Long-Term Outcomes and Cost-Effectiveness of Artificial Intelligence for Breast Cancer Screening: A Modeling Study

Artificial intelligence (AI)-assisted breast cancer screening may improve diagnostic accuracy; however, the long-term health outcomes and cost-effectiveness of AI-assisted screening is unknown. We estimated benefits, harms, and cost-effectiveness of incorporating an AI product, Saige-Dx, into standard screening with digital breast tomosynthesis (DBT).

We developed a microsimulation model using nationally representative data from Surveillance, Epidemiology, and End Results, the Breast Cancer Surveillance Consortium, and published data on AI performance. The model compared biennial screening for women aged 40 to 74 using DBT to DBT plus AI. We estimated false-positive and false-negative screens, breast cancer cases by stage, and breast cancer deaths per 1000 women screened over a lifetime. We also estimated quality-adjusted life-years (QALYs), costs, and the incremental cost-effectiveness ratio.

In a cohort of 1000 women screened from ages 40 to 74, AI-assisted screening reduced false-negative screens by 2.1 and false positives by 49 resulting in 0.33 fewer advanced breast cancer cases (regional or metastatic cancer) at diagnosis and 0.13 fewer breast cancer deaths compared with DBT alone. Screening with AI resulted in 3.09 additional QALYs and an increase in lifetime costs of $936,430 per 1000 women, yielding an incremental cost-effectiveness ratio of $303,279 per QALY. In 98% of simulations, AI was not cost-effective at a $100,000/QALY willingness-to-pay threshold. Findings were not sensitive to changes in test characteristics likely to be observed in routine practice.

AI-assisted breast cancer screening yielded modest reductions in breast cancer mortality, but at current pricing, AI is not cost-effective.