Background: Breast MRI has the potential to non-invasively measure response to neoadjuvant chemotherapy (NACT). We studied the effect of varying two analytic parameters used to define MRI-measured tumor volume in the prediction of pathologic complete response (pCR) to NACT and to determine if optimization of these parameter thresholds would improve the prediction of pCR.
Methods: Women with locally advanced breast cancer (tumor size ≥ 3cm) were enrolled in the ACRIN 6657 / I-SPY 1 TRIAL. Each patient had up to four dynamic contrast-enhanced MRI examinations: before NACT (MR1), after one cycle of NACT (MR2), between the anthracycline-based regimen and taxane (MR3), and after NACT and prior to surgery (MR4). Breast cancer was stratified by subtypes of hormone receptor (HR), and human epidermal growth factor receptor 2 (HER2) status: HR+/HER2, HER2+, and triple negative ((TN) HR-/HER2-). MRI-measured functional tumor volume (FTV) and change in FTV (ΔFTV) were investigated as predictors of the outcome pCR. FTV is defined as the image volume with enhancement kinetics exceeding both an early percentage enhancement threshold (PEt) and a signal enhancement ratio threshold (SERt). Primary study analysis used empirically determined values. For this study PEt was varied from 30% to 200% in 10% intervals, and SERt was varied from 0.0 to 2.0 in 0.2 unit intervals. FTV was measured at each examination (FTV1, FTV2, FTV3, FTV4). ΔFTV was measured relative to the first examination (ΔFTV2, ΔFTV3, ΔFTV4). For each pair of varied PEt and SERt thresholds, the absolute and relative FTVs were re-measured and analyzed for discrimination of pCR using the area under the curve (AUC) of the receiver operating characteristic curve.
Results: A total of 116 patients were included from the ACRIN 6657 / I-SPY 1 TRIAL who had complete data on all four MRI visits, HR/HER2 status, and pCR outcome. Mean age was 48 years old (range 29-69). The full cohort of 116 patients was divided into subgroups: 45 (39%) HR+/HER2-; 39 (34%) HER2+; and 30 (26%) TN. When stratified by subtypes, lower AUCs with less variation were observed in patients with HER2+ cancer than patients with HR+/HER2- and TN breast cancer. When examining prediction by visit, maximum AUCs were found at later time points in all patient cohorts. Specifically, maximum AUC was observed for the full cohort at ΔFTV3 with AUC of 0.78 (CI: 0.69–0.87) when PEt=130% and SERt=0; for HR+/HER2- subtype at ΔFTV3 with AUC of 0.9 (CI: 0.84–0.97) when PEt=130% and SERt=0 were the same as in the full cohort; for HER2+ subtype at FTV3 with AUC of 0.77 (CI: 0.62–0.92) when PEt=70%/SERt=1.4; for triple negative at FTV4 with AUC of 0.89 (CI: 0.76–1) when PEt=40%/SERt=2.0.
Conclusion: This analysis suggests that the thresholds of MRI quantitative DCE measurements may need to be adjusted by breast cancer subtype to improve the predictive performance. The PEt threshold may need to be set higher in HR+/HER2- than other subtypes, which may be due to higher background parenchymal enhancement among HR+ patients. SER threshold may need to be set at higher level for triple negative subtype. A validation is underway in I-SPY 2, with a larger patient population.