Making a Human Impact with System Dynamics: A Healthcare Case Study
Vincenza Nigro is executive director of medical affairs at Veloxis Pharmaceuticals A/S in Edison, NJ. Rob Nachtrieb is a senior lecturer of System Dynamics at MIT and lead scientist of Lutron Electronics in Allentown, PA. Both Nigro and Nachtrieb are members of the MIT EMBA class of 2012.
When you are evaluating a new drug, there are several key things you need to understand. For example, how is the drug metabolized in the human body? How quickly does it work? And how fast is it eliminated from the body? System dynamics can describe the behavior of almost any system, but it’s particularly good at describing a complex physical system like the human body. When applied to these types of pharmacokinetic questions, system dynamics is a novel approach and can provide incredibly important and useful insights.
We’ve been using system dynamics models to study a new drug product designed to help kidney transplant patients. It’s an immunosuppressant that works to prevent rejection of the new organ and, as a result, can extend patients’ lives. The actual drug is not novel, but it’s new because it uses a different delivery technology than the drug currently on the market. As a result, we need to understand differences between the two drugs, especially in relation to those questions above. Although it may not be possible to explain or understand everything, system dynamics allows us to model the differences in absorption, exposure and metabolism. Most importantly, we can simulate the delay in these parameters when comparing a competitor’s drug.
Our system dynamics analysis of the new drug has been so significant that we presented our results in September at the European Society of Transplantation Meeting in Vienna, Austria. Here’s how we built our models:
Start out Simple
To build a system dynamics model to understand the pharmacokinetic parameters of the new drug, we started out as simply as possible. We looked at the existing clinical data to get an idea of the dynamics we need to understand. We then built what is known as a stock-and-flow model to explain that data in the most basic way possible. The actual model is proprietary, but here is a simple example with a single stock.
The simplest simulation reproduces the concentration of a drug that is absorbed in the blood as soon as it is ingested. (Sugar is quite similar.) The figure shows the concentration as a function of time that would result from a dose taken every 24 hours. The concentration is highest just after the dose is taken, and then the concentration decreases as the body eliminates the drug by metabolism.
This type of model is helpful, but the formulation we are interested in is not absorbed immediately. Instead, it is released into the bloodstream over time. Building on the example of the single-stock model, we add another stock to the model to represent the unreleased drug that has been ingested. You can see this step below with the addition of a second stock:
Again, the patient takes a daily dose of the drug, but now the peak concentration occurs approximately six hours afterwards, as the drug is released into the blood stream. (See red curve in Figure 2.) As before, after the drug has been released into the blood stream the body starts to eliminate it and the drug concentration decreases.
In the model we actually used — which was not that much more complicated — we adjusted the number of stocks and the parameters representing the rate coefficients to improve the fit between the model output and our clinical trial data. Then we ran stress tests of the model and iterated until the model matched the data.
Beyond the Data
Once the model matched the data, we could really put the model to work by exploring “what if” scenarios not considered in the original data. We could explore things like a missed dose or taking a dose outside of the window. Through the model, we can see how long it takes for the drug levels to drop below the minimum therapeutic level in the body.
Adding Real Value
Through our work, we’ve been able to provide specific guidance to the company. By taking existing data and modeling through system dynamics, we can manipulate the data to predict dosing, visually compare absorption and metabolic characteristics of the drug, model a missed dose, or model the behavior of the drug in specific subgroups, such as elderly, children, and patients with specific genotyping. Now that we have the model, we can further add complexity as new data is gathered and mined.
While these are important benefits, the biggest one is for the patients who will eventually take the new drug. System dynamics can help teach physicians how to more effectively use the drug and tailor that use based on patient characteristics. By using these simple models to show the interaction of feedbacks in a system, we can make a real human impact and extend lives.
Have you applied system dynamics to processes in your work? What kind of benefits have you seen?