Interpreting interaction effects in estimates of the risk of traffic injury associated with the use of illicit drugs.

Interactions characterise the relationship between use of amphetamines, cannabis and opiates as a driver and the risk of traffic injury associated with the use of these drugs. Inverse risk curves have been found for these drugs, meaning that the higher the proportion of drivers in normal traffic testing positive for the drugs, the lower is the increase in risk associated with them. The inverse risk curves can arise in many ways. The paper discusses ten different interpretations of the curves; seven of these are methodological and claim that the risk curves are statistical artefacts. Some support for these interpretations is found; however, this does not rule out that substantive interpretations, proposing causal mechanisms underlying the curves may also be correct. Unfortunately, there is insufficient evidence to assess the support for the substantive interpretations. There is, accordingly, a large element of uncertainty about how the inverse risk curves arise and whether they can be modified.