333749 An Automated System for the Design of Optimal Personalized Chemotherapy Protocols for the Treatment of Leukemia
An automated system for the design of optimal personalized chemotherapy protocols for the treatment of Leukemia.
E. Pefania, N. Panoskaltsisb, E. Vellioua, M. Fuentesa, A. Mantalarisa, M.C. Georgiadisa, E.N. Pistikopoulosa
aCentre for Process Systems Engineering, Department of Chemical Engineering, Imperial College London, South Kensington Campus, London SW7 2AZ, UK
b Department of Hematology, Imperial College London, Northwick Park & St. Mark's Campus, London, HA1 3UJ, UK
Keywords: Mathematical modeling, chemotherapy optimization, cell cycle models, pharmacokinetics, pharmacodynamics
In the UK (Cancer Research UK data, 2008), it is estimated that more than 1 in 3 people will be afflicted with cancer in their lifetime. For one such cancer, leukemia - a neoplasm of the blood and bone marrow (BM) - 1 in 71 men and 1 in 105 women will be affected, with incidence sharply rising in adults over the age of 50. Approximately 40% of those affected with leukemia will have Acute Myeloid Leukemia (AML), a cancer of the BM and blood wherein blood cells are unable to develop or function normally, are overproduced at an immature stage of development and overtake any normal elements remaining in the BM and blood. This uncontrolled growth compounds the morbidity and mortality due to the disease by inhibiting development of healthy blood and immune cells through multiple mechanisms (Panoskaltsis 2003; Panoskaltsis, 2005).
The most common treatment for most types of leukemia is intensive chemotherapy. It is well-known that this therapy can be life-threatening with only relatively few patient-specific and leukaemia-specific factors are considered in current protocols; choice of chemotherapy, intensity and duration often depends on either the availability of a clinical trial, the treating physician's experience or the collective experience of the treating centre, with significant international protocol variability.
There is therefore a need to optimize current treatment schedules for cancers such as AML to limit toxicities and to improve clinical trial pathways for new drugs to enable personalized healthcare. Towards this end, mathematical modeling is undoubtedly a useful tool for the automation of treatment design due its advantages in exploring extensive datasets in combination with scientific principles in order to capture a systems dynamics and subsequently provide better insight for process enhancement (Dua et. al, 2008; Pefani et. al, 2013).
In this work, the prospective advances to the clinical treatment design through a mathematical model are discussed. A mathematical model and model analysis is presented for induction chemotherapy treatment for AML using Daunorubicin (DNR) and Cytarabine (Ara-C) anti-leukemic agents, a standard intensive treatment protocol for AML. The proposed model combines critical targets of drug actions on the cell cycle, together with pharmacokinetic (PK) and pharmacodynamic (PD) aspects providing a complete description of drug diffusion and action after administration. The current developed model has been created in the gPROMs environment (PSE, 2003) and consists of a model simulator and an optimizer comprised of a closed-loop system for the design of optimal personalized chemotherapy protocols (Figure 1).
Figure SEQ Figure \* ARABIC 1: Schematic representation of the proposed closed loop system for the design of optimal personalized chemotherapy protocols.
The required inputs for both the simulator and the optimizer consist of patient, disease and drug information. Patient information includes gender, age, weight and height, whereas, disease information is that acquired as standard practice from the BM diagnostic test – leukemic blast percentage in the BM aspirate and the BM cellularity from the trephine biopsy. Extra information to that which is currently used in clinical practice will be the cell cycle characteristics of the S-phase duration and the total cell cycle duration. Moreover, PK information of drug elimination rate in liver and kidneys is required and is readily available in the product specification provided by the pharmaceutical manufacturer. PD information is also required that consists of drug dependent parameters characterising the effect of the anticancer agent on the cell population.
Results of sensitivity analysis indicate that the cell cycle times are the critical model parameters affecting treatment outcome. Clinical data of 6 patients are used in order to estimate the cell cycle time distribution and characterise the intra- and inter- patient variability on cycling times. Moreover, the derived mathematical model is used as an application for an optimization problem. The benefits of optimization are presented and an optimization problem is developed for chemotherapy treatment schedule. The objective of chemotherapy treatment in AML is the reduction of cancer cells to eventual eradication and to ensure that more normal cells remain in the BM. Moreover, the number of normal cells is constrained to have a maximum total reduction of 3-log during one chemotherapy cycle. An optimization algorithm is presented in the current work for the scheduling of the optimal treatment design based on the control drug use, dose load and dose duration. This optimization problem is solved for one of the patient case studies (P016). P016 received two chemotherapy cycles of treatment and by treatment completion residual disease was detected. For this reason optimization is applied and the optimal treatment schedule for this specific case study is derived. At completion of the optimized treatment protocol, the leukemic cell population is further minimized with a difference of 3·109 cells when compared to that of the standard treatment protocol. Furthermore, at treatment completion the normal cell population is also lower than that prior to treatment, but it is still higher than the leukemic cell count with a difference of 5.3·109 cells, as is the desired treatment outcome.
Model analysis reveals the utility of mathematical modeling in gaining better insight into the disease and into normal tissue dynamics during treatment with chemotherapy. There is future potential for the amelioration of treatment design that will be defined on a case-by-case basis and would be dependent on disease characteristics (tumor-specific characteristics, tumor burden, cell cycle times, normal cell population) as well as patient-specific characteristics (gender, age, weight and height). This design would provide the opportunity to personalize treatment protocols through the use of optimization methods.
Acknowledgment
This work is supported by European Research Council (MOBILE, ERC Advanced Grant, No: 226462), the Richard Thomas Leukaemia Fund and CPSE Industrial Consortium.
References
Dua, P., Dua, V., Pistikopoulos, E., 2008. Optimal delivery of chemotherapeutic agents in cancer. Computers & Chemical Engineering, 32(1-2), 99-107.
gPROMS, Introductory user's guide, release 2.2, Process Systems Enterprise Limited, London, U.K; 2003.
Panoskaltsis N, Reid CDL, Knight SC. Quantification and cytokine production of circulating lymphoid and myeloid cells in acute myelogenous leukemia (AML). Leukemia 2003; 17:716-730.
Panoskaltsis N. Dendritic cells in MDS and AML – cause, effect or solution to the immune pathogenesis of disease? Leukemia 2005; 19:354-357.
Pefani E., Panoskaltsis N., Mantalaris A., Georgiadis M. C., Pistikopoulos E. N.. Design of optimal patient-specific chemotherapy protocols for the treatment of Acute Myeloid Leukaemia (AML). Computers and Chemical Engineering Journal, In Press, Available Online 1 March 2013.
See more of this Group/Topical: Food, Pharmaceutical & Bioengineering Division