitudinal studies, we will illustrate, using lung cancer case control and smoking history data, the validity and utility of such approaches. We will assess the underlying assumptions yielded by this methodological drift and will exemplify the rich statistical inference these approaches are able to provide.

Conclusions We will finally introduce potential extensions over this framework that include omics biomarkers to model genetically-driven susceptibility and/or to identify the stage (s) at which exposure (s) are more likely to mediate their effects.