In current clinical practice, clinical breast examination, mammography, US, MRI, and PET-CT have been used to evaluate response in patients receiving NAC. In a comparative study of imaging modalities in 43 patients, Shin and colleagues showed increasing correlation with pathologic tumor size, with intra-class correlation coefficients of 0.65 for clinical breast exam, 0.69 for mammography, 0.78 for US and 0.97 for MRI (33). In a recent meta-analysis of 969 patients from 18 studies, MRI had a summary ROC (sROC) AUC of 0.89 (13). When comparing PET-CT to MRI in a meta-analysis of 13 studies with 575 patients, MRI was more accurate, summary ROC (sROC) AUC: 0.88 vs. 0.84 (34). In contrast, in a meta-analysis of 10 studies, PET-CT outperformed MRI during NAC treatment (15). Mammography and/or breast MRI are recommended by NCCN (35) and CT, MRI, FDG PET are endorsed by RECIST 1.1 (36). US has not been recommended for monitoring disease status in clinical trials due to its perceived subjectivity and operator dependence (37).
However, because of its low cost and accessibility, US has been utilized and evaluated in several studies (9–12). In the GepharTrio trial of 2090 patients bidimensional US measurements at EOC2 were used to classify response and randomize patients to treatment with 6 vs 8 cycles of NAC (11). Candelaria and colleagues evaluated mid-treatment breast US in 159 patients and showed association of percentage change in tumor US measurements with RCB in TNBC and hormone receptor (HR) + and HER2– tumors but not in HR– and HER2 + tumors (10). Marinovich and colleagues applied RECIST 1.1 and WHO 1D and 2D measurement criteria in 832 patients who underwent US at EOC2 and demonstrated an average increase in AUCs of 2% and 3% to 0.79 and 0.80 respectively, with the addition of US to patient characteristics including biomarkers (9). In this study, we found that the fractional change of US maximum diameter (%US), measured at EOC1 (AUC = 0.83) was more predictive than EOC2 (AUC = 0.68). We further showed a substantial 11% increase, AUC = 0.77 to 0.88 at EOC1, in patients with known ER + or HER2 + disease. We believe that early US measurements at EOC1 can substantially avoid measurement uncertainty caused by treatment induced inflammatory and fibrotic changes in the tumor bed.
Recently studies of shear wave elastography (SWE) have shown that increased extracellular matrix stiffness is associated with tumor progression and NCT resistance (38–39). SWE measures significant differences in tumor elasticity changes in pCR vs. non-pCR cases and AUCs of 0.613 (baseline), 0.745 (EOC1), 0.685 (EOC2) have reported in a recent clinical trial (40). This tread is similar to our study using gray scale US.
Diffuse optical tomography and diffuse optical spectroscopy exploit changes in tumor vascularity and metabolism and have demonstrated the potential for early prediction of breast cancer pathological response (18–26). Studies have shown accurate predictions in the neoadjuvant setting by utilizing pretreatment hemoglobin levels and changes in hemoglobin early in the course of treatment (19,21–26), or by monitoring changes of blood oxygen saturation sO2 at day 1 of dose dense treatment (18) or day 10 during early treatment (20). In the recent ACRIN 6691 trial evaluating 36 patients, the authors derived a tissue optical index (TOI), a product of deoxygenated Hb and water concentration over lipid, and reported that the mid-treatment TOI can predict pCR with AUC 0.6 to 0.83 (19). Gunther and colleagues developed a dynamic diffuse optical tomography system that could distinguish between response groups. An ROC analysis showed that this method could identify patients with a pCR two weeks into the treatment with AUC = 0.85 (17). In an earlier investigation using data from two studies with limited patients treated with new dual HER2 blockade regiment (19), Zhu and colleagues developed prediction models identifying HER2 status and HbT as the best pretreatment predictors of pCR (AUC = 0.88). The pretreatment predictors “ER status and HbT”, and “TNBC and HbT” predicted response with moderate AUC accuracies of 0.69 and 0.72, which are similar to the single predictor HbT (AUC = 0.75). With known HER2 positivity, the best window to accurately predict response was at the completion of the first and second cycles of NAT (AUC = 0.96, AUC = 0.97). For ER+/HER2- or TNBC subtype, the best window was at the completion of the first cycle of NAT and the best predictors were HbT and %HbT (AUC = 0.95).
In this new cohort of 38 patients, “HER2, ER and pretreatment HbT” has shown good prediction, AUC = 0.80 and HbT alone has shown moderate prediction AUC = 0.71. However, fractional reduction of HbT (%HbT) is a much more powerful predictor of response as is fractional reduction of maximum diameter measured with US (%US) in the first three cycles. In particular, combining tumor HER2 and ER status, %US EOC1 and %HbT at EOC1 provided the best early indicator of treatment response, AUC = 0.941, and remained powerful even without biomarker data, i.e. AUC = 0.910. As expected from log cell-kill kinetics of cytotoxic drugs, a given dose kills a constant proportion of a tumor cell population rather than constant number of cells (41). Therefore, for chemo-sensitive tumors, there are more total cells killed in the first cycle of treatment and more tumor neovasculature damage which may cause a significant decrease in tumor hemoglobin measured by the DOT system and size reduction measured by US. Overall, the highest accuracy, AUC = 0.974, was achieved with the combination of %US EOC1 and %HbT at EOC3 irrespective of biomarker status. To our knowledge, these AUC values are among the highest reported results using NAT regimens in current clinical practice.
From an exam delivery perspective, the combination of US and US-guided DOT provides important potential benefits for assessing NAT response compared to PET-CT and MRI. In the system used in this study, a commercial US probe was used for US measurements and then placed within the fiber-optic array to guide DOT. Furthermore, in future iterations US could be completely integrated into a hybrid US – DOT system. The combined exam has no ionizing radiation, has low intrinsic cost, is easily adaptable to current equipment, and could be made widely available and portable. In contrast PET-CT and MRI are high cost and have limited access. Additionally, MRI has associated medical risks and contraindications. Disadvantages to US/US-guided DOT include: US-guided DOT imaging is not real-time and reconstruction currently takes a minimum of 20 to 30 minutes in computation time. US-guided DOT is not suitable for imaging tumors in the dark nipple-areolar complex, and requires a sonographically visible index tumor and normal contralateral reference tissue. US-guided DOT data acquisition requires training of operators to match the location of the lesion side and the contralateral reference side for differential measurements and can be performed by radiologists or US technologists. Dark skin absorbs more light than white skin, however, the DOT system has adequate source power to acquire measurements with good signal to noise ratio for a range of skin colors.
Our study has a number of limitations. This is a single institution experience and the imaging procedure was performed (US) and supervised (US-guided DOT) by dedicated breast imaging radiologists, which may affect the generalizability of results. The treatment regimens were based on current practice at a research institution and were not limited to a single regimen. The choice of systemic therapies for breast cancer patients are based on multiple factors, including tumor biology, stage, patient characteristics and wishes, clinical trial availability. The study population was not large (n = 38) and to develop prediction models, we included data from 22 patients from an earlier study with similar but not identical study criteria (19). While attempts were made to image the index tumor in the same orthogonal scan planes, i.e., radial-antiradial or transverse-sagittal, in order to optimize tumor conspicuity during treatment this was not always possible.
We chose our comparison groups as pCR and near pCR as “responders”, (Miller-Payne grades 4–5) versus “non-responders” (Miller-Payne grades 1–3), because the MP grading system is used to evaluate index tumor while residual cancer burden (RCB) additionally incorporates DCIS and lymph node status, which are not assessed by our DOT procedure. In our study Miller-Payne grade and RCB were highly correlated. Symmans et al. using the Residual Cancer Burden (RCB) system and another separate dataset for evaluating tumor response after neoadjuvant chemotherapy also found that pCR and near-pCR had very similar survivorship curves after surgery (30). Given that Ogston et al. originally reported that Miller-Payne grades 4 and 5 tended to track together with regard to 5 year disease free survive after neoadjuvant chemotherapy (29), we felt justified in considering MP 4 and MP 5 collectively as the responder group.
Only one patient was treated with an antiestrogen regimen, i.e. anastazole, in our study cohort. Due to the limited sample size, we have grouped this patient with the rest of the patients and used pCR as surrogate endpoint. However, pCR is uncommon in neoadjuvant endocrine treatment and a poor surrogate for disease free survival. Future studies assessing response to neoadjuvant endocrine therapy will likely rely on change in the proliferation marker Ki67 and the preoperative endocrine prognostic index (PEPI), a composite score of post-treatment ER, Ki67, tumor size and axillary nodal status, as a surrogate for endocrine sensitivity. Higher Ki67 and PEPI scores have shown to correlate with an increased risk of relapse (42).
Our study has substantial implications for the combined use of tumor subtypes, conventional US and near-infrared-measured tumor hemoglobin content in accurately predicting pathological response as soon as one treatment cycle is completed. A recently phase 3, open-label trial involving patients with HER2-positive early breast cancer who were found to have residual invasive disease has shown that adjuvant trastuzumab emtansine (T-DM1) for 14 cycles reduced the risk of recurrence of invasive breast cancer or death by 50% as compared with trastuzumab alone (6). Another trial of 910 HER2– residual invasive breast cancer patients after neoadjuvant chemotherapy showed that adjuvant capecitabine was safe and effective in prolonging disease-free survival and overall survival (43). If the residual disease could be accurately estimated earlier in the NCT, patients with an unsatisfactory response could be switched to investigational therapies or even definitive surgery as soon as cycle 1 is completed, allowing for personalized treatment. This ability will gain value as our armamentarium of interventions increases and responses can more effectively tailor the therapeutic agents selected.
