Screening performance and characteristics of breast cancer detected in the Mammography Screening with Artificial Intelligence trial (MASAI): a randomised, controlled, parallel group, non-inferiority, single-blinded, screening accuracy study

Background Emerging evidence suggests that artificial intelligence (AI) can increase cancer detection in mammography screening while reducing screen-reading workload, but further understanding of the clinical impact is needed. Methods In this randomised, controlled, parallel-group, non-inferiority, single-blinded, screening-accuracy within the Swedish national screening programme, women recruited at four screening sites in southwest (Malm & ouml;, Lund, Landskrona, and Trelleborg) who were eligible for mammography screening were randomly (1:1) to AI-supported screening or standard double reading. The AI system (Transpara version 1.7.0 ScreenPoint Nijmegen, Netherlands) was used to triage screening examinations to single or double reading and as detection highlighting suspicious findings. This is a protocol-defined analysis of the secondary outcome measures cancer detection, false-positive rates, positive predictive value of recall, type and stage of cancer detected, reading workload. This trial is registered at ClinicalTrials.gov, NCT04838756 and is closed to accrual. Findings Between April 12, 2021, and Dec 7, 2022, 105 934 women were randomly assigned to the intervention control group. 19 women were excluded from the analysis. The median age was 53<middle dot>7 years (IQR 46<middle dot>5-63<middle dot>2). supported screening among 53 043 participants resulted in 338 detected cancers and 1110 recalls. Standard among 52 872 participants resulted in 262 detected cancers and 1027 recalls. Cancer-detection rates were 6<middle dot>4 (95% CI 5<middle dot>7-7<middle dot>1) screened participants in the intervention group and 5<middle dot>0 per 1000 (4<middle dot>4-5<middle dot>6) in the control ratio of 1<middle dot>29 (95% CI 1<middle dot>09-1<middle dot>51; p=0<middle dot>0021). AI-supported screening resulted in an increased detection cancers (270 vs 217, a proportion ratio of 1<middle dot>24 [95% CI 1<middle dot>04-1<middle dot>48]), wich were mainly small lymph-node cancers (58 more T1, 46 more lymph-node negative, and 21 more non-luminal A). AI-supported screening resulted in an increased detection of in situ cancers (68 vs 45, a proportion ratio of 1<middle dot>51 [1<middle dot>03-2<middle dot>19]), with of the increased detection being high-grade in situ cancer (12 more nuclear grade III, and no increase grade I). The recall and false-positive rate were not significantly higher in the intervention group (a ratio [95% CI 0<middle dot>99-1<middle dot>17; p=0<middle dot>084] and 1<middle dot>01 [0<middle dot>91-1<middle dot>11; p=0<middle dot>92], respectively). The positive predictive value of significantly higher in the intervention group compared with the control group, with a ratio of 1<middle dot>19 (95% CI p=0<middle dot>012). There were 61 248 screen readings in the intervention group and 109 692 in the control group, a 44<middle dot>2% reduction in the screen-reading workload. Interpretation The findings suggest that AI contributes to the early detection of clinically relevant breast reduces screen-reading workload without increasing false positives.