ideasIdea
Roll out AI-supported mammography nationally as a stepped-wedge trial
Sweden's MASAI trial showed AI can safely replace one of two radiologists. Rolling it out region by region in a randomised order would prove it works at national scale and that interval cancers do not rise.
MASAI (Lund) demonstrated a 44% workload reduction and higher cancer detection with AI-supported reading. Rather than piecemeal adoption, propose national programmes adopt in a randomised stepped-wedge order so every region serves as its own control, with interval cancer rate and recall rate as co-primary endpoints.
Hypothesis
National AI-supported reading reduces radiologist reading load by at least 40% with no increase in interval cancers and no increase in DCIS-only detection.
Rationale
Stepped-wedge gives regulator-grade evidence at the cost of an implementation programme, and detects harms (recall, DCIS inflation) that single-centre trials cannot.
What would test it
Run a three-year stepped-wedge across about ten regions of a national programme with a pre-registered analysis.
Maturity
being tested at scale
Who has to act
policy
Cost to try
Medium ($1M to $50M)
Years to first evidence
4
Bottlenecks it attacks
- Most lethal cancers are found late · Screening exists for only a few cancers. Pancreatic, ovarian, liver, oesophageal and most lung cancers are found when cure is unlikely.
- Not enough oncologists, nurses, pathologists, physicists · The number of people with cancer is rising faster than the workforce trained to treat them.