E.H. Reason1, S.M. Thomas2,3, J.K. Plichta2,4, A.M. Botty Van Den Bruele2,4, E.S. Hwang2,4, B.C. Tong2,4, N.A. Larrier5, J.E. Grilley-Olson2,6, L.H. Rosenberger2,4 1Duke University School of Medicine, Durham, NC, USA 2Duke Cancer Institute, Durham, NC, USA 3Duke University Medical Center, Dept. Of Biostatistics And Bioinformatics, Durham, NC, USA 4Duke University Medical Center, Dept. Of Surgery, Durham, NC, USA 5Duke University Medical Center, Dept. Of Radiation Oncology, Durham, NC, USA 6Duke University Medical Center, Dept. Of Medicine, Durham, NC, USA
Introduction: Phyllodes tumors (PT) are rare breast neoplasms classified as benign, borderline, and malignant based on scaled histologic features. Local recurrence rates increase with higher grades, and distant recurrence events have a dismal prognosis. The Singapore nomogram predicts any recurrence event utilizing stromal cytologic atypia, mitoses per 10 high-power field (HPF), presence of stromal overgrowth and final surgical margins. The nomogram was superior in predicting recurrence compared to a histological score. We aimed to test the Singapore nomogram for accuracy of PT recurrence events in a US cohort and compare to the histological score.
Methods: Women with PT were selected from a prospective institutional database. Histological parameters and margin status were used to estimate the (1) nomogram score and (2) histological score, per patient, as previously defined. Univariate and multivariate Cox proportional hazards models were used to estimate the association of (1) individual factors, (2) nomogram and (3) histological scores with recurrence-free survival (RFS). Harrel’s C-index was estimated for each model.
Results: Of 53 PT cases, 28.3% were benign (N=15), 35.8% borderline (N=19), and 35.8% malignant (N=19). Recurrences occurred in 32.1% (N=17). Univariable analyses demonstrated higher mitotic count per 10 HPF (HR 1.04, 95% CI 1.02-1.06, p=0.001), stromal overgrowth (HR 6.69, 1.86-24.0, p=0.004), and mitotic rate ≥10 (vs <5, HR 14.29, 1.92-100, p=0.009) were associated with significantly lower RFS. The adjusted RFS analysis including the four factors utilized in the Singapore nomogram performed well (C-index of 0.81). However, despite a higher nomogram score being associated with increased risk of recurrence (HR 1.02, 1-1.05, p=0.05), the individual numeric scale defined in the nomogram only moderately fit our data (C-index of 0.64), Table 1. Patients with higher histological scores also had increased risk of recurrence (HR 1.53, 1.19-1.97, p<0.001) with a C-index of 0.78.
Conclusion: Both nomogram and histological scores can be used to predict PT recurrence. Unlike other international cohort analyses, histological score served as a more accurate predictor of PT recurrence, and the nomogram performance was not consistent with prior studies. This may be due to the high proportion of borderline and malignant PT in this cohort with a known skewed distribution. The Singapore nomogram, developed in a cohort with a low proportion of malignant PT, may perform better in cohorts with lower proportions of malignant PT (5.9%, C-index of 0.93, 7.0%, C-index 0.90), and worse when the portion of malignant PT rises (13.7%, C-index 0.76). Refining the nomogram scores may result in improved performance in malignant PT.