Abstract P4-01-14: Can MRI predict the response to neoadjuvant chemotherapy in breast cancer accurately?

Abstract

Abstract Introduction: Magnetic resonance imaging (MRI) has been increasingly used to assess response to neoadjuvant chemotherapy (NAC). The aim of this retrospective study was to investigate the accuracy of MRI in predicting pathological complete response (pCR) and residual disease in patients treated with NAC. Methods: 152 patients with invasive breast cancer were treated with NAC (2005–2009). In our institute the response to neoadjuvant anthracycline (AC) based chemotherapy is monitored with MRI scanning before NAC and after 2 and 4 cycles. Taxane based chemotherapy was substituted for AC chemotherapy if MRI response was deemed inadequate. Response to NAC was measured using the modified response evaluation criteria in solid tumours (RECIST). Tumour extent and response on final MRI were correlated with the pathological findings on post-surgical specimens. pCR was defined as absence of invasive cancer. Sensitivity and specificity of MRI reporting was determined against response on histopathology. Potential predictive factors for response on MRI were assessed for significance. Results: In total, patients had 2(14.5%), 3(60.5%) or 4(25%) MRI scans. pCR was seen in 37(24.3%) patients. After 2 cycles of NAC, positive response (partial or complete) was seen on MRI in 76(50%) patients. Response was more likely in focal tumours versus multifocal tumours (chi square (χ2) 3.83, p = 0.05). In the non-responding group (n = 76), 74 had their treatment switched and a later response was detected in 50(67.6%) cases on MRI. In the 133 patients with 3 or more MRI scans, final MRI size was compared to tumour size on pathology. Median tumour extent on final MRI was 25.5mm (range 0–120) and median whole tumour size on pathology was 22.5mm (range 0–120). MRI response was apparent in 115 (75.7%) cases (sensitivity = 92.3%, specificity = 56.3% and PPV = 93.9%). In a 2-tailed analysis MRI response showed strong correlation with actual pathological response (spearman's = 0.501, p < 0.0001). MRI correctly diagnosed pCR in 15 of 37(40.5%) patients with pCR. Linear regression shows that overall tumour extent on MRI was highly predictive of whole tumour size on pathology (coefficient = 0.459, p < 0.0001). However, MRI overestimated residual disease in 24(18%) cases (median size difference of 39mm, range 12–120mm) and underestimated minimal residual disease in 12(9%) cases. In univariate analysis, time of response to NAC was a predictor of pCR (p = 0.013), with greater number of early responders having pCR than late responders. Complete response on MRI was a significant predictor of pCR (p < 0.0001), with a greater proportion of patients having pCR compared to those not showing complete response on MRI. Both complete response on final MRI (p = 0.276) and pCR (p = 0.069) show favourable disease-free survival (DFS), however neither reaches statistical significance on Kaplan-Meier curve (Median follow up of 43 months, range 13–78). Conclusions: Early responders and tumours achieving complete response on MRI are significant predictors of histopathological response. Serial MRI is predictive of response to NAC and possibly also a predictor of DFS. In this series MRI overestimated residual disease in nearly 1 in 5 patients. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P4-01-14.