Hawaii and Utah Add Bills + MASAI Trial
New State Insurance Bills
Two new state bills for expanded coverage for breast imaging: Hawaii HB2366, Utah HB0468. For existing state insurance laws, see our MAP.
MASAI Trial
A large, randomized, prospective, controlled, single-blinded non-inferiority trial including nearly 106,000 women in the Swedish breast cancer screening program was conducted to compare interval cancer rates between women screened with standard double reading versus Transpara AI-supported mammography to triage mammograms that should undergo double reading. The study included a mix of tomosynthesis and 2D mammography. Interval cancers were defined as primary breast cancers diagnosed between two screening rounds (typically every 1.5 – 2 years in average-risk women) or within 2 years after the last scheduled screening that were not detected at screening.
- The interval cancer rate was lower among women who received AI-supported mammography compared to those who had standard double reading.1
- This effect was consistent across age and breast density.
- Sensitivity was higher in the AI-supported mammography group2 and specificities were equal among both groups.3
AI-supported mammography improved cancer detection, reduced interval cancer rates and reduced radiologist workload by decreasing the need for double reading. In the United States, double reading is not routinely performed and isolated AI interpretation is not yet accepted: there are barriers to implementation. Further, AI is not generally covered by insurance so facilities in the USA may opt to charge patients an out-of-pocket fee when offering AI.
1 Interval cancer rates were 1.55/1000 (82/53043) (95% CI 1·23–1·92) and 1·76/1000 (93/52872) (95% CI 1·42–2·15) in the intervention and control groups respectively, with a non-inferior proportion ratio of 0·88 (95% CI 0·65–1·18; p=0·41).
2 Sensitivities were 80.5% (338/420) and 73.8% (262/355) in the AI-supported and standard double reading groups, respectively.
3 Specificity was 98.5% in both groups (51852/52623 in the AI supported group vs. 51752/52517 in the double reading group).