This session is about Non-compartamental Analysis, Pharmacokinetics, and JMP Clinical. I'm Chris Kirchberg, Principal Systems Engineer of Global Technical Enablement, and with me is Meichen Dong, Senior Research Statistician Developer from the Life Science Development Group from JMP. We're here to present to you the new pharmacokinetics report that's coming in JMP Clinical 19. But first, what is pharmacokinetics and pharmacodynamics? Pharmacokinetics or PK, is really what a body does to a drug that you either take or is injected.

Those are things like how is it absorbed, how it's metabolized by the body, how it's eliminated, how it's distributed. We typically can measure that in serum or urine, for example. Pharmacodynamics is what the drug does to the body. Those are things like what effect does it have on the kidney or the liver or other body's organs? Could also be in how it affects metabolism within the body. The safety, does it do its job or efficacy? Also, those kinds of dosing-responsing relationships that we need to be aware of.

Why is the importance of PK analysis so great? To be clear, we're only going to focus on the PK part of this drug delivery system, not pharmacodynamics.

Well, PK analysis is important because it helps us ensure adequate drug exposure, looking at the toxicity levels within an organism or a person and trying to avoid them, even looking for things like bioequivalents. When people make generic drugs, we want to make sure they're just as safe and just as effective or basically equivalent.

Typically, we do these things in healthy volunteers. Sometimes we look at patient subpopulation studies to try to see if those are equivalent. It could be seen in topical drug safety as well. We want to make sure sometimes that something doesn't get absorbed. There's the reverse of this as well. These are typically where we might find these things and why we might do these types of studies.

It's a little bit of background. JMP has a tool called Fit Curve. It's a Fit Curve platform, and we can do some pharmacokinetic analysis. We have three model-based methods, one-compartment oral dose, two-compartment IV bolus dose, and even a biexponential four-parameter dose. An example of that output, you can see on the right-hand side, where we try to fit a curve based on specific assumptions. What shape it would be, what input it would be, and even if we're going to compare to test for equivalents.

But the PK parameters that are calculated are limited. There's also, as I mentioned, a lot of assumptions. We're expecting the shape of that curve to be a particular way. There's a certain mathematical model that we're expecting to do this. This might have some limitations in trying to get the information we need when doing these kinds of PK analysis.

The industry has then gravitated to something else in order to do this, something called non-compartmental analysis or NCA. The reason being is that there's minimal assumptions. It's model independent. You're taking a look at the shape of the curve itself as they take a look at the graph on the right, and we're going to just fit it. We're going to use standard tools and methods to try to extract information regardless of what that shape is. We're using the data as is.

We can calculate things like what is the area underneath the curve? Or what is the maximum peak concentration? What is the time to that peak of concentration? Or even what we'll call elimination, the half life to eliminate. Those are the kinds of things we look at, and non-compartmental analysis allows us to calculate these kinds of parameters.

We've decided that it's time to update our tools, the JMP products. In JMP Clinical, we're adding a report that does this NCA analysis, and specifically looking at the clinical trial data that one might have in a safety review, looking at domains like PC, PP, AD, PC, or ADPP domains from the CDISC data models. It allows us to do a little bit more comprehensive of a review of not just this specific information, but are there any adverse events? Are there other things that are suggestive of its efficacy? A bit more comprehensive review since we have the data.

We're going to find out there's all kinds of options and capabilities within this report. There's going to be options to separate the data or buy variables. There's going to be different methods of area under curved calculations, be able to change the time type options. We've added two different types of reporting. We've added some nice exploratory data analysis as well as the parameters that we'll calculate. Sure, about 35 different parameters. But we'll also be able to include multiple analytes. It's not just fixated on one, looking at different subgroupings, demographic information or covariance, as you might hear it called. More importantly, there's no coding needed. There are a lot of packages out there that require some level of coding, but we will give you this next exploratory and NCA analysis with no coding.

In order for us to demonstrate this today, we've had to create a new simulated study. Theophylline is this drug. It's based on this article published by Pinheiro and Bates. They had 12 individuals at 11 time points over 24 hours. We use this data to simulate 100 patients. We broke it up into two dosing groups, 300 milligrams and 400 milligrams. We wanted to sign other various attributes or demographic information such as age, sex and race and weight, as well as sites, and try to simulate a somewhat balanced demographic distribution here to give us a feel of things that we might want to look at.

Let's take a little quick look. Some of the reports we're going to be able to see are the individual plots of concentration over time for exploratory data analysis. We're going to see plots like treatment summary over time. We're also going to see the same plot, but this time via covariant that you get to choose.

Then we'll see a bunch of reports for the NCA analysis. Those reports would be things like the estimated parameters versus a covariant. Or looking at the correlation between different PK parameters and a covariant, even taking a look at the actual versus the regression-fitted values as well and be able to explore that. Then give you a table of all the estimated parameters per patient and be able to filter down with that.

Well, this gives you an introduction of what this report is, but I think it's more important to actually see it in action. Meichen, I think you're all ready. Meichen is going to give us a live demonstration of how to use this report.

Let's start with the JMP Clinical UI. Under the Reviews panel, click this Start new review. We'll get a comprehensive report from the Pharmacokinetics here. Click Okay. The report we get consists of two big sections. For the first section, we explore what concentration distribution is like over time. For the second section, we perform the non-compartmental analysis and visualize and summarize the results. Let's look at the details.

In the first section, we were inspired by the book A Picture is Worth a Thousand Tables. We use these figures to help identify unusual patterns of data from an individual's perspective and from a subgroup perspective. The repeated measures from the same person are naturally correlated. To reflect this relationship, we plot each subject's concentration over time profile together in this first figure. If your data has multiple parameters or epochs or study base, we provide this filter on the left to allow you to look at a specific scenario.

We also provide this plot by time type option to allow users to switch between planned time and actual time on the x-axis. We also want to illustrate how the individual's concentration curve compared to the general trend of the data. We include the shaded areas which represent different summary statistics. In this figure, we're looking at concentration data by treatment. The gray area is the overall range of concentration. The light blue is treatment-specific range of concentration. Dark blue is the treatment-specific interquartile of concentration. The curve in black is the treatment-specific median, and the curve in red is the mean concentration.

Select a specific individual. We'll be able to see their pharmacokinetics in the context of their treatment arm as well as the overall patient population. This provides us a useful reference for an initial intuitive assessment of individual's curve, whether it shows a meaningful pattern or just a random fluctuation. It also serves as a quick sanity check of the data. We can also break down the concentration data by demographic groups, such as sex, race, or click this plus sign to include any other categorical variables of interest.

Doing this helps us come up with hypothesis, catch outliers, see how much variation there is within groups, and spotify any unbalanced sampling. Here we show subgroup-specific statistics similar to the last figure. For example, if we know that genetic differences in drug metabolizing enzymes vary by race. Race can have a big impact on how drug is absorbed, distributed, metabolized, and cleared from the body. This gives us a quick and intuitive way to check how concentration vary across different subgroups.

With this understanding and data quality check, let's move on to the second part, the non-compartmental analysis. Let me briefly explain how the NCA was done. We first create PK profiles based on the subject ID, group ID, PC test, and epochs. Then for each profile, we estimate the PK parameters. Finally, we visualize and summarize the parameters into a table.

This first figure shows us the distribution of the estimated parameters by the key covariates. This figure can serve as a screening tool to explore potential exposure or subgroup differences. As expected, we can see that higher dose gives us higher AUC values on average. We can also look at this information across sex, race, or other variables of interest. It gives us the ideas like whether clearance is slower in Asians compared to Whites. Users may run t-test or ANOVA to formally compare the subgroups.

To help us better understand how physiology, namely the biological processes that influence how the body handles a drug, we look at the estimated PK parameters against each other. This helps us explore relationships between PK parameters and ask questions like, does the relationship hold consistently across a subgroup? We can identify outliers or subgroups and ask questions like, are there subgroups specific trends or shifts? Are there outliers in certain subgroups? Do you observe covariate PK parameter interactions?

With that understanding, we also want to make sure that we're fitting the appropriate models. We visually assess the model fit for each individual by presenting this figure with the observed concentration curve in blue overlaid with the fitted curve over the elimination phase in red. Select a specific profile. We can check the fit at an individual level and spotify unusual individual PK patterns.

Finally, we summarize the estimated parameters in the table. Users can select specific subjects to look at their parameters. If you go to this Show Tables icon, you'll be able to see three underlying tables. The first table gives you the demographic information along with the estimated PK parameters. Users can make use of this table to explore more. Simply click this, Explore All Rows in JMP, this middle icon, and the table will pop up. Users can use all the advanced platforms to do more analysis.

The second table is grouping variables. This is showing us the details for how we created the PK profiles. We also provide this initial version of the PP domain if the users need to compare to their own a PP domain or to create a new PP domain. From here, I will hand it back to Chris for some final insights.

Thank you, Meichen. It was great to see all of that in action and show us a better feel for what the PK report actually behaves in JMP Clinical. What did we see? First of all, we saw JMP Clinical, specifically JMP Clinical 19, simplifies non-compartmental analysis for the clinical trial reviewer. A simple report on the data generates the tables and figures, allows us to explore the data as well as confirm some of the parameters that need to be calculated.

It also allows us for early stage exploratory analysis in a clinical trial. Many of these particular studies are done in the early phases, and be able to monitor these things along with their safety data makes it easy for us to communicate with other departments when it becomes time for moving on to the next clinical stages of the trial.

Also, we have seen that there's very interactive visualizations, and it can really help us in the interpretation or find outliers as Meichen has just showed us. This can also be very ideal for clinical teams who may have limited statistical background and also may have some limited experience in PK analysis.

There are some limitations, though, things that will need to be worked on in the future. These types of analysis do assume that we have complete PK curves per patient. Right now, we don't currently support sparse sampling, but maybe we will in the future. We really didn't design this to be as a replacement for the current methods that are used for the NDA submission, but really a complement and to use along with that safety review as well, because there might be something that goes on that a person's PK profile may be an indicator that something else is going on where a particular profile may match a particular safety, and we may want to drill down further into those patients in that safety information.

Some of our future considerations are to incorporate a domain that's been around for a while that's specifically for NCA analysis, and that's the ADNCA and the ADPPK, and try to take better use of those domains. Who knows? We might put other reports or other different types of reports and outputs into this report to make it more convenient and more informative to people who might submit.

Thank you for spending time with us today. We appreciate the time and the thought into taking a look at where we have. I hope you have a good day.