True vs. Observed Effect – Part Two

Continuing the discussion regarding effect modeling that was started in this blog post. The chart shows another example of the type of analyses we can do using effect models. By changing the threshold for success, we can analyze the risk of making incorrect investment decisions.

Read: True vs. Observed Effect – Part 1

This is made possible by modeling the true effect as an unknown entity and separating the truth from what can be observed in the clinical trial - much like in a real drug development project!

The chart in the picture shows five different thresholds in the observed Response Rate needed to declare success. This is based on a case Captario had a couple of years ago working on an oncology development project. If RR is high (45%), we will increase the risk of incorrectly terminating the project but lower the risk of progressing a bad compound. By lowering the RR threshold, we increase the risk of false negatives, but we will also increase the likelihood of correctly declaring success and thus increase the probability of success.

Let's grab a virtual coffee and discuss this further!

Read also: Clinical Effect – A Key Aspect in Drug Development