22 Exposure-Response Modeling: Methods and Practical Implementation
= 0.215) for a 1000 ng/mL concentration incr ease, which shows a significant
increase of the risk of 20 ms prolongation at the 5% level. However, when fit-
ting the model to the moxifloxacin data only, the log-OR becomes -0.104 (SE
= 0.557), indicating no ER relationship. In fact, this situation is not uncom-
mon in exposure–re sponse modeling when control (i.e., no exposure) data are
available, since fitting the same model with and without control data may lead
to co mpletely different results. The difference often sugg ests that the dose–
response relationship in the whole exposure range is complex and may need
a more complex model. A simple model is likely to be misspecified at some
exp ...