Workshop 3: Control of Disease: Personalized Medicine Across Heterogeneous Populations
The state of the cardiovascular system can be assessed from time-series signals including heart rate and blood pressure. Characteristics of these signals are used to determine pathophysiology. Experienced clinicians can visually inspect signals and with high level of certainty determine key observed dynamics however analysis with computational models can uncover the underlying mechanisms driving these dynamics. This talk will address how mathematical modeling can be adapted to match patient specific behavior and how the optimized system equations can be used to predict emergent behavior. Sensitivity analysis, parameter estimation, and uncertainty quantification are mathematical tools used analyze models for individual patients. Focus will be on studying dynamics observed in patients diagnosed with postural orthostatic tachycardia (increased heart rate brought on by change in posture, e.g. sit to stand) observed in some girls after vaccination against the human papilloma virus (HPV). Girls exhibiting side-effects to this vaccine often feel dizzy, light-headed, and tired. It is believed that these symptoms are related to pathophysiology within the cardiovascular control system, which is supposed to keep heart rate and blood pressure constant. To understand how the system is impacted we use models to analyze patient specific changes observed during head-up tilt, Valsalva (breath holding), and deep breathing.
Aurelio de los Reyes V
A model of the human cardiovascular-respiratory system (CVRS) is developed to describe its response to various ergometric workloads. An optimal control for time-varying workloads is obtained by using the Euler-Lagrange formulation of the optimal control problem. Variations in the heart rate and the alveolar ventilation are considered essential controls of the CVRS in which arterial pressure of CO2 is regulated close to 40 mmHg. Penalization terms are also included in the cost functional to match the metabolic need for O2 and metabolic production of CO2 with O2- and CO2- transport by blood. In this work, sensitivity analysis on the parameters of CVRS model under a constant ergonometric workload is performed to identify which are most/least influential to the arterial systemic pressure, for which experimental data are available. Three different methods are considered - traditional sensitivity analysis, partial rank correlation (PRC) and extended Fourier amplitude sensitivity test (eFAST) analysis. For each of the three methods, a ranking of parameters is obtained according to sensitivity and a set of five parameters across the three methods is identified. Furthermore, parameter estimates are obtained on different datasets and indicated good fitting results.
The tailoring of medical treatment to the individual characteristics of each patient (definition of "Personalized Medicine"â€™ by the President's Council of Advisors on Science and Technology) among many other requirements puts new challenges for mathematical modeling of physiological subsystem. In the presentation we shall first discuss some of the definitions of personalized medicine or precision medicine available in the literature, then the challenges on mathematical modeling and finally illustrate these for some concrete examples.
In this presentation I will provide an overview of some robust and adaptive approaches to the problem of controlling patient response to sedative anesthetic agents in clinical settings. In current practice, anesthesiologists are responsible for monitoring and adjusting the delivery of anesthetic agents to the patient, with the main goal being to maintain a desired level of sedation, analgesia and muscle relaxation. At the same time, the attending anesthesiologist must ensure proper cardiovascular and respiratory functioning of the patient, for example maintaining appropriate heart rate (HR), blood pressure (BP), oxygen saturation and end-tidal (exhaled) carbon dioxide levels, amongst other patient indicators. That is, the anesthesiologist performs the role of a multivariable feedback controller for a highly complex process. Our goal is to incorporate partially automated anesthesia delivery into the process, allowing the anesthesiologist to concentrate on urgent safety-critical events that arise during surgery. Two necessary ingredients towards automating anesthesia delivery are (1) adequate means of sensing the patient's level of sedation, analgesia and muscle relaxation, and (2) mathematical models capturing the patient response to anesthetic agents. In this presentation, we focus on efforts aimed at modeling and controlling the level of sedation via automated feedback methods, while at the same time maintaining the patient's blood pressure in a given safe range. We discuss the development of a control design approach in which we use system identification to construct multivariable patient models and apply advanced control methods, so that patientsâ€™ sensed sedation levels track a desired reference trajectory.
The first published work on closed-loop control of anesthesia authored by Mayo, Bickford and Faulconer dates back to 1950. However, feedback control of anesthesia has yet to become adopted for clinical use despite the fact that since 1950, and especially since the availability of depth-of-hypnosis monitors many systems have been developed and clinically tested. For clinical adoption to take place, benefits to the patient have to be demonstrated while patient safety has to be guaranteed. This presentation will focus on the engineering design process that can ensure patient safety by design.
A major characteristic of the anesthesia closed-loop control problem is the significant inter and intra-patient variability and that under such circumstances, the control system has to perform adequately and safely for a patient it has never seen and with very limited learning opportunity. This is why it is crucial for such systems to be designed by expert control engineers well versed in robust control theory. I will briefly describe the theory of robust control and demonstrate its use for closed-loop anesthesia in order to guarantee closed-loop stability and a minimum level of performance acceptable to the clinician.
Furthermore, the system has be able to handle safely a number of exceptions with well-designed fallback modes. This is where the field of safety-preserving control, a concept developed for aeronautic applications, can provide a number of attractive solutions. This will be demonstrated through work performed at the University of British Columbia when developing our closed-loop TIVA control system, iControl.
Finally, because the anesthesiologist will remain in the loop and has to be able to take over in a safe manner should the need arise, the Human-Computer Interaction (HCI) aspects have to carefully studied and designed. I will cover some of our nascent work on this topic, also inspired by aeronautics applications.
An Optimal Control Approach to Structured Treatment Interruptions for HIV Patients: A Personalized Medicine PerspectiveHien Tran
Highly Active Antiretroviral Therapy (HAART) has changed the course of human immunodeficiency virus (HIV) treatments since its introduction. However, for many patients, long term continuous HAART is expensive and can include problems with drug toxicity and side effects, as well as increased drug resistance. Because of these reasons, some HIV infected patients will voluntarily terminate HAART. Some of these patients will also interrupt the continuous prescribed therapies for short or long periods. After discontinuing HAART, patients will usually experience a rapid increase in viral load coupled with a immediate decline in CD4+ counts. The canonical example of a patient undergoing unsupervised breaks in HAART is that of the â€œBerlin patientâ€?. In this case, the patient was able to control viral load in the absence of treatment by cycling HAART on and off due to non-related infections. Due to this patient, interest in the use of structured treatment interruptions (STI) as a mechanism to regulate an HIV infection piqued. This talk discusses an optimal control approach to determine STI regimen for HIV patients. The optimal STI was implemented in the context of the receding horizon control (RHC) using a mathematical model for the in-vivo dynamics of an HIV type 1 infection. Using available clinical data, we calibrate the model by estimating on a patient specific basis a best estimable set of parameters using sensitivity analysis and subset selection. We demonstrate how customized STI protocols can be designed through the variation of control parameters on a patient specific basis.
Inflammation trigs and drives leukemia and the related Myeloproliferative Neoplasm (MPNs) diseases through the innate immune system while leukemia and MPNs stimulates the inflammatory responds of the adaptive immune system fighting the malign cells of the diseases. Where the two-way coupling of tumorous cancer and the adaptive immune system has drawn some attention during the last decades and inspired to immuno- and gene-therapy, leukemia and MPNs have been left unnoticed with respect to such coupling. Furthermore, the two-way coupling of leukemia and MPNs, and the innate immune system is by the large left unstudied with respect to treatments and preventive measures. We pose a novel mathematical model of development of leukemia and MPNs taking these two-way couplings into account. The model is validated against human data. It follows that the innate immune response is crucial in the development and treatment of leukemia and MPNs. Steady states and their stability are determine analytically and it is discuss how the model may be used for optimizing treatment. Geometric singular perturbation theory suggest a reduced model which shows excellent agreement with the full model.
B. Wayne Bequette
Pursuit of a closed-loop artificial pancreas that automatically controls the blood glucose of individuals with type 1 diabetes has intensified during the past decade. Here we discuss the recent progress and challenges in the major steps towards a closed-loop system. Continuous insulin infusion pumps have been widely available for over two decades, but â€˜â€˜smart pumpâ€™â€™ technology has made the devices easier to use and more powerful. Continuous glucose monitoring (CGM) technology has improved and the devices are more widely available. Physiological models that include subcutaneously delivered insulin pharmacokinetics and pharmacodynamics enable extensive simulated clinical trials using proposed closed-loop algorithms before testing in human clinical trials.
A number of approaches are currently under study for fully closed-loop systems; most manipulate only insulin, while others manipulate insulin and glucagon. Algorithms include onâ€“off (for prevention of overnight hypoglycemia), proportionalâ€“integralâ€“derivative (PID), model predictive control (MPC) and fuzzy logic based learning control. Two major disturbances include meals and exercise. While feedforward action using a â€œmeal announcementâ€? is desirable it places an addition burden on individuals, so we review techniques for meal detection/prediction; similarly, we review exercise detection methods. Finally, we examine fault detection algorithms for insulin infusion set failure and sensor signal attenuation.
Diabetes has become the epidemic of the 21st century affecting more than 300 million people globally. Treatment regimens and medication dosing for diabetes have relied on weight based formulas or non-customized dosing regimens for decades, and not surprisingly, treatment outcomes have been far from satisfactory. The individualization of diabetes pharmacotherapy to address the variability in response to medications and side effects has been introduced to improve diabetes treatment, however has been unattainable until the advances in technology made it a possibility.
The development of the closed-loop (CL) systems has become one of the most striking examples of customized insulin treatment for people with type 1 diabetes. The CL, commonly referred to as the Artificial Pancreas systems, consist of a continuous glucose monitor (CGM), insulin delivery algorithm and an insulin pump that work in tandem to automate insulin treatment based on real-time CGM values. The CL systemsâ€™ key innovative factor has been the continuous adjustment of insulin delivery based on individual patientâ€™s blood glucose response as opposed to standard of care treatment by quarterly insulin dose adjustments during clinic visits. The findings from diabetes technology trials have demonstrated the benefit of better fitting treatments that are achieved using increased computational power, technological insulin delivery tools and the advent of mobile and wireless capability. While these systems are not completely customized for each patient, they provided the platform to advance personalized diabetes management and generated more questions to overcome barriers in implementing individualized diabetes treatment.
The first-generation diabetes technology systems will evolve as we gain more experience in designing customized diabetes treatment regimens with skillful application of individualized insulin action profiles, continue to incorporate new technology into current systems and collaboratively work to improve diabetes management for an ultimate goal of keeping people with diabetes complication and burden-free until a diabetes cure becomes a reality.
Diabetes has serious impact on the kidney: it elevates solute and water transport along the nephron and induces glomerular hyperfiltration. A new class of diabetes drugs, namely the sodium-glucose co-transporter 2 (SGLT2) inhibitors, enhance glucose excretion and lower hyperglycemia in diabetes. These drugs work by targeting Na+ and glucose reabsorption along the proximal convoluted tubule (a proximal segment of the nephron). The goal of this study is to explore how SGLT2 inhibition affects solute transport and oxygen consumption along the rat nephron in chronic kidney disease, in the presence but also in the absence of diabetes.
Addressing the pressing global health challenges will require new and innovative approaches to healthcare delivery. The widespread availability of low cost but powerful digital technology has provided opportunities to collect detailed information from front-line health workers everywhere. Clinical decision making can be supported and enhanced with the timely and evidence based treatment and referral recommendations. The complexity of these decisions that have traditionally required extensive training and experience can now be reliably undertaken by healthcare workers in the community with minimal training or experience but supported by powerful data driven prediction models. The goal of these prediction models is to significantly change the trajectory of the typical disease process. Early identification of the patient at risk can change healthcare behavior, allow for timely administration of treatment before significant deterioration in health and allow for early referral when a higher level of care is required. The development of these personalized prediction models is a rapidly evolving science that will leverage the era of big data to make a big impact on health outcomes globally.
One challenge to understanding mechanisms of behavior change (MOBC) completely among individuals with alcohol use disorder (AUD) is that processes of change are theorized to be complex, dynamic (time varying), and at times non-linear, and they interact with each other to influence alcohol consumption. We used dynamical systems modeling to better understand MOBC within a cohort of problem drinkers undergoing treatment. We fit a mathematical model to ecological momentary assessment data from individual patients who successfully reduced their drinking by the end of the treatment. The model solutions agreed with the trend of the data reasonably well, suggesting the cohort patients have similar MOBC. This work demonstrates using a personalized approach to psychological research, which complements standard statistical approaches that are often applied at the population level. Our efforts collaborative efforts with our group (Kidist Bekele-Maxwell, R. A. Everett, Lyric Stephenson) and psychologists Alexis Kuerbis (Hunter College, CUNY) and Sijing Shao and Jon Morgenstern (Northwell Health).