SCOPE (Michigan) University Of Michigan

Overview

The University of Michigan Simulation of Cancer Outcomes and Policy Evaluation (SCOPE) model is a stochastic, individual-level microsimulation model of prostate cancer natural history, screening, diagnosis, treatment, survival, and mortality. The model extends the University of Michigan Self-Consistency Analysis of Surveillance (SCANS) analytic framework by preserving key SCANS natural history and survival components while adding an explicit individual-level representation of longitudinal PSA dynamics and latent tumor growth. It is designed to evaluate screening and diagnostic policies, including risk-adaptive strategies that use evolving patient information to guide screening intervals, diagnostic testing, biopsy decisions, and stopping ages.

SCOPE simulates individual life histories beginning with date of birth and other-cause mortality. Age at latent prostate cancer onset follows the SCANS formulation, with onset defined as the earliest age at which cancer could potentially be detected by biopsy. After onset, each individual is assigned a latent tumor growth trajectory. Tumor extent evolves over time according to an exponential growth model with subject-specific growth rates. This latent tumor extent is the central unobserved disease process in the model and drives PSA dynamics, stage progression, grade progression, and clinical detection.

PSA is modeled longitudinally on the logarithmic scale as the sum of a non-cancer component, a cancer-related component, and measurement error. The non-cancer PSA component follows a mixed-effects model with subject-specific intercepts and slopes, while the cancer-related component is proportional to latent tumor extent after onset. This formulation generates realistic PSA trajectories before and after disease onset and supports screening policies that depend on observed PSA levels over time.

Disease progression is represented by stage and grade processes that evolve as functions of latent tumor extent. Individuals begin with localized, Gleason 2-6 disease at onset and may transition to regional and distant stage and to Gleason grade 7 and 8+ categories over time. The joint distribution of stage and grade at clinical diagnosis is calibrated to targets derived from the SCANS model. Clinical diagnosis occurs according to a post-onset hazard that depends on tumor extent and is calibrated to reproduce the SCANS sojourn time distribution.

Screening and diagnostic pathways are superimposed on individual life histories according to externally specified policies. PSA screening determines downstream decisions such as additional testing and biopsy, and screening can advance the time of diagnosis relative to clinical detection in the absence of screening. Secondary tests, including MRI and biomarkers, are represented as binary outcomes. In the current implementation, MRI sensitivity depends on disease grade and the selected PI-RADS threshold, while biopsy sensitivity depends on disease grade and biopsy approach, including systematic biopsy, MRI-targeted biopsy, or combined biopsy approaches.

After diagnosis, treatment assignment depends on disease characteristics, age, race, and mode of detection under either idealized or population-based treatment scenarios. Treatment options include conservative management, radical prostatectomy, and radiotherapy. Conservative management is interpreted as active surveillance for screen-detected low-risk disease and as no curative treatment for higher-risk or regional disease. Active surveillance is modeled explicitly with confirmation biopsy, repeat biopsies, and possible transition to curative treatment based on progression, detection of upgrading, age, clinical diagnosis under the natural history, other-cause death, or a modeled transition-to-treatment time.

Prostate cancer-specific survival is modeled using baseline survival functions that depend on age at diagnosis, stage, grade, and race. Survival may be modified for secular improvements and for curative treatment. The benefit of screening is represented through a lead-time dependent cure mechanism for screen-detected localized or regional disease, in which a fraction of individuals are reassigned to die from other causes rather than prostate cancer. Prostate cancer death competes with other-cause mortality, which is modeled using age- and cohort-specific U.S. life tables.

Model outputs include individual life histories and aggregate outcomes such as screening tests, MRI tests, biopsies, screen-detected diagnoses, clinical diagnoses, overdiagnoses, overtreatments, treatment counts, prostate cancer deaths, other-cause deaths, life-years, and life-years without diagnosed prostate cancer. Incidence and mortality outputs can be summarized by age, calendar year, PSA, stage, grade, and mode of detection. Natural history outputs include age at onset, sojourn time, tumor growth rate, preclinical disease duration, stage and grade distributions, and PSA values at onset and diagnosis. The model has been evaluated against SEER incidence patterns and Cluster Randomized Trial of PSA Testing for Prostate Cancer (CAP) screening outcomes and has been used to compare candidate prostate cancer screening strategies, including MRI-based confirmation pathways and biopsy approaches, in the general U.S. population and Black men.

References

1. Tsodikov A, Szabo A, Wegelin J. A population model of prostate cancer incidence. Statistics in Medicine. 2006;25(16):2846-2866.
2. Tsodikov A, Chefo S. Generalized self-consistency: Multinomial logit model and Poisson likelihood. Journal of Statistical Planning and Inference. 2008;138(8):2380-2397.
3. Chefo S, Tsodikov A. Stage-specific cancer incidence: An artificially mixed multinomial logit model. Statistics in Medicine. 2009;28(15):2054-2076.
4. Wang S, Tsodikov A. A self-consistency approach to multinomial logit model with random effects. Journal of Statistical Planning and Inference. 2010;140(7):1939-1947.
5. Tsodikov A, Liu LX, Tseng C. Likelihood transformations and artificial mixtures. Statistical Modeling for Biological Systems: In Memory of Andrei Yakovlev. Springer; 2020. p. 191-209.
6. Tsodikov A, Gulati R, Heijnsdijk EA, Pinsky PF, Moss SM, Qiu S, de Carvalho TM, Hugosson J, Berg CD, Auvinen A, et al. Reconciling the effects of screening on prostate cancer mortality in the ERSPC and PLCO trials. Annals of Internal Medicine. 2017;167(7):449-455.
7. Tsodikov A, Gulati R, de Carvalho TM, Heijnsdijk EA, Hunter-Merrill RA, Mariotto AB, de Koning HJ, Etzioni R. Is prostate cancer different in black men? Answers from 3 natural history models. Cancer. 2017;123(12):2312-2319.
8. Etzioni R, Tsodikov A, Mariotto A, Szabo A, Falcon S, Wegelin J, Ditommaso D, Karnofski K, Gulati R, Penson DF, et al. Quantifying the role of PSA screening in the US prostate cancer mortality decline. Cancer Causes & Control. 2008;19(2):175-181.