Work

Chance, Change, or Shock? A Stochastic Landscape Theory of Escape from Cancer Dormancy and Its Signatures in Relapse-Time Data

Year: 2026
Type: preprint
Abstract: Many cancers recur years after apparently curative treatment, because residual tumour cells persist in a dormant, immune-controlled state before resuming growth. Why and when dormancy ends remains a fundamental open question. We formulate this question mathematically and show that it can be answered from data.We model the residual tumour burden as a one-dimensional Itô diffusion with logistic proliferation, saturating (Hill-type) immune killing and environmental multiplicative noise. In dimensionless form the model has two parameters and a closed-form bistability region bounded by a cusp, in which a dormant state and a progressive state coexist. We prove that the process is well posed and positive, give the stationary density in closed form, derive the exact mean relapse time and its Eyring–Kramers asymptotics, and validate all of them against Fokker–Planck and Monte Carlo computations.We then distinguish three mechanisms of dormancy escape: chance (noise-driven crossing of a fixed barrier), change (slow erosion of the barrier, e.g. immune decline) and shock (a transient perturbation, e.g. surgery-induced inflammation). We show that these mechanisms leave distinct, testable signatures. Chance yields an asymptotically constant hazard of relapse. Change yields a strictly increasing hazard, a coefficient of variation below one and rising variance and autocorrelation in serial biomarkers. Shock yields a transient hazard spike followed by a return to baseline.Because any mixture of constant hazards has a non-increasing hazard, an increasing cohort-level relapse hazard falsifies pure chance-driven escape, whatever the heterogeneity between patients. These results turn the question “chance or change?” into hypotheses that existing relapse data and serial biomarker data can test.
Language: English
Open Access status: green
OpenAlex ID: W7234087177
Cites:
Cited by:
Related to: 10