Comparison of predictive models for local recurrence of DCIS after breast conserving surgery.

Abstract

21 Background: Efforts have been made to identify women for whom radiotherapy (RT) after breast-conserving surgery (BCS) for ductal carcinoma in-situ (DCIS) can be omitted. These include development of the Memorial Sloan-Kettering (MSK) nomogram, the Van Nuys Prognostic Index (VNPI), and the DCIS Score. We assessed correlation of the predictive qualities of these 3 approaches to determine risk of local recurrence (LR) and need for RT. Methods: We calculated VNPI score and used the MSK nomogram to determine LR risks for 48 women on whom DCIS Scores were available since January 2012. Given that the DCIS Score does not account for HT, MSK risks were calculated with and without its use. 10-year LR risks from DCIS Score reports were compared with MSK nomogram estimates using the Pearson correlation coefficient. These comparisons were done for MSK LR risks with and without HT. VNPI scores yield low, intermediate and high risk categories. To compare the 3 predictive models, each patient was assigned an RT recommendation per instrument. Two 10-year LR risk “thresholds” to recommend RT were used - 10 and 15% - to account for M.D. preference. Results: For LR risk, the MSK nomogram (excluding HT) and the DCIS Score showed weak positive correlation (R = 0.346, p < 0.02), weaker when including HT (R = 0.179, p: 0.224). VNPI Score could not be calculated on 5 patients and they were excluded in comparisons involving VNPI. DCIS Score, using 15% LR risk as the RT recommendation threshold, and VNPI RT recommendations correlated in 72.1% of cases. Lowering the threshold to10% reduced the correlation to 32.6%. Comparison of the 3 approaches’ RT recommendations with a 15% threshold showed agreement in 65.1% (28/43) of cases, and a 10% threshold reduced the agreement to 18.6% (8/43). Conclusions: Our analysis shows some correlation between the MSK nomogram and the DCIS Score, more when HT is excluded. Lowering the threshold 10-year LR risk to recommend RT from 15 to 10% reduced agreement among the 3 instruments.