Control of confounding in modeling the association between exposure to particulate matter and drug consumption: old issue, new outcome