A mathematical model to predict for pre-malignant or malignant diagnosis among patients with Birad 4 breast lesions

Abstract

10578 Background: Preoperative cytological or histological diagnosis of breast lesions is mandatory in order to avoid unnecessary surgical biopsies, but on the other side preoperative work-up may dangerously delay specific care of breast cancer. Solid lesions or microcalcifications (M) scored as Birad 4 are increasingly prevalent in western countries. A highly variable proportion of these lesions (20–80%) are breast cancer. Tools to help clinicians recognize cancers and preneoplastic lesions among true benign conditions may be very helpful in clinical practice. Methods: Radiological, clinical and pathological data of consecutive patients with Birad 4 breast M (N = 384 biopsies among 354 patients) or nodular lesions (N = 172 FNAC among 167 patients) seen in a multidisciplinary breast clinic were prospectively recorded. A multivariate analysis of factors predicting for a final cancer or pre-malignant diagnosis was performed and two nomograms were constructed using the R statistical package for both nodular lesions and M. They were validated by bootstrapping. Variables tested included age, size and palpability of lesion, Gail score, menopausal status, HRT use, progression of lesion (M), and presence of associated symptoms. Results: Median age was 57 years (18–92) for the entire population. Patients with nodular lesions were menopausal in 64.5%, median size of their lesion was 12 mm (4–50), 32% were palpable; 43% had a final diagnosis of breast cancer and 3.5% of atypical hyperplasia or LCIS. 69% of patients with M were menopausal, 31.25% had a final diagnosis of breast cancer and 9.8% of atypical hyperplasia or LCIS. Among patients with nodular lesions, age and palpability were the sole independent predictors of cancer or precancerous lesions (p = 0.04 and 0.004), but the other variables (Gail, menopause, HRT) added discrimination with a concordance index of 0.71. Among patients with M, the only independent predictive variable was the recent progression of the lesions (p = 0.01). The nomogram had a concordance index of 0.69. Conclusion: Our study provides two original nomograms for the prediction of the pre-malignant or malignant nature of recently discovered solid breast lesions and M. Gail model alone is not a highly useful tool in daily individual cancer prediction. No significant financial relationships to disclose.