630P - Determining a prognostic score using imaging to assess the benefit of combo anti-PD1 + anti-CTL4 vs anti-PD1 in patients with metastatic MSI/dMMR colorectal cancer (mCRC MSI)

Background

The aim is to determine, using quantitative imaging features from baseline CT-scans and clinical factors, the response to mono vs combo immunotherapy for metastatic patients with mCRC MSI.

Methods

This retrospective study included 105 patients treated either with a single anti-PD1 (mono) or a combination of anti-PD1 and anti-CLA4 (combo). Baseline scanners were annotated manually in 2D on the axial plane slice with the largest diameter for each visible lesion by 6 expert radiologists. Tumor volume was computed by summing the volume of each lesion, and reproducibility was assessed using Inter-correlation coefficient (ICC2k). Cox models were used to predict overall survival (OS) on bootstrap samples, and the concordance index was used to assess its performance on out-of-bag samples. Progression free survival (PFS) was evaluated. Recursive feature elimination (RFE) was used to reduce the number of variables. Cutoffs were determined using maximally selected rank statistics and were used to binarize features, and a risk score was built using weights derived from the Hazard ratios. The final model was modified to exclude the type of treatment, and the cutoff was adjusted.

Results

In total, 1633 lesions were annotated. ICC2k was equal to 0.95 for total tumor volume. Of the 13 variables studied, 5 were selected after the RFE: age, type of treatment, presence of peritoneal carcinomatosis (PC), number of lesions and total tumor volume. The cutoffs for total tumor volume, number of lesions and age were 73 cm3, 20 lesions and 60 years old. The weights for total tumor volume, number of lesions, age and PC were 1.13, 0.96, 0.91 and 0.38 and the cutoff of the risk score was 1.51. Patients with a low risk score had similar OS and PFS independently of treatment, while those with a high score had significantly worse OS and PFS if treated with monotherapy (p=0.004 and p=0.0001).

Conclusions

Total tumor volume had the highest weight amongst other quantitative imaging measures and age, emphasizing the need to compute tumor burden before initiating treatment. This score could allow oncologists to identify mCRC MSI patients who need combo therapy, limiting its unnecessary side effects compared to monotherapy.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

Gustave Roussy.

Funding

DATAIA and Gustave Roussy.

Disclosure

All authors have declared no conflicts of interest.