A general two-stage progressive model of cancer natural history to project downstaging due to multi-cancer screening tests

TL;DR

This paper introduces a two-stage model to estimate how multi-cancer screening tests could reduce advanced-stage cancer diagnoses by detecting cancers earlier, using data-driven projections validated with lung cancer data.

Contribution

The paper develops a general, data-driven two-stage model to project the impact of multi-cancer early detection tests on stage shifts across multiple cancer types.

Findings

Model accurately predicts stage shift in lung cancer.

Projects significant reduction in advanced-stage diagnoses for liver, pancreas, and bladder cancers.

Framework can be applied to evaluate population benefits of MCED tests.

Abstract

Multi-cancer early detection (MCED) tests offer to screen for multiple types of cancer with a single blood sample. Despite their promising diagnostic performance, evidence regarding their population benefit is not yet available. Expecting that benefit will derive from detecting cancer before it progresses to an advanced stage, we develop a general two-stage model to project the reduction in advanced-stage diagnoses given stage-specific test sensitivities and testing ages. The model can be estimated using cancer registry data and values for the mean overall and advanced-stage preclinical sojourn times. We first estimate the model for lung cancer and validate it against the stage shift observed in the National Lung Screening Trial. We then estimate the model for liver, pancreas, and bladder cancer, which have no recommended screening tests, and we project stage shifts under a shared MCED testing protocol. Our framework transparently integrates available data to project reductions in advanced-stage diagnoses due to MCED testing.