BACKGROUND: Responder thresholds facilitate the interpretation of meaningful within-person change in clinical outcome assessment (COA) scores and support development of label claims. The FDA PRO guidance (2009) recommends the use of an anchor-based method as the primary method for estimating meaningful change. Increasingly, cumulative distribution function (CDF) and probability density function (PDF) plots are produced to illustrate the distribution of change in COA scores. There is limited COA-specific literature on interpreting CDF and PDF plots and their use to support estimates of meaningful within-person change.
OBJECTIVE: To demonstrate the interpretation of CDF and PDF plots, as well as their complementary strengths and limitations.
METHODS: Simulated data mimicking COA results from clinical trials will illustrate and compare the information provided by CDF and PDF plots. Examples will explore the effect of sample sizes, the strength of the association between the anchor measure and change in COA, and the smoothing of the distributions on the interpretability of both plots. Interpretation of each plot will be discussed, as will the strengths and limitations of each display.
IMPLICATIONS: Examples will help clinical trial designers and researchers better understand the type of evidence CDF and PDF plots can present and their various uses.
CONCLUSIONS: PDF and CDF plots are valuable graphical tools that can be used to support responder threshold estimation and facilitate the interpretation of meaningful change from the patient perspective in clinical trials. While CDFs provide a direct visualization of the separation between anchor levels at the median, PDFs enhance the evaluation of the appropriateness of the anchors. In most cases, the information provided in CDF and PDF plots is complementary and we recommend that sponsors and researchers provide both in regulatory communications in order to enhance the interpretation of change in COA scores.
