Predictive value of DCE-MRI for early evaluation of pathological complete response to neoadjuvant chemotherapy in resectable primary breast cancer: A single-center prospective study

Abstract

ObjectiveThis study proposed to establish a predictive model using dynamic enhanced MRI multi-parameters for early predicting pathological complete response (pCR) to neoadjuvant chemotherapy (NAC) in breast cancer. MethodsIn this prospective cohort study, 170 breast cancer patients treated with NAC were enrolled and were randomly grouped into training sample (136 patients) and validation sample (34 patients). DCE-MRI parameters achieved at the end of the first cycle of NAC were screened to establish the predictive model by using multivariate logistic regression model according to pCR status. Receiver operating characteristic curves were conducted to assess the predictive capability. The association between MRI-predicted pCR and actual pCR in survival outcomes was estimated by using the Kaplan–Meier method with log-rank test. ResultsMultivariate analysis showed ΔAreamax and ΔSlopemax were independent predictors for pCR, odds ratio were 0.939 (95%CI, 0.915 to 0.964), and 0.966 (95%CI, 0.947 to 0.986), respectively. A predictive model was established using training sample as “Y = −0.063*ΔAreamax − 0.034*ΔSlopemax”, a cut-off point of 3.0 was determined. The AUC for training and validation sample were 0.931 (95%CI, 0.890–0.971) and 0.971 (95%CI, 0.923–1.000), respectively. MRI-predicted pCR patients showed similar RFS (p = 0.347), DDFS (p = 0.25) and OS (p = 0.423) with pCR patients. ConclusionThe multi-parameter MRI model can be potentially used for early prediction of pCR status at the end of the first NAC cycle, which might allow timely regimen refinement before definitive surgical treatment.