99P Development of a risk assessment and prediction model for DCIS recurrence: Insights from analysis of 600 patients

This is the authors' abstract. We don't add key points for this paper.

Predicting DCIS recurrence remains challenging. Our study validates local recurrence risks in DCIS patients based on criteria from landmark and ongoing trials, aiming to aid treatment decisions and develop a novel prediction model. Analyzing patients diagnosed with DCIS who underwent breast-conserving surgery at our institution over a span of 20 years, from 1999 to 2021. We assessed recurrence risk using Kaplan-Meier estimates. Rates of initial local recurrence were compared across various trial recommendations and eligibility criteria. We used penalized Cox regression and reported AuROC to assess predictive performance. Our study comprised 600 DCIS patients, with median follow-up durations of 10.69 and 6.76 years in recurrence and non-recurrence groups, respectively. Among them, 71 patients experienced a first local recurrence (11.80%). Recurrence included DCIS in 36 cases (50.70%) and invasive cancer in 35 cases (49.30%). Most recurrences were ipsilateral (53 cases, 74.60%), followed by contralateral (16 cases, 22.50%), and bilateral (2 cases, 2.80%) breast tumor recurrence. In the table: LORD showed the lowest risk (0% over 10 years) with grade 1 histology. LORETTA, including up to grade 2 histology, tumor size ≤ 2.5 cm, and ER positivity, demonstrated the second lowest risk (6.7% over 10 years). These findings prioritize grade 1 histology, followed by tumor size ≤ 2.5 cm and ER status. Our preliminary analysis found three significant variables: hormonal treatment, tumor grade, and re-wide excision using penalized Cox regression. Radiotherapy notably predicted lower local recurrence rates.Table: 99PIIBTR from Landmark or Ongoing trialsTotalLowHighHRC-indexRTOG 9804 10 year 15 year60094 (16%) 8.4% 21.7%506 (84%) 16.2% 21.7%1.60 (0.73-3.46) p=0.2390.53COMET 10 year 15 year600195 (32.5%) 10.4% 24.9%405 (67.5%) 17.3% 21.3%1.58 (0.90-2.76) p=0.1100.58LORD 10 year 15 year60019 (3%) 0% 0%581 (97%) 15.7% 22.9%NANALORETTA 10 year 15 year60059 (10%) 6.7% 15.2%541 (90%) 16.0% 22.5%1.96 (0.71-5.38) p=0.1910.54LORIS 10 year 15 year60064 (11%) 9.1% 16.0%536 (89%) 15.9% 22.6%1.46 (0.63-3.36) p=0.3790.53Low1Low2HighECOGAcrinE5194 10 year 15 year600302 (50%) 14.9% 24.8%105 (18%) 13.6% 18.7%193 (32%) 16.1% 18.2%High vs Low1 1.17 (0.71-1.95) p=0.539 High vs Low2 1.43 (0.67-3.07) p=0.3480.56 Open table in a new tab Based on our 10-year data validation, Grade 1 histology, tumor size ≤ 2.5 cm, and ER-positive status are key in identifying low recurrence risk. Factors like no re-wide excision and radiotherapy may also reduce recurrence. Future plans include refining risk assessment in DCIS patients with a prediction model.