Optimising screening intervals and BI-RADS management pathways in ultrasound-based breast cancer screening: real-world evidence from Shenzhen (2018-2023).

PURPOSE: To identify screening intervals and BI-RADS-specific management pathways that balance cancer detection yield against recall burden in ultrasound-based breast cancer screening in China. METHODS: We analysed 1,694,838 ultrasound examinations reported with BI-RADS (2018-2023). For BI-RADS 1/2, inter-screen intervals were evaluated by recall rate, cancer detection rate (CDR), positive predictive value (PPV), and early-stage proportion. For BI-RADS 0/3/4/5, we assessed associations between biopsy waiting time and PPV/early-stage proportion, and examined imaging follow-up outcomes for BI-RADS 4A. RESULTS: Among 337,497 screening intervals following BI-RADS 1/2, longer intervals were associated with higher CDR (0.13-0.54 per 1,000) but higher recall (17.4%-22.1%) and lower PPV (61.1%-31.0%), while early-stage proportion remained stable. The interval that best balanced yield and downstream burden was 13-24 months for women aged ≥ 40 years and 25-36 months for women aged < 40 years, achieving higher detection yield and PPV without a material increase in recall. Among 1,949 BI-RADS 4/5 biopsies, PPV was low for 4A (12.3%-25.0%), declined markedly with biopsy delay for 4B (60.5% to 25.0%), and remained consistently high for 4C/5 (> 60%/>93%). Among BI-RADS 4A lesions 64.7% (10,909/16,871) showed imaging downgrading during follow-up. In 487 BI-RADS 3 cases undergoing biopsy, delayed biopsy increased PPV without changing early-stage proportion. CONCLUSION: A 2-year screening interval is recommended for women aged 40-64 years, extendable to 3 years for those aged < 40. Prompt biopsy is warranted for BI-RADS 4B/4C/5. The high image downgrade rate of BI-RADS 4A suggests overestimation and supports imaging-based triage, requiring further validation. Imaging surveillance is preferable for BI-RADS 3 lesions.