Pharmacokinetic (PK) and electrocardiogram (ECG) measurements from clinical trials provide critical insight into drug behavior and cardiac safety. PK data describe how drug concentrations change over time, while ECG data, particularly heart rate-corrected QT intervals (QTc), indicate cardiac risk. Both data types are longitudinal and are routinely summarized at the population level in ways that can obscure individual patient experiences, such as a subject whose concentration trajectory deviates sharply from the population trend or whose QTc interval exceeds notable thresholds. Alternatively, patient-specific profiles of longitudinal results provide no insight into how the responses compare to the larger population.

This presentation discusses a standardized visualization framework that displays individual patient trajectories alongside population-level summary statistics within a unified graphical display. This framework is implemented as an interactive JMP add-in using the JMP Scripting Language (JSL), with support for annotations to communicate clinically relevent thresholds or notable observations (through reference lines or markers, respectively) to highlight potential safety concerns for individual patients. Additional functionality allows for the computation of basic PK parameters such as AUC, Cmax, and Tmax. The integrated visualization enables the rapid review of individual patients using animation through a local data filter, providing greater insight compared to patient- or population-only graphical displays. These findings demonstrate that standardized tools combining individual and population perspectives deliver a more complete and clinically actionable view of clinical trial data.

 

 

Hello, everyone. My name is Divya. I'm a recent graduate from UNC Chapel Hill, Gillings School of Global Public Health, and I'm here with a short presentation today on Exploring Patient Patterns Within Population Trends in Treatment Response. Overall, this has been a presentation or a project that I've been working on for the last year, and it's been a great honor to be able to work within JMP in regards to this.

The main problem or the main question we were trying to answer throughout this project was, how can we preserve individual-level information without losing the population context? Overall, it's pretty common to have the typical patient response be hidden within the overall population trends, and we just wanted to make a way to be able to see both contexts very efficiently and improved detection of clinically relevant patterns.

Two areas in which this has been shown for the profile plot has been within pharmacokinetic and electrocardiogram measurements from clinical trials, which will help provide clinical insight into drug behavior and cardiac safety. PK, or pharmacokinetic, data describes how the drug concentrations change over time, while electrocardiogram data, or ECG data, particularly, correlate heart rate-corrected QT intervals, which is a heart rate adjusted measurement, which is typically recorded in ECGs, and they indicate cardiac risk.

Both of these data types are longitudinal, and they can routinely be summarized at a population level in ways that can obscure individual patient experiences, such as a patient whose concentration trajectory deviates sharply from population trends, or for example, whose QTc interval exceeds notable thresholds. A key criterion, that would be the Torsades de Pointes. It's a life-threatening heart rhythm deviation within the lower chambers of the heart that create a distinct turning point in ECG or EKG monitors. That's something that would be really specific and important for safety measurements when it comes to clinical trials.

Alternatively, patient-specific profiles of longitudinal results will provide no insight into how the responses are actually compared to the larger populations, and this is the problem that we've tried to resolve through the creation of the profile plots. As you can see, there are multiple, both of the data sets, so there's both the pharmacokinetic profile results as well as electrocardiogram profile results. Here, you can see the typical workflow, which I can show you right here. The typical workflow or the dialogue, per se, would allow you to place the specific variables as needed to be able to create your plot.

By assigning the specific columns and the required columns, you'd be able to create your own profile plot, so long as it is longitudinal data sets. If you are doing, for example, AUC, Cmax, or Tmax, which are specific measurements for pharmacokinetic data sets that are typically common, which is the area under the curve, the concentration maximum, and the time maximum, if you were to select those options, you'd also be provided with those overall summary data sets.

We can take a look at the pharmacokinetic profile results. You can see right here that it has been separated by the treatment type. There is the Theophylline 300 mg and the Theophylline 400 mg. You can see that there is a drastic shift when it comes to the concentration value, and you can also see that there's a massive jump from all of the metrics down at the bottom when it comes to the area under the curve, the Cmax, and the Tmax. Over here, you can see that this is the threshold value, which was also selected in the previous dialogue. You can also see the flags, which is of the yes and no values in regards to if it's an outlier or not. You can see that there's the outlier flag right here, which would be put into that category. You also have a limit of detection that would serve as your threshold value.

When it comes to the QTc or the electrocardiogram plots, we can see two separate values here. We can see both a female version and a male version because they have different threshold values, and that can also be seen right here with the 460 for females, 460 milliseconds, and 440 milliseconds when it comes to males. You can see how the prolonged and non-prolonged are also as a result based off of their sex-related threshold values.

When it comes to Torsades de Pointes, there's also a flag value associated with that. If it is a Torsades de Pointes value, then it will be flagged right here, as you can see. If it is not a Torsades de Pointes, it will not be visible. I will go ahead and show you a live demo of this working and how you would be able to put it into action.

When you first enter your JMP Home, when you go into the Add-ins menu right here, if you have downloaded it from the marketplace, you should be able to see the Profile Plot Builder. Here, you can see the additional four values associated with it. You will be able to see a manual. You'll be able to see both the sample data sets that I've used as well when it comes to pharmacokinetic, the PK, and the electrocardiogram, which is the QTc sample data. You will also be able to pull up the dialogue, which will be the way that we do create these plots, and the manual.

If I open up the manual. This is a user manual. You should be able to see different key and clinical uses along with key features and allow you to use the add-in more appropriately as well. When it comes to actually using the add-in, first I'm going to go ahead and open up the PK sample data set, and then I'm also going to go ahead and open up the dialogue. As you can see, it's been pre-populated to show up in the available columns. I'm just going to go ahead and assign these columns. Now that I go ahead and assign these columns, I'm going to go ahead and create profile along with selecting these options so that I could get the PK metric values. Click 'Create Profile.' Here you go. You can see the overall PK metrics. This will be per patient. You can just see an overall summary of all of them. I'm going to close out of this for a second.

Now I can go ahead and click on this animation control, which will allow me to go through majority of the patients and take a look. As you can see, this would be the 300 mg Theophylline values based off of the concentration, and I can see how different patients are actually moving compared to the overall population. I can see each individual patient and how they are in response to the overall population metrics. You can take a look at that there. As mentioned, you'd be able to see those specific flag values as they show up as it correlates to the outlier values or points of interest.

Then I'll also go ahead and show you how it would work for the QTc data points and the different flag values there. I'm going to open up the QTc sample data set, and I'm also going to open up the dialogue. Here, I'm going to assign the patient ID to the subject. I'm going to use a grouping variable. I'm going to use the Torsades de Pointes flag. There we go. Once again, I'm going to go ahead and click on the animation control to be able to see the different points.

As you can see, even though these values are above the threshold value of 460, as this is a female, these are not highlighted because they are not a Torsades de Pointes flag. As you can see here, you can see that the Torsades de Pointes flag is stated as a no, so it will not change in color. When you go to the next patient, you will be able to see that this is a Torsades de Pointes flag, which is above the value of 500 milliseconds. That is now changing in value. When you click through the animation, you can see the different patients that have that point and which patients don't.

Overall, this presentation or this project has been trying to integrate individual patient trajectories and population-level summary statistics into a single reproducible display, as you've just seen. The flexible framework will allow for longitudinal clinical data sets without modifying its underlying code, and it would streamline the visualization and parameter deviations within one workflow.

Overall, it makes it very efficient, and the additional functionalities for the computation of basic PK parameters also enables the rapid view of individual patients using animation through local data filters, and can overall provide a greater insight compared to just patient-specific or population-only graphical displays. The overall findings demonstrate that there is a standardized tool to help combining individual and population perspectives to overall develop a more complete and clinically actionable view of a clinical trial data set. Thank you so much. Once again, I really appreciate the opportunity to speak.

0 Comments
Presented At Discovery Summit 2026

Presenter

Skill level

Intermediate
  • Beginner
  • Intermediate
  • Advanced
Published on ‎07-15-2026 03:40 PM by Community Manager Community Manager | Updated on ‎10-30-2026 03:35 PM

Pharmacokinetic (PK) and electrocardiogram (ECG) measurements from clinical trials provide critical insight into drug behavior and cardiac safety. PK data describe how drug concentrations change over time, while ECG data, particularly heart rate-corrected QT intervals (QTc), indicate cardiac risk. Both data types are longitudinal and are routinely summarized at the population level in ways that can obscure individual patient experiences, such as a subject whose concentration trajectory deviates sharply from the population trend or whose QTc interval exceeds notable thresholds. Alternatively, patient-specific profiles of longitudinal results provide no insight into how the responses compare to the larger population.

This presentation discusses a standardized visualization framework that displays individual patient trajectories alongside population-level summary statistics within a unified graphical display. This framework is implemented as an interactive JMP add-in using the JMP Scripting Language (JSL), with support for annotations to communicate clinically relevent thresholds or notable observations (through reference lines or markers, respectively) to highlight potential safety concerns for individual patients. Additional functionality allows for the computation of basic PK parameters such as AUC, Cmax, and Tmax. The integrated visualization enables the rapid review of individual patients using animation through a local data filter, providing greater insight compared to patient- or population-only graphical displays. These findings demonstrate that standardized tools combining individual and population perspectives deliver a more complete and clinically actionable view of clinical trial data.

 

 

Hello, everyone. My name is Divya. I'm a recent graduate from UNC Chapel Hill, Gillings School of Global Public Health, and I'm here with a short presentation today on Exploring Patient Patterns Within Population Trends in Treatment Response. Overall, this has been a presentation or a project that I've been working on for the last year, and it's been a great honor to be able to work within JMP in regards to this.

The main problem or the main question we were trying to answer throughout this project was, how can we preserve individual-level information without losing the population context? Overall, it's pretty common to have the typical patient response be hidden within the overall population trends, and we just wanted to make a way to be able to see both contexts very efficiently and improved detection of clinically relevant patterns.

Two areas in which this has been shown for the profile plot has been within pharmacokinetic and electrocardiogram measurements from clinical trials, which will help provide clinical insight into drug behavior and cardiac safety. PK, or pharmacokinetic, data describes how the drug concentrations change over time, while electrocardiogram data, or ECG data, particularly, correlate heart rate-corrected QT intervals, which is a heart rate adjusted measurement, which is typically recorded in ECGs, and they indicate cardiac risk.

Both of these data types are longitudinal, and they can routinely be summarized at a population level in ways that can obscure individual patient experiences, such as a patient whose concentration trajectory deviates sharply from population trends, or for example, whose QTc interval exceeds notable thresholds. A key criterion, that would be the Torsades de Pointes. It's a life-threatening heart rhythm deviation within the lower chambers of the heart that create a distinct turning point in ECG or EKG monitors. That's something that would be really specific and important for safety measurements when it comes to clinical trials.

Alternatively, patient-specific profiles of longitudinal results will provide no insight into how the responses are actually compared to the larger populations, and this is the problem that we've tried to resolve through the creation of the profile plots. As you can see, there are multiple, both of the data sets, so there's both the pharmacokinetic profile results as well as electrocardiogram profile results. Here, you can see the typical workflow, which I can show you right here. The typical workflow or the dialogue, per se, would allow you to place the specific variables as needed to be able to create your plot.

By assigning the specific columns and the required columns, you'd be able to create your own profile plot, so long as it is longitudinal data sets. If you are doing, for example, AUC, Cmax, or Tmax, which are specific measurements for pharmacokinetic data sets that are typically common, which is the area under the curve, the concentration maximum, and the time maximum, if you were to select those options, you'd also be provided with those overall summary data sets.

We can take a look at the pharmacokinetic profile results. You can see right here that it has been separated by the treatment type. There is the Theophylline 300 mg and the Theophylline 400 mg. You can see that there is a drastic shift when it comes to the concentration value, and you can also see that there's a massive jump from all of the metrics down at the bottom when it comes to the area under the curve, the Cmax, and the Tmax. Over here, you can see that this is the threshold value, which was also selected in the previous dialogue. You can also see the flags, which is of the yes and no values in regards to if it's an outlier or not. You can see that there's the outlier flag right here, which would be put into that category. You also have a limit of detection that would serve as your threshold value.

When it comes to the QTc or the electrocardiogram plots, we can see two separate values here. We can see both a female version and a male version because they have different threshold values, and that can also be seen right here with the 460 for females, 460 milliseconds, and 440 milliseconds when it comes to males. You can see how the prolonged and non-prolonged are also as a result based off of their sex-related threshold values.

When it comes to Torsades de Pointes, there's also a flag value associated with that. If it is a Torsades de Pointes value, then it will be flagged right here, as you can see. If it is not a Torsades de Pointes, it will not be visible. I will go ahead and show you a live demo of this working and how you would be able to put it into action.

When you first enter your JMP Home, when you go into the Add-ins menu right here, if you have downloaded it from the marketplace, you should be able to see the Profile Plot Builder. Here, you can see the additional four values associated with it. You will be able to see a manual. You'll be able to see both the sample data sets that I've used as well when it comes to pharmacokinetic, the PK, and the electrocardiogram, which is the QTc sample data. You will also be able to pull up the dialogue, which will be the way that we do create these plots, and the manual.

If I open up the manual. This is a user manual. You should be able to see different key and clinical uses along with key features and allow you to use the add-in more appropriately as well. When it comes to actually using the add-in, first I'm going to go ahead and open up the PK sample data set, and then I'm also going to go ahead and open up the dialogue. As you can see, it's been pre-populated to show up in the available columns. I'm just going to go ahead and assign these columns. Now that I go ahead and assign these columns, I'm going to go ahead and create profile along with selecting these options so that I could get the PK metric values. Click 'Create Profile.' Here you go. You can see the overall PK metrics. This will be per patient. You can just see an overall summary of all of them. I'm going to close out of this for a second.

Now I can go ahead and click on this animation control, which will allow me to go through majority of the patients and take a look. As you can see, this would be the 300 mg Theophylline values based off of the concentration, and I can see how different patients are actually moving compared to the overall population. I can see each individual patient and how they are in response to the overall population metrics. You can take a look at that there. As mentioned, you'd be able to see those specific flag values as they show up as it correlates to the outlier values or points of interest.

Then I'll also go ahead and show you how it would work for the QTc data points and the different flag values there. I'm going to open up the QTc sample data set, and I'm also going to open up the dialogue. Here, I'm going to assign the patient ID to the subject. I'm going to use a grouping variable. I'm going to use the Torsades de Pointes flag. There we go. Once again, I'm going to go ahead and click on the animation control to be able to see the different points.

As you can see, even though these values are above the threshold value of 460, as this is a female, these are not highlighted because they are not a Torsades de Pointes flag. As you can see here, you can see that the Torsades de Pointes flag is stated as a no, so it will not change in color. When you go to the next patient, you will be able to see that this is a Torsades de Pointes flag, which is above the value of 500 milliseconds. That is now changing in value. When you click through the animation, you can see the different patients that have that point and which patients don't.

Overall, this presentation or this project has been trying to integrate individual patient trajectories and population-level summary statistics into a single reproducible display, as you've just seen. The flexible framework will allow for longitudinal clinical data sets without modifying its underlying code, and it would streamline the visualization and parameter deviations within one workflow.

Overall, it makes it very efficient, and the additional functionalities for the computation of basic PK parameters also enables the rapid view of individual patients using animation through local data filters, and can overall provide a greater insight compared to just patient-specific or population-only graphical displays. The overall findings demonstrate that there is a standardized tool to help combining individual and population perspectives to overall develop a more complete and clinically actionable view of a clinical trial data set. Thank you so much. Once again, I really appreciate the opportunity to speak.



Start:
Wed, Oct 21, 2026 04:00 PM EDT
End:
Wed, Oct 21, 2026 04:45 PM EDT
Ped 8
0 Kudos