The value of chest CT for prediction of breast tumor size: comparison with pathology measurement

Abstract

Background: Little information is available on the use of chest computed tomography (CT) to predict breast tumor size in breast cancer, despite the fact that chest CT examinations are being increasingly used. The purpose of this study was to evaluate the value of chest CT for predicting breast tumor size using pathology measurements as reference standards. Methods: Tumor sizes (defined as greatest diameter) were retrospectively measured on the preoperative chest CT images of 285 patients with surgically proven unifocal, invasive breast carcinoma. Greatest tumor diameters as determined by chest CT and pathologic examinations were compared by linear regression and Spearman's rho correlation analysis. Concordance between CT and pathology results was defined as a diameter difference of <5 mm. Subgroup analyses were also performed with respect to tumor size (<20 mm or >= 20 mm) and histological subtype (invasive ductal carcinoma(IDC) or non-IDC). Results: CT and pathology measured diameters were found to be linearly related (size at pathology = 1.086 x CT determined tumor size - 1.141; Spearman's rho correlation coefficient = 0.84, P<0.001). Most tumors (n = 228, 80.0%) were concordant by chest CT and pathology, but 36 tumors (12.7%) were underestimated by CT (average underestimation, 11 mm; range, 6-36 mm) and 21 tumors (7.4%) were overestimated (average overestimation by CT, 10 mm; range, 6-19 mm). The concordance rate between the two sets of measurements was greater for tumor of <20 mm and for IDC (P<0.001 and P = 0.011, respectively). Conclusions: Tumor size by chest CT is well correlated with pathology determined tumor size in breast cancer patients, and the diameters of the majority of tumors by chest CT and pathology differed by <5 mm. In addition, the concordance rate was higher for breast tumors of <20 mm and for tumors of the IDC histologic subtype.