IntroductionThe integration of artificial intelligence (AI) in oncological diseases has opened new avenues for early cancer detection, accurate risk assessments, and personalized treatment protocols. The objective of our study was to describe the shift in the treatment indications and prognosis of ductal carcinoma in situ (DCIS) in a prospective cohort of 998 patients managed over 18 years at a single University Hospital in France. MethodsThis analysis included all patients managed for DCIS at the University Hospitals of Strasbourg between January 14, 2002, and December 18, 2019, from the prospective SENOMETRY cohort, which initially included 9599 patients with both in situ and invasive breast cancer. Data were analyzed using Onconum, an AI-based natural language processing tool, to extract structured information from medical records. ResultsThe incidence of DCIS remained stable at 65 new cases per year. The mean age at diagnosis increased from 55 years in 2002 to 61 years in 2019. There was a significant rise in high-grade DCIS (DIN3) cases from 25% to 35%. The re-excision rate decreased from 57% in 2002 to 18% in 2019. Most DCIS cases were managed with breast-conserving surgeries (682), while total mastectomies accounted for 385 cases. Sentinel lymph node biopsy was performed in 39% of cases, primarily in high-grade and multifocal DCIS. Specific mortality was 0%, with a recurrence rate of 2.2%, predominantly invasive and occurring earlier in high-grade DCIS. Discussion: Over 18 years, there has been a notable shift in the clinicopathological characteristics of DCIS, with an increase in patient age at diagnosis and higher histopathological grades. Therapeutic management evolved significantly, with reduced surgical margins and fewer adjuvant treatments, while maintaining low and stable recurrence rates. AI significantly enhanced data extraction and analysis efficiency, contributing to better clinical decision-making. ConclusionThe study confirms the possibility of therapeutic de-escalation in DCIS, supported by AI-driven data analysis, allowing individualized treatment approaches and leading to optimized patient outcomes.
Read full abstract