Threshold-awareness in adaptive cancer therapy.

Area of Science:

• Oncology
• Mathematical Biology
• Computational Science

Background:

• Adaptive cancer therapy integrates evolutionary dynamics but often ignores stochasticity.
• Random perturbations in heterogeneous tumors affect treatment policy performance.
• Existing models rarely account for the random nature of cancer evolution.

Purpose of the Study:

• To develop an efficient method for selecting optimal adaptive cancer treatment policies under stochastic tumor dynamics.
• To maximize the probability of achieving treatment goals (stabilization or eradication) within a defined cost budget.
• To address the random nature of treatment cost, including drug amount and treatment duration.

Main Methods:

• Utilized a novel Stochastic Optimal Control formulation.
• Applied Dynamic Programming to derive "threshold-aware" optimal treatment policies.
• Developed an efficient algorithm to compute policies for various budget thresholds.

Main Results:

• "Threshold-aware" policies significantly improve therapy success probability under budget constraints.
• Optimized policies correlate with lower overall drug usage compared to deterministic approaches.
• The method demonstrated effectiveness in illustrative examples with stochastic cancer models.

Conclusions:

• The proposed method offers a robust framework for optimizing adaptive cancer therapies considering evolutionary stochasticity.
• This approach provides a valuable tool for balancing treatment efficacy with cost constraints in cancer care.
• It enhances the reliability of adaptive treatment strategies in the face of unpredictable tumor evolution.