I am requesting the addition of confidence intervals to Relative Potency report in Parallelism Test in the Fit Model platform for all models with Parallelism Test, especially all the 4 and 5p sigmoidal cures.
This is important for US and European Pharmaceutical that use a biological test method for measuring potency. Potency is measured on every lot of drug substance and drug product. Comparison of a reference lot to a new lot is very common, and calculations are daunting for most scientists who are required to routinely assess potency.
Request for confidence intervals on potency are have become more frequent and answers in the community related to this topic are vague and not particularly actionable or practical.
The EP guidance document says this about CI's for potency
"The maximum likelihood estimators of the slope and location of the curves can be found only by applying an iterative procedure. There are many procedures which lead to the same outcome, but they differ in efficiency due to the speed of convergence. One of the most rapid methods is direct optimisation of the maximum-likelihood function (see Section 7.1), which can easily be performed with computer programs having a built-in procedure for this purpose. Unfortunately, most of these procedures do not yield an estimate of the confidence interval, and the technique to obtain it is too complicated to describe here." . p 580, link to document attached below
Recently (2019) Don McCormack presented the alternative method of using Probit to estimate CI's for relative potency, an assumption heavy method which is also described in the the EP guidance. (link to Don's talk below). While Don's paper and attached examples directly from EP5.3 are very helpful, the method is sufficiently complex to cause the common JMP user pause. The method also requires the use of other platforms to obtain a result. (user cannot stay in the flow of analysis)
EP 5.3 STATISTICAL ANALYSIS OF RESULTS OF BIOLOGICAL ASSAYS AND TESTS
Using JMP for Statistical Analysis of Biological Assays - Discovery Paper, Don McCormack
