Multistage Clonal Expansion Model Examines Early-onset Cancer Incidence

Recent studies have reported an alarming rise in early-onset cancer diagnosis in adults who are under 50 years old. However, scientists do not yet know whether this troubling trend reflects true effects or apparent effects. True effects are environmental or genetic factors that can heighten cancer risk, trigger driver gene mutations, and increase cell proliferation; in contrast, apparent effects do not signify an actual causal outcome and may result from overdiagnosis or earlier detection due to improved technology and/or better screening policies. During the 2025 SIAM Conference on Computational Science and Engineering, which is currently taking place in Fort Worth, Texas, Navid Mohammad Mirzaei of Columbia University extended a multistage clonal expansion (MSCE) model to account for tumor size at diagnosis and better understand the incidence of early-onset cancers. “Is a true biological factor contributing to this?” he asked.

MSCE constitutes a class of continuous-time Markov chain models that capture carcinogenesis (i.e., the formation of cancer) from initiation through malignant conversion (see Figure 1); most cancer types only require three rate-limiting driver gene mutations for normal cells to become malignant. Unfortunately, traditional MSCE models have two major limitations: (i) they assume that cancer is detected at the first sign of malignancy, and (ii) they are not structurally identifiable.

To improve upon these shortcomings, Mirzaei derived an extended model from an existing kinetic formulation. This extended model incorporates tumor size at diagnosis and includes a time-dependent probability function based on the birth process of malignant cells, which resolves non-identifiability issues in the classic model and improves parameter estimation. “This model solves for the hazard of cancer,” he said. “If \(f(t)\) is a known input, then the model becomes structurally identifiable.”

Next, Mirzaei fit the model to three types of cancer: breast cancer in females (see Figure 2), colorectal cancer in males and females, and thyroid cancer in males in females. He selected these diseases because the National Cancer Institute’s Surveillance, Epidemiology, and End Results Program database noted recent increases in early-onset diagnosis (see Figure 3). After conducting parameter estimation and calculating the sojourn time (in this case, the time from the occurrence of the first malignant cell until detection), Mirzaei found that recent cohorts show accelerated carcinogenesis and heightened cancer aggression in early and mid-life patients; the traditional model is less sensitive to these changes.

Mirzaei then performed counterfactual analysis, which measures the effect of intervention by comparing patient outcomes in the presence and absence of screening. He found that the model projection showed a much higher incidence than the data points for past screening. “If the patients were not screened, it would have been disastrous,” he said.

Mirzaei concluded his presentation with a discussion about the distinct impacts of preventative screening for different cancers. For instance, colorectal screening—in the form of a colonoscopy—reduces cancer incidence because practitioners can remove precancerous polyps during the procedure. “But breast cancer screening is not as convenient,” Mirzaei said. Unlike colorectal screening, mammograms increase early detection; even if a patient is diagnosed during the test, they must undergo several rounds of treatment—excision, immunotherapy, radiation, and/or chemotherapy—to hopefully eliminate the cancer. In this case, a raise in detection is associated with a raise in incidence. These results collectively emphasize the importance of tumor size at the time of diagnosis, account for apparent effects, and indicate a true recent increase in early-onset cancer risk.