Validation of the Michigan General Cancer Screening Model (M-GCSM) Using the National Lung Screening Trial Dataset

Principal Investigator

Name
Tanner Caverly

Degrees
M.D.

Institution
Department of Veterans Affairs

Position Title
Associate Professor, Research Investigator

Email
tanner.caverly@va.gov

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-1513

Initial CDAS Request Approval
May 11, 2026

Title
Validation of the Michigan General Cancer Screening Model (M-GCSM) Using the National Lung Screening Trial Dataset

Summary
This project seeks to validate the Michigan General Cancer Screening Model (M-GCSM), a parsimonious cancer screening simulation designed to be broadly applicable across cancer types and accessible to a wide range of users. The M-GCSM uses readily available epidemiological and RCT-derived inputs to model the dynamics of cancer screening from a health systems perspective. This model will enable a straightforward way to compare the effects of competing risk across cancer types and changes in screening policy using the same simulation structure and assumptions. We will use NLST data to derive transition probabilities for lung cancer incidence that are a function of stage and histology, as well as stage-specific lung cancer mortality and competing risk. NLST data will also be used to validate the simulation by comparing predicted relative risk ratios of lung cancer mortality and comparing stage specific incidence and mortality curves.

Aims

1. Create a strategically simplified simulation model for cancer screening that can be used to probe the factors and dynamics across a wide range of cancer types within a shared set of assumptions and simplifications.
2. Derive lung cancer transition probabilities from NLST participant-level data, including incidence, stage at diagnosis, and survival stratified by stage and histology, as well as competing risk probabilities.
3. Validate the M-GCSM against observed NLST outcomes by comparing model-predicted lung cancer-specific and all-cause mortality to empirical trial results.
4. Explore deriving probabilities for simplified cancer stage designation (early-stage, late-stage) and assess the impact on performance and accuracy after simplification.

Collaborators

Tanner Caverly Department of Veterans Affairs
Rod Hayward University of Michigan