Relationship between functional imaging and immunohistochemical markers and prediction of breast cancer subtype: a PET/MRI study

Purpose: The aim of this study was to determine if functional parameters extracted from the hybrid positron emission tomography/magnetic resonance imaging (PET/MRI) correlate with the immunohistochemical markers of breast cancer (BC) lesions, to assess their ability to predict BC subtype. Methods: This prospective study was approved by the institution’s Ethics Committee, and all patients provided written informed consent. A total of 50 BC patients at diagnosis underwent PET/MRI before pharmacological and surgical treatment. For each primary lesion, the following data were extracted: morphological data including tumour-node-metastasis stage and lesion size; apparent diffusion coefficient (ADC); perfusion data including forward volume transfer constant (Ktrans), reverse efflux volume transfer constant (Kep) and extravascular extracellular space volume (Ve); and metabolic data including standardized uptake value (SUV), lean body mass (SUL), metabolic tumour volume and total lesion glycolysis. Immunohistochemical reports were used to determine receptor status (oestrogen, progesterone, and human epidermal growth factor receptor 2), cellular differentiation status (grade), and proliferation index (Ki67) of the tumour lesions. Correlation studies (Mann–Whitney U test and Spearman’s test), receiver operating characteristic (ROC) curve analysis, and multivariate analysis were performed. Results: Association studies were performed to assess the correlations between imaging and histological prognostic markers of BC. Imaging biomarkers, which significantly correlated with biological markers, were selected to perform ROC curve analysis to determine their ability to discriminate among BC subtypes. SUVmax, SUVmeanand SUL were able to discriminate between luminal A and luminal B subtypes (AUCSUVmean= 0.799; AUCSUVmax= 0.833; AUCSUL= 0.813) and between luminal A and nonluminal subtypes (AUCSUVmean= 0.926; AUCSUVmax= 0.917; AUCSUL= 0.945), and the lowest SUV and SUL values were associated with the luminal A subtype. Kepmaxwas able to discriminate between luminal A and luminal B subtypes (AUC = 0.779), and its highest values were associated with the luminal B subtype. Ktransmax(AUC = 0.881) was able to discriminate between luminal A and nonluminal subtypes, and the highest perfusion values were associated with the nonluminal subtype. In addition, ADC (AUC = 0.877) was able to discriminate between luminal B and nonluminal subtypes, and the lowest ADCmeanvalues were associated with the luminal B subtype. Multivariate analysis was performed to develop a prognostic model, and the best predictive model included Ktransmaxand SUVmaxparameters. Conclusion: Using multivariate analysis of both PET and MRI parameters, a prognostic model including Ktransmaxand SUVmaxwas able to predict the tumour subtype in 38 of 49 patients (77.6%, p < 0.001), with higher accuracy for the luminal B subtype (86.2%)