Equivalence Margin Development - The Visualization tool
For all of you who haven't been there last time and who just don't have the connection: Our topic at the moment is the Equivalence Margin Development Document (from Bioassay 25). The last article (Equivalence Margin Development) first of all looked at the basics, so that today we can start directly with our topic of the heart: The EMD visualization tool.
What does the visualization tool offer us?
At this point I have to come back to my claim “Verification of QRAs” and “Visualization via Plots” are not mutually exclusive. Do you remember? They are not. But they also have nothing to do with each other, but rather stand autonomously next to each other. Verification of QRAs compares the results of our QRA documents declared as "verification data" with the tolerated margins of our test system. If you are interested there will be a detailed article here as well. We cannot stress often enough: Communicate with us, let us know your wishes. In terms of PLA, individual functions and calculations, reporting, and also the topics we cover here on the blog!
The Visualization via Plots does not use any actual assay data. Instead, it generates new random curves that would pass our test strategy with all the developed or manually entered margins.
Why should we use the tool?
Test strategies can be generated with different objectives. On the one hand, a strategy created in the EMD can serve to check test strategies already in use. The result of the investigation should clarify which of our tests are particularly relevant for our results, which are perhaps not needed at all or whether we still need other tests to support our results. On the other hand, the EMD can also be used to develop new test strategies and to compare two or more strategies with each other. Different combinations of tests and margins can be used for the same input data. For each test used, we can decide individually whether the margins calculated by the EMD (development data) should be used or whether own margins should be entered manually. This selection can also be made separately for the upper and lower limits. In this way, test systems can be designed in very small parts.
Historical data can be used to check whether our typical assays pass the individual test strategy. But: The visualization tool offers us a look beyond the end of our nose. If our interest is in exploring which more fancy curves and data our strategy would still tolerate. we are more likely to work with random curves. The visualization tool can therefore give us hints about hidden pitfalls of our test system. It is about presenting acceptable curves that may have nothing to do with our actual results. However, we can sensitize ourselves to the unwanted results that could pass our designed test system.
What are all the settings for?
The visualization tool provides us with several setting options within our PLA Content Editor. With the help of Max Configurations, we can decide how many plots of acceptable graphs the simulation should deliver at most. During our development work, we noticed that with a benchmark of 30 plots, quite good results can be achieved, and a relatively clear image of our test system can be obtained.
The curves for Standard and Test are each displayed in a common plot. A graph is always based on a combination of the needed parameters (according to the appropriate model), whereby for each of these parameters a random value is generated. Then it is checked, if this simulated configuration passes all tests of the test strategy. If that is the case, the combination of parameters is used to generate one of the visualization plots. Max Configurations should be considered here urgently, since a graph for the corresponding parameter values cannot always be found. It can happen that individual simulated parameter values will often contradict each other in tests where interactions of different parameters are relevant, e.g. ratio of parameters tests. Here Max Simulations gets its meaning. If we would enable the calculation of our document to perform any number of simulations to find a suitable curve, the duration of this action could increase immeasurably. This element therefore serves to limit the number of simulation attempts per configuration to a maximum. PLA suggests a default value of 10000 simulations.
Last but not least: Since some of the suitability tests do not work with parameter values but with confidence intervals, the standard error can be defined for each individual parameter below the element Simulation Space. Beyond that the PLA Creatable Elements Editor additionally provides the following elements for visualization: Model, Logarithm Base, Degrees of Freedom, Transformed Dose Range and Response Range. However, the use of these elements only makes sense if we work exclusively with the visualization and the margin development is not used. At the same time, this also means that we have to manually enter the lower and upper margins for our test System. In case assay references are available, the respective properties from these documents are used for the simulation and values entered here are ignored.
And now, please enjoy the possibilities and advantages of our EMD document!