DU-CAM Duke University

To simulate uterine cancer incidence and survival in order to gain insight into factors affecting rising incidence and mortality, project future trends, and evaluate strategies for reducing the burden of disease through primary prevention, screening, improved diagnosis, and improved treatment.

Contact: Laura Havrilesky laura.havrilesky@duke.edu

Summary

The Duke University Uterine Cancer Model (DU-CAM) is a multistage clonal expansion (MSCE) model that simulates the natural history of uterine cancer, while accounting for age, race, trends in reproductive history (RH), body mass index (BMI), and prior hysterectomy. DU-CAM relates individual reproductive and BMI histories to cellular processes that contribute to the development of uterine cancer, using clonal expansion models to explicitly simulate processes of cell division, mutation, and promotion.

The DU-CAM model combines a stochastic two-stage clonal expansion (TSCE) model of premalignant cell initiation, clonal expansion, malignant transformation, and malignant lag-time, with a stochastic single stage clonal expansion (SSCE) model that begins with a single malignant cell that may undergo malignant clonal expansion, possible metastatic spread, and incidence (Figure 1).

DU-CAM Schematic

Grapic showing schematic of DU-CAM

Figure 1: Schematic of DU-CAM

Incident cancers are staged according to the American Joint Committee on Cancer (AJCC) stages I, II, III, and IV based on SSCE model stochastic threshold probabilities for malignant clone size and metastatic status, allowing estimation of the time-dependent probability of uterine cancer incidence at each age. The premalignant TSCE model is linked to the malignant SSCE model by calibrating the SSCE model mean time to cancer incidence to equal the estimated malignant lag time from the calibrated TSCE model.

We harmonized demographic, BMI, and RH variable data by race/ethnicity for NHANES-III (1991) and all NHANES surveys between 2000-2020. Using individual RH and corresponding BMI measurements, we estimated age- and year-specific BMI percentile distributions to simulate BMI trajectories across RH events. The advantage of this approach is that it inherently incorporates complex correlations between race, age, BMI and RH without the need to explicitly estimate those correlations.

DU-CAM models separate parallel cohorts stratified by race/ethnicity (non-Hispanic White, non-Hispanic Black, and Hispanic) and histology (endometrioid--EM, non-endometrioid—non-EM, and sarcoma). Individual histories of BMI and RH from the National Health and Nutrition Examination Surveys (NHANES) were tested using likelihood-based dose-response model comparisons to find the best fitting models for each race/ethnicity and histology group.

Calibration targets were SEER-18 uterine cancer incidence (EM, Non-EM, Sarcoma), AJCC stage distributions, and uterine cancer mortality. DU-CAM calibration was performed by single year of age and year of incidence for each histologic category (EM, Non-EM, and sarcoma) by diagnosis year and race/ethnicity.

The clonal expansion approach allows evaluation of the effect of trends in BMI and RH on specific aspects of the carcinogenic pathway to inform projections of future disease burden and estimate the potential effect of interventions designed at mitigating these effects through primary prevention, screening, improved diagnosis, and improved treatment.