Early prediction of response to neoadjuvant chemotherapy in luminal breast cancer using total lesion glycolysis measured on 18F-FDG PET images: A prospective validation study

Objectives Luminal A and B breast tumors are characterized by variable and mostly limited response to neoadjuvant chemotherapy (NAC). We prospectively investigated the value of several 18F-FDG PET image-derived parameters for early prediction. Prediction based on PET was also compared to that obtained by clinical, histological and molecular markers.

Methods 64 luminal breast cancer patients were included and underwent 18F-FDG PET scans at baseline and before the third cycle of chemotherapy. Surgery was performed after 8 cycles of NAC and pathological response was assessed using the Sataloff scale. SUVmax and TLG image-derived parameters were extracted from PET images. The accuracy of image-derived parameters (delta between the two PET scans) or molecular markers such as progesterone or luminal status in identifying responders was assessed through receiver operating characteristic (ROC) analysis.

Results There were 27 responders and 37 non-responders. The best accuracy was obtained using ΔTLG, with an area under the ROC curve (AUC) of 0.81 (vs. 0.73, 0.71 and 0.63 for SUVmax, luminal status, and progesterone status, respectively). Median ΔTLG was -49±31% in non-responders (vs. -73±25% for responders; p<0.0001). ΔTLG had a sensitivity of 89% and a specificity of 74% in identifying non-responders.

Conclusions When they respond to neoadjuvant treatment, luminal tumors present more pronounced partial shrinkage at early evaluation. ΔTLG seems to be more adapted for early prediction than ΔSUVmax and also a more powerful predictor than biological parameters.