SUMMARY
In this note we discuss combing therapeutics, diagnostics and quantitative system pharmacology modelling to provide a patient personalized precision dosing strategy, for achieving therapeutic efficacy whilst reducing toxicology.
BACKGROUND
Modelling in the biopharma industry has been a part of the toolset for over 30 years, with Silicon Graphics machines embedded in Medicinal Chemistry Departments by the mid-nineties. Back then, modelling was narrow and specific, focused on small molecule and binding site interactions.
Nowadays, modelling is also used at a systems level, such as in Quantitative Systems Pharmacology (QSP), which models the dynamic interaction between drugs and biological systems. There is also the related subject of Quantitative Systems Toxicology (QST), which aims to reduce the toxicological impact of drug therapy rather than focusing solely on broader pharmacological activity. The aims of QSP and QST are mutually compatible, as both aim to optimize drug efficacy while minimizing harmful effects.
CLINICAL APPLICATION OF QSP
Although QSP/QST are primarily applied in pre-clinical drug discovery efforts, in this note we discuss how these models can be used in clinical studies and post-market launch as a means of connecting diagnostics and therapeutics. The strategy would involve implementing a more personalized precision dosing strategy for a therapeutic, based on either monitoring biomarkers or tracking the drug concentration in the patient’s body. We envision the digital health solution as a combination of the therapeutic, a diagnostic tool, and an adaptive pharmacological model that can guide clinicians and/or patients. This is illustrated in Figure One.
In our connected world, where smartphones, the Internet of Things, Edge Computing, and Cloud Computing prevail, the QSP model between the therapeutic and the diagnostic doesn’t have to be static. Instead, the model could be adaptive and continuously learning over time. The analogy here is that the QSP model adapts to create a more accurate digital twin of the patient. While the core model would be population-based, the coefficients can be adjusted as the model learns the connection between precision dosing and the patient’s diagnostic results (see Figure Two). The author acknowledges that a PID controller could serve as the connection between the therapeutic and the diagnostic. However, considering that PID controllers are over one hundred years old, the modern efforts in QSP should definitely add value, especially in a post-machine learning world.
PRECEDENCE
The linking of diagnostics with therapeutics occurs millions of times a day. For instance, individuals with Type 1 diabetes control their insulin injections based on the biomarker blood glucose measured at home using a blood glucose meter. Three examples of diagnostic and therapeutic combinations includes:
- Dosing of Coumadin based on at-home PT-INR measurements.
- Partial control of insulin pumps by continuous glucose meters (CGM). The concept of the Artificial Pancreas, where Medtronic has an FDA-approved CGM and Insulin Pump combination that stops the insulin pump if the CGM detects low glucose values.
- Calculation of Insulin Boluses based on Self-Monitored Blood Glucose. Individuals with Type 1 diabetes test their capillary blood using an at-home electrochemical blood glucose meter for glucose concentration, and use a ‘simple’ calculation (Equation One) to determine the required insulin dosage. What is interesting is that the Correction Factor (sensitivity) is partly determined by the patient, who uses personal experience and learning to adapt the calculation for themselves. It could be argues that the model is adaptive.
In this note, one could argue that adaptive models to control therapeutics based on diagnostics already exist and are accepted, but these models are run by the patient. Therefore, they are highly patient-sensitive and could be seen as exclusionary, favouring patients with a STEM inclination/background/education. In this note, we simply propose that an evolution of QSP and QST should be directed towards the patient, helping to guide dosing in a dynamic and personalized manner.
DIAGNOSTICS
The thesis in this note is that QSP/QST can expand into other phases and not only be a modelling tool in the drug development phase. Furthermore, QSP/QST can also be the key component in a revolution of precision dosing. However, for this to happen, rapid tests/diagnostics have to exist, preferably electronic test that do not require the user to interpret the signal or line. At ZP, we believe that we have a technology stack that can serve as the diagnostic part of the diagnostic-QSP-therapeutic system for patient centric precision dosing.
