Kystis (Brown)

Purpose

Kystis is a discrete-event microsimulation model of bladder cancer in the U.S. population.¹ It simulates the natural history of bladder cancer, including symptom development, disease progression, clinical detection, treatment, and mortality. The model generates hypothetical individuals born in specific years. It can approximate the U.S. population as stacked birth cohorts but does not account for migration. It was developed to assess strategies for bladder cancer prevention, detection, and management.

Kystis models a series of events occurring either in parallel or sequentially using non-homogeneous Poisson point processes (NHPPP).² Each simulated person is assigned demographic attributes such as sex and race. The model simultaneously simulates mortality from other causes and exposure history from birth. The lesion instantiation process begins when a person enters the simulation (spawn moment) and ends with a terminal event (death from bladder cancer or other causes). An individual may develop zero, one, or multiple lesions, with the risk of lesion development influenced by demographic attributes and exposure history. Currently, the exposure generator is limited to smoking, a major environmental risk factor for bladder cancer, and is based on an extensively tested and verified reimplementation of the CISNET Lung Cancer Group Smoking History Generator (versions 5.2.1 and 6.0.0).³ Once a lesion develops, it grows and progresses through a series of states, influencing symptom development and, consequently, the clinical detection of bladder cancer.

Kystis can generate both precancerous and benign (papilloma) lesions. It models non-muscle-invasive bladder cancer (NMIBC) at the Ta and Tis stages, which can progress to T1. However, it does not differentiate between AJCC stages of muscle-invasive bladder cancer (T2, T3) and distant metastases (M1a, M1b), nor does it model lymph node involvement (N).⁴ Lesion growth follows a generalized logistic (Verhulst) growth model, with growth rate and capacity varying by lesion type. For Tis lesions, the carrying capacity is the number of cells that line half the bladder surface, whereas for all other tumors, it is the number of tumor cells occupying one-third of a distended bladder’s volume. The "steepness" of the growth curve is determined by the growth rate, which varies based on lesion type. The model includes four lesion growth patterns: flat urothelial carcinomas, lesions with low invasive potential, lesions with high invasive potential, and non-urothelial malignancies.

Kystis accounts for two types of symptoms: macroscopic hematuria (visible blood in the urine) and irritative voiding symptoms (e.g., frequent or painful urination). The instantaneous rate of voiding symptoms is influenced by lesion location and the proportion of the bladder surface covered by lesions, while the rate of hematuria is determined by lesion size. The probability of clinical detection depends on race, sex, symptom duration, and the presence of advanced cancer (MIBC or metastases). Diagnostic delays are assumed to result in part from symptom misattribution, such as urinary tract infections in women, or barriers to seeking care.

A person can die from bladder cancer or other causes, with time to death modeled using competing NHPPP processes. The earliest event determines the cause of death. Mortality from causes other than bladder cancer is assumed to align with general population mortality rates, stratified by age, sex, race, and smoking history. Death from bladder cancer is only modeled for individuals with MIBC or metastatic disease, as the probability of dying from NMIBC is considered negligible.

Public health impact

The model is calibrated to fit U.S. epidemiological data and can infer metrics that are not observed directly, such as the age at which tumors first emerge, how soon they become detectable, and the timing of other key events in bladder cancer progression. Kystis enables both short- and long-term projections of bladder cancer burden and helps evaluate the feasibility and impact of screening and surveillance interventions in the general population and high-risk subgroups.

References

1. Jalal H, Kang S, Trikalinos TA. Comparative Modeling of the Burden of Bladder Cancer in the United States. 2025.
2. Trikalinos TA, Sereda Y. The nhppp package for simulating non-homogeneous Poisson point processes in R. PLoS One. 2024;19(11):e0311311. doi:10.1371/journal.pone.0311311
3. Holford TR, Levy DT, McKay LA, et al. Patterns of birth cohort-specific smoking histories, 1965-2009. Am J Prev Med. Feb 2014;46(2):e31-7. doi:10.1016/j.amepre.2013.10.022
4. Hensley PJ, Panebianco V, Pietzak E, et al. Contemporary Staging for Muscle-Invasive Bladder Cancer: Accuracy and Limitations. Eur Urol Oncol. Aug 2022;5(4):403-411. doi:10.1016/j.euo.2022.04.008