CDFs and PDFs: Comparing Visual Methods Supporting Responder Thresholds in Clinical Trials
Qin S, Coles T, Nelson L, McLeod L, Williams V, Williams N, Reaney M
Return to RTI-HS ISPOR Baltimore webpage to see what else was presented
During our project work, cumulative distribution function (or CDF) and probability density function (or PDF) plots are often requested by FDA. They are used to support anchor-based, clinically-meaningful within-person change, or responder definition. However, the guidance on the generation and interpretation of these plots are not widely available or discussed, so confusion sometimes arises among medical researchers. For this reason, our group prepared a poster to provide suggestions on how to generate and interpret those plots. We also discussed their strengths and limitations.
In our poster, we show how CDF and PDF plots are mathematically related. Example plots were provided to show how they would appear in FDA label claim submissions. Both ideal and suboptimal cases were discussed and popular figure generating methodologies were presented. In sum, both plots are very useful in evaluating responder definitions. To make the best of them, we must be careful in method choices and be clear on our rationale behind the choices.