Background: The aim of the study was to determine factors and develop the model that predict the risk of invasion in patients with ductal carcinoma in situ (DCIS) diagnosed by preoperative biopsy.

Methods: We selected 497 patients underwent surgical treatment for preoperatively diagnosed DCIS of the breast between 1997 and 2009. Multivariate analysis was used to identify relevant clinical, radiological and pathological factors that may predict upstaging. A prediction model was developed based on significant factors and measured using the area under the receiver operating characteristics (ROC) curve. This prediction model was subsequently validated using the dataset of 149 cases who were diagnosed with DCIS preoperatively between 2010 and 2012.

Results: Of 24.75%, there was an upstaging to the invasive cancer. We found that larger size (≥1.5cm) of DCIS was most significant factor, followed by palpable lesion on physical examination, no expression of estrogen receptor (ER)/progesterone receptor (PR), high nuclear grade, non-cribriform subtype and absence of necrosis significantly associated with presence of invasion.

The prediction model with these factors showed that excellent predictive accuracy (the area under the ROC curve= 0.851, 95% CI: 0.817–0.881) and showed similar findings in the validation data set (AUC = 0.868, 95% CI: 0.800–0.916). After ranking the significant factors, we selected 4 factors to simplify the model. This simple model also showed good prediction power (AUC = 0.821, 98% CI: 0.784–0.854).

Conclusion: We developed the prediction model to predict upstaging in patients diagnose with DCIS preoperatively based on 4 most significant factors. By this prediction model, we suggest the axilla exploration with SLNB in case of DCIS with more than 1.5 cm size, palpable lesion, no expression of HR and high NG to avoid 2nd operation.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P2-10-41.