Pharmacokinetic (PK) analysis of high temporal resolution (tRes) DCE-MRI data has been shown effective for early prediction of breast cancer response to neoadjuvant chemotherapy (NACT) (1,2). However, high tRes breast DCE-MRI studies are currently limited to research and early phase clinical trial settings. Due to the trade-off of tRes and spatial resolution (sRes) in data acquisition and clinical needs for bilateral full breast coverage and high sRes, low tRes (60-120 s) breast DCE-MRI protocols are commonly used in large-scale clinical trials and clinical practice. Consequently, because of inaccuracies in PK parameter estimation from low tRes data (3,4), semi-quantitative analysis (such as uptake slope, etc.) is often employed for low tRes data. Unlike quantitative PK parameters (such as K^trans) which are direct measures of biological properties, semi-quantitative metrics are directly related to MR signal change, not tissue biology, and the values are often dependent on data acquisition protocols and scanner platforms and settings, making it difficult to compare studies across institutions. There has been no literature evidence on whether PK analysis of low tRes data can still provide useful early prediction of breast cancer therapy response despite expected PK parameter errors. Here we report our initial results comparing PK analyses of low and high tRes breast DCE-MRI data for early prediction of NACT response, using data sets from the same patient cohort.

15 breast cancer patients enrolled in a multicenter ISPY-2 NACT trial consented to high tRes (14-18 s) research DCE-MRI (2) at visit 1 (V1, before NACT), V2 (after 1 NACT cycle), V3 (at NACT midpoint), and V4 (after NACT). They also underwent a low tRes (80-100 s) ISPY-2 DCE-MRI protocol at the same four time points. PK analyses of the low and high tRes DCE-MRI data were performed using the Shutter-Speed model (SSM) (2) which takes into account transcytolemmal water exchange kinetics. Tumor mean PK parameter values were calculated by averaging tumor voxel parameter values from all slices covering the tumor, which included K^trans, v_e, k_ep (=K^trans/v_e), and the SSM-unique τ_i parameter, mean intracellular water lifetime.

Estimated PK parameters from the low and high tRes data at V1 and V2, and the percent changes (V21%, V2 relative to V1) were compared, and correlated with pathologic response status (determined from resection specimens after NACT) to assess abilities for early prediction of response through ROC analysis. A nonparametric method was used to compare ROC AUC between results from the two tRes data sets.

Following NACT, 4 patients had pathologic complete response (pCR) while the other 11 had non-pCR. Table 1 lists tumor mean ± SD values of V1, V2, and V21% PK parameters estimated from the high and low tRes data, showing statistically significant underestimations of K^trans, k_ep, and τ_i and overestimation of v_e from the low tRes data compared to the high tRes data. However, there were no significant differences in V21% values of these parameters. For example, Fig. 1 shows scatter plots of V2 and V21% K^trans from the high and low tRes data. Table 2 lists the ROC AUC values of several DCE-MRI metrics for early discrimination of pCR vs. non-pCR. For each metric there was no statistically significant difference in ROC AUC between the two tRes data sets.

The findings of K^trans underestimation and v_e overestimation from the low tRes data are consistent with a previous study (3). The errors in PK analysis of low tRes data are largely systematic with PK parameter values changing in the same direction going from low to high tRes. This is why there are no significant differences in V21% values, and the likely reason that DCE-MRI metrics that are good early predictors of NACT response when obtained from the high tRes data perform comparably well in early prediction when obtained from the low tRes data (Table 2). This preliminary study suggests that despite expected errors in estimated PK parameters, PK analysis of low tRes DCE-MRI data could be useful for assessment of breast cancer therapy response. Since low tRes data is usually collected in large-scale breast cancer clinical trials, the utility of PK analysis of low tRes data for therapy response evaluation may have significant impact on future imaging biomarker development, taking advantage of the large, retrospective database from the past and current trials that include breast DCE-MRI.