DCE-MRI and MBI in Assessing Tumor Response to Chemotherapy in Patients With Triple Negative Breast Cancer

The ability to assess and predict response will be compared among the imaging modalities and with standard pathological evaluation. These volume changes will be correlated with residual cancer burden (RCB) status after surgery and will be used to classify patients into predicted responder or non-responder categories. Predictive accuracies among the imaging modalities will be compared using paired ROC curve analyses.

The index tumor will be measured on both cranial-caudal (CC) and medio-lateral oblique (MLO) views and 3 dimension will be recorded. The percentage change in volume at time points 2 and 3 relative to baseline imaging (time point 1) will be calculated. Data acquisition and image processing algorithms will be developed from having conjugate views of the breast in the MBI examination. Appropriate methods to correct the image from loss of contrast due to scatter and loss of signal from photon attenuation as it transits breast tissue will be explored and implemented. Will investigate the correlations of absolute and relative values of baseline standardized uptake value (SUVb) with pathological tumor response.

Correlations of pathological response with absolute and relative values of MBI SUVb will be investigated. The longest dimension of the residual tumor will be measured. If only foci of disease are seen the longest dimension of tumor cell distribution will be measured. The M D Anderson Cancer Center (MDACC) Residual Cancer Burden Calculator will be used to categorize cancer burden: RCB-0 (no residual disease in breast or in lymph nodes), RCB-1 (minimal residual disease), RCB-2 (moderate residual disease), or RCB-3 (extensive residual disease).

Assessed by baseline SUVb and baseline genomic signature. Will fit one logistic regression model with baseline SUVb and baseline genomic signature as covariates and response to the upfront chemotherapy as the endpoint and one model with baseline DCE-MRI tumor volume and baseline genomic signature as covariates and response to the upfront chemotherapy as the endpoint. Will assess the predictive accuracy of this model by estimating the area under the ROC curve using the predicting response probability and the observed response outcome. ROC curves will be generated for each modality and each TNBC molecular subtype and compute the areas under these curves along with 95% confidence intervals. Will determine a cut-point in the predicted probabilities to classify patients as predicted responders and non-responders based an appropriate trade-off between sensitivity and specificity.