A mathematical model developed by researchers at The University of Texas at Arlington may help resolve one of melanoma treatment’s most persistent contradictions: why continuous targeted therapy has outperformed planned treatment breaks in clinical trials, even though laboratory experiments have suggested that intermittent dosing could slow the evolution of drug resistance.
The study, published in Mathematical Biosciences, examines how melanoma tumors respond when therapy is adjusted over time rather than delivered according to a rigid, predetermined schedule. Its central conclusion is striking: the most effective strategy may not be continuous full-dose treatment or repeated cycles of treatment and withdrawal, but a carefully controlled taper that begins aggressively, moves to an intermediate dose and ultimately stops without restarting.
The research comes as melanoma remains a major public health concern. The American Cancer Society estimates that approximately 112,000 new melanoma cases will be diagnosed in the United States in 2026, with about 8,510 deaths from the disease. Although newer targeted therapies have transformed outcomes for many patients, tumors can evolve under treatment pressure. Resistant cells that survive therapy may eventually expand, allowing the cancer to return or progress.
The apparent conflict between laboratory and clinical results has centered on treatment timing. In laboratory systems, intermittent dosing can sometimes preserve drug-sensitive tumor cells, which compete with or suppress resistant cells. When the drug is removed, sensitive cells may regain a growth advantage, potentially preventing resistant populations from taking over. Yet clinical trials using simple on-and-off treatment schedules have generally failed to produce the expected benefit, and continuous therapy has often remained more effective at controlling melanoma.
Souvik Roy, an associate professor of mathematics at UTA, and collaborators Natalia Komarova of the University of California–Irvine and her student Anthony Zamora used mathematical modeling and optimal control theory to investigate why. Their model represents the tumor as a dynamic population containing cells with different responses to therapy. It then calculates how drug exposure changes the size and composition of the tumor over time, while also accounting for the harmful effects of treatment on the patient.
The model is based on a bilinear control framework, a mathematical structure in which the treatment level interacts with the biological state of the tumor. In practical terms, the impact of a drug is not treated as a simple constant reduction in tumor size. Instead, its effect depends on how many sensitive and resistant cells are present and how those populations influence one another. The researchers then use optimal control techniques to search for a treatment schedule that minimizes a combined measure of tumor burden and toxicity.
This approach produces a result that differs from the familiar “treat, stop, treat again” pattern. The model favors beginning with the highest treatment intensity to eliminate drug-sensitive cells quickly. It then recommends reducing the dose to an intermediate level before eventually discontinuing therapy. Crucially, the predicted schedule does not call for restarting treatment after the final reduction. The sequence is designed to limit the time available for resistant cells to emerge while avoiding unnecessary exposure once the balance of tumor populations has changed.
The distinction between tapering and intermittent therapy is biologically important. Repeated treatment holidays can allow surviving tumor cells to recover and create multiple opportunities for resistant populations to expand when therapy is reintroduced. A gradual reduction, by contrast, changes the selective pressure in a more controlled way. The model suggests that the timing and direction of dose changes may matter as much as the total amount of drug administered.
The analysis also places toxicity at the center of treatment design. Standard oncology protocols frequently rely on fixed schedules, whether they involve chemotherapy, immunotherapy, surgery or targeted drugs. Such schedules are practical and familiar, but they may not reflect the changing biology of a tumor during treatment. A dynamically adjusted protocol could potentially reduce adverse effects, lower the financial burden associated with managing toxicity and reduce unnecessary healthcare costs, while maintaining stronger tumor control.
Roy emphasized that the work is a computational framework rather than a ready-to-use prescription for patients. Mathematical models can test thousands of possible treatment schedules before any strategy is evaluated in a clinical trial, but their predictions must still be validated using laboratory experiments, patient data and carefully designed prospective studies. Tumors differ between individuals, and real treatment decisions must account for factors that may not be fully captured by a model, including immune responses, drug pharmacology and the location of metastatic disease.
The melanoma study extends Roy’s earlier work applying mathematical methods to esophageal and colon cancers. Its broader significance lies in the possibility of using adaptive therapy to manage cancer as a changing ecological system rather than treating it as a static target. Researchers and clinicians are already exploring adaptive treatment concepts in clinical settings, and the new model could potentially be modified for other diseases in which efficacy and toxicity must be balanced over time. If future studies confirm its predictions, a treatment plan that gradually changes instead of following a fixed calendar could become an important tool in delaying resistance and improving the long-term control of melanoma.
Subject of Research: Mathematical modeling of adaptive treatment strategies for melanoma
Article Title: Optimal melanoma treatment protocols: A bilinear control model
Web References: https://doi.org/10.1016/j.mbs.2026.109764
References: Mathematical Biosciences, DOI: 10.1016/j.mbs.2026.109764; American Cancer Society melanoma estimates for 2026
Keywords: Melanoma, cancer treatment, targeted therapy, drug resistance, adaptive therapy, optimal control, mathematical modeling, tumor evolution, treatment toxicity, bilinear control model
