A prediction model based on dual-layer spectral detector computed tomography for distinguishing nonluminal from luminal invasive breast cancer.

BACKGROUND

The identification of the molecular subtypes of breast cancer is critical to determining appropriate treatment strategies and assessing prognosis. This study aimed to evaluate the ability of dual-layer spectral detector computed tomography (DLCT) metrics to differentiate luminal from nonluminal invasive breast cancer.

METHODS

A total of 220 patients with invasive breast cancer who underwent routine DLCT examination were included in the study. The molecular subtypes of breast cancer were identified through immunohistochemical staining of biopsies or postoperative pathological specimens. DLCT quantitative parameters were compared between the luminal and nonluminal types of breast cancer. The diagnostic efficacy of these parameters was determined via receiver operating characteristic (ROC) curves. Univariate and multivariate regression analyses were conducted to identify independent predictors that could differentiate nonluminal from luminal breast cancer. A nomogram prediction model was established based on multivariate regression analysis. The performance of the nomogram model was assessed with ROC curve and calibration curve analyses.

RESULTS

Among the DLCT quantitative values, eight were significantly lower in the luminal type than in the nonluminal type of breast cancer (P<0.001-0.011). The area under the curve (AUC) values for these significant DLCT quantitative parameters ranged from 0.604 to 0.694. Multivariate logistic regression analysis identified CT-reported lymph node metastasis status [hazard ratio (HR) =4.214; P<0.001], the Hounsfield unit (HU) value of the virtual monoenergetic image at 40 keV (HU) (HR =2.628; P=0.012), and the normalized iodine concentration (nIC) (HR =2.182; P=0.041) as independent predictors of the nonluminal type, with an AUC of 0.754 [95% confidence interval (CI): 0.688-0.820]. The nomogram based on multivariate logistic regression analysis exhibited good discrimination and calibration (Hosmer-Lemeshow test; P=0.835). An average AUC value of 0.75 was obtained for the internal validation data.

CONCLUSIONS

DLCT quantitative parameters are valuable noninvasive indexes for differentiating between the luminal and nonluminal types of invasive breast cancer. Furthermore, the nomogram constructed in this study could guide individualized predictions of molecular subtypes in patients with invasive breast cancer.