Toxicity-Centric Cancer Chemotherapy Treatment Design
| Autor: | Joseph T. Liparulo, Robert S. Parker, Timothy D. Knab |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
0209 industrial biotechnology
Mathematical optimization Optimization problem Computer science 020208 electrical & electronic engineering 02 engineering and technology Optimal control Scheduling (computing) Model predictive control 020901 industrial engineering & automation Docetaxel Control and Systems Engineering Non-linear least squares Toxicity 0202 electrical engineering electronic engineering information engineering medicine Relaxation (approximation) medicine.drug |
| Zdroj: | IFAC-PapersOnLine. 53:16353-16358 |
| ISSN: | 2405-8963 |
| DOI: | 10.1016/j.ifacol.2020.12.666 |
| Popis: | Cancer chemotherapy scheduling in the mathematics and engineering literature has generally focused on optimal control formulations and tumor kill, using constraints on dose magnitude and duration to implicitly mitigate toxicity. We introduce a framework for scheduling that focuses on clinically-relevant toxicity mitigation allowing clinicians to specify toxicity limits in terms they understand. Building from the model predictive control framework, we explicitly use the pharmacokinetic model of drug distribution as well as pharmacodynamic models of both antitumor effect and drug toxicity in the optimization problem. Clinical and logistical constraints round out the treatment design problem. Rather than direct inversion, we synthesize the optimization problem in an input-discretized form and solve via graphical processing unit (GPU) calculation. The resulting suboptimal solution is shown to be clinically indistinguishable from an optimal solution (calculated via nonlinear least squares (NLS) from a relaxation of the input and logistical constraints to continuous variables). Using a docetaxel administration case study, the algorithm controlled neutropenia within user-specified toxicity constraints while maintaining tumor eradication rates equivalent to, or better than, clinically-implemented dosing schedules. Changes in patient response - both antitumor efficacy and toxic drug sensitivity are captured via a nonlinear least squares (NLS) calculation at the end of each treatment cycle and updated in the next cycle design. By explicitly controlling treatment toxicity, this algorithm has the potential to improve patient quality-of-life. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|