53P - Baseline PET/CT deep radiomics signature apply for identifying bevacizumab sensitivity of RAS-mutant colorectal cancer liver metastases patients

Background

Bevacizumab has significantly improved the resectability, response rate and survival of patients with RAS-mutant colorectal cancer liver metastases (CRLM). However, more than half of these patients were insensitive to bevacizumab therapy. Identification of patients who are sensitive to bevacizumab therapy may improve the response rate and reduce adverse events. In this study, we aimed to construct and validate a PET/CT deep radiomics signature to predict bevacizumab efficacy in initially unresectable RAS-mutant CRLM patients.

Methods

We retrospectively collected 208 RAS-mutant CRLM patients. Training cohort (n=74) included the members of armA (mFOLFOX plus bevacizumab) from the BECOME study (NCT01972490). Internal validation cohort (n=65) and external validation cohort (n=29) were collected, during January 2018 to December 2018, from the consecutive bevacizumab-treated RAS-mutant CRLM patients of Shanghai Zhongshan Hospital and First Hospital of Wenzhou, respectively. In order to exclude the effect of chemotherapy alone, a negative validation cohort (n=40) enrolled the members of armB (mFOLFOX alone) from the BECOME study. The PET/CT image features were extracted using a deep learning signature, and we converted them into a multi-scale representation by a Gaussian mixture model. This representation was further combined with relevant clinical factors to form the final radiomics signature.

Results

Our deep radiomics signature fitted well in the training cohort (AUC 0.982 [0.926-1.0]). As for internal validation cohort, our signature achieved a promising performance in predicting bevacizumab sensitivity (AUC 0.846 [0.794-0.869], sensitivity 0.752 [0.723-0.794], specificity 0.776 [0.743-0.814]), and the external validation cohort shows a similar outcome (AUC 0.768 [0.732-0.846], sensitivity 0.684 [0.647-0.734], specificity 0.696 [0.645-0.751]). But for the negative validation cohort, our signature failed with chemotherapy (AUC of 0.534 [0.467-0.592]).

Conclusions

A baseline PEC/CT deep radiomics signature was constructed and was able to specifically identify bevacizumab-sensitive RAS-mutant CRLM patients. This tool deserves to be validated by further prospective study.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The authors.

Funding

Fujian Provincial Health Commission Project (2021GGB032) and National Natural Science Foundation of China (82072653).

Disclosure

All authors have declared no conflicts of interest.