MODEL-BASED DESIGN OF CANCER CHEMOTHERAPY TREATMENT SCHEDULES

摘要: Cancer is the name given to a class of diseases characterized by an imbalance in cell proliferation and apoptosis, or programmed death. Once cancer has reached detectable sizes ($10^{6}$ cells 1 mm$^3$), it assumed have spread throughout body, systemic form treatment needed. Chemotherapy commonly used, effects both healthy diseased tissue. This creates dichotomy for clinicians who need develop schedules which balance toxic side with efficacy. Nominally, optimal schedule --- where defined as amount frequency drug delivered one found be most efficacious from set evaluated during clinical trials. In this work, model based approach developing was developed. chemotherapy modeling typically segregated into pharmacokinetics (PK), describing distribution organism, pharmacodynamics (PD), delineates cellular proliferation, on organism. work considers two case studies: (i) preclinical study oral administration antitumor agent 9-nitrocamptothecin (9NC) severe combined immunodeficient (SCID) mice bearing subcutaneously implanted HT29 human colon xenografts; (ii) theoretical intravenous engineering literature.Metabolism 9NC yields active metabolite 9-aminocamptothecin (9AC). Both 9AC exist lactone inactive carboxylate forms. Four different PK structures are presented describe plasma disposition 9AC: three linear models at single dose level (0.67 mg/kg 9NC); nonlinear dosing range 0.44 -- 1.0 9NC. Untreated tumor growth modeled using approaches: exponential growth; switched transitioning between rates critical size. All PK/PD considered here bilinear kill terms decrease proportional effective concentration current The combining best accurately responses ten experimental administered 0.67 myschedule (Monday-Friday weeks repeated every four weeks). coupled PD captured response data multiple levels. Each problem formulated mixed--integer programming (MILP), guarantees globally solutions. When minimizing volume specified final time, MILP algorithm much possible end window (up cumulative toxicity constraint). While numerically optimal, that exponentially growing tumor, driven would experience same time regardless when long {it amount} administered. An alternate objective function selected minimize along trajectory. more clinically relevant better represents clinician (eliminate tissue rapidly possible). resulted eliminated burden rapidly, can recursively each cycle efficacy toxicity, per practice.The second consists intravenously first order elimination treating under Gompertzian growth. system also MILP, objectives above were considered. original authors solution qualitatively similar solutions originally control vector parameterization techniques. attempted administer interval. then posed receding horizon trajectory tracking problem. again, returned promising results; eliminated.