Wen Li^{1}, Lisa J Wilmes^{1}, David C Newitt^{1}, John Kornak^{2}, Ella F Jones^{1}, Savannah C Partridge^{3}, Jessica Gibbs^{1}, Bo La Yun^{1}, Matthew S Tanaka^{1}, Laura J Esserman^{4}, and Nola M Hylton^{1}

Diffusion weighted MRI can be used to characterize water mobility and cellularity of tumor by measuring the apparent diffusion coefficient (ADC). Functional tumor volume (FTV) measures size change along the treatment. This study showed that after only 3 weeks of pre-surgery chemotherapy, ADC added value to the logistic regression model of using FTV alone to predict pCR or RCB as outcomes for patients with advanced breast cancer in I-SPY 2 TRIAL. The effect of adding ADC is statistically significant and increase the estimated area under the ROC curve after adjusted for breast cancer subtype categorized by HR and HER2 status.

1. Hylton NM, Blume JD, Bernreuter WK, et al. Locally advanced breast cancer: MR imaging for prediction of response to neoadjuvant chemotherapy--results from ACRIN 6657/I-SPY TRIAL. Radiology. 2012;263(3):663-672.

2. Hylton NM, Gatsonis CA, Rosen MA, et al. Neoadjuvant Chemotherapy for Breast Cancer: Functional Tumor Volume by MR Imaging Predicts Recurrence-free Survival-Results from the ACRIN 6657/CALGB 150007 I-SPY 1 TRIAL. Radiology. 2016;279(1):44-55.

Figure 1 The plot of correlation associations
among FTV and ADC variables. ADC0—ADC3 represent mean ADCs for the whole tumor
at T0—T3. ADC 10—ADC30 represent the percent change at MRI time point T1–T3,
compared to T0. Similarly, FTV0—FTV3 represent absolute FTVs and FTV10—FTV30
represent percent changes. Correlation coefficients are plotted as circles,
colored by the values. Positive correlations are displayed in blue and negative
correlations in red color. Color intensity and the size of the circles are
proportional to the correlation coefficients. Correlations with p-value >
0.05 are marked with “X”.

Table 1 The logistic regression analysis for
predicting pCR. FTV10 and ADC10 represent the percent changes of FTV and ADC at
early treatment T1 compared to their values at pre-treatment T0. The estimated
odds ratios of FTV10 and ADC10 means that every 10% increase in FTV change at
the early treatment will increase the odds of being non-pCR at the surgery by
16% while 10% increase in ADC change will decrease the odds of being non-pCR by
13%, after adjusted for breast cancer subtypes.

Table 2 The logistic regression analysis for
predicting RCB. FTV10 and ADC10 represent the percent changes of FTV and ADC at
early treatment T1 compared to their values at pre-treatment T0. The estimated
odds ratios of FTV10 and ADC10 means that every 10% increase in FTV change at
the early treatment will increase the odds of being non-pCR at the surgery by
15% while 10% increase in ADC change will decrease the odds of being non-pCR by
16%, after adjusted for breast cancer subtypes.

Table 3 The AUCs of logistic regression models. A
cohort of 262 I-SPY 2 patients were included in the regression analysis of
predicting pCR. Among 262 patients, 31.3% achieved pCR at the surgery. Percent
changes of FTV and ADC at early treatment (FTV10 and ADC10) were tested in the
model alone or combined. Breast cancer subtype were also introduced as a
confounding variable. In this patient cohort, 10 were missing RCB outcome so
252 in total were included in the regression analysis of predicting RCB. As
shown in this table, 53.6% patients had RCB2 or RCB3 at the surgery.