1108P - Impact of quantitative imaging of all disease on the prognostic value of ¹⁸F-FDG PET/CT in patients treated with immunotherapy for metastatic melanoma

Background

¹⁸F-FDG PET/CT is crucial in guiding immunotherapy for patients with metastatic melanoma (MM). Quantitative imaging offers a comprehensive view of disease burden, however optimal use of this information is undetermined. This study aimed to investigate the prognostic value of quantitative imaging information derived from FDG PET/CT for survival analysis.

Methods

103 MM patients received immunotherapy: pembrolizumab (n=60), ipilimumab (n=7), nivolumab (n=12), or ipilimumab + nivolumab (n=24). TRAQinform IQ software (AIQ Solutions) tracked lesion-ROI on FDG PET/CT scans retrospectively collected from baseline (BL) and first on-treatment follow-up (FU) images. Imaging features, including size, intensity, percent change, and heterogeneity, were extracted. Cox Proportional Hazards CoxPH models assessed Overall Survival (OS) and Progression Free Survival (PFS) with varied input features and lesion-ROI subsampling. Model performance, evaluated using the C-index (C) of 1000 bootstrap iterations, was compared with paired t-tests.

Results

Model outcomes are summarized in the table. For OS: three CoxPH models with all lesion-ROI and subset of features reached max C=0.81; four models with all features and subset of lesion-ROI achieved max C=0.85. Top model, C=0.87, included all features and lesion-ROI. For PFS: three models with all lesion-ROI and subset of features reached max C=0.58; four models with all features and subset of lesion-ROI achieved max C=0.73. Best model, C=0.77, included all features and lesion-ROI. Paired t-tests among bootstrap samples reveal significant p-values (

Conclusions

FDG PET/CT image analysis enabled the extraction of image features including characterizing lesion heterogeneity of change. The inclusion of all lesion-ROI and all features including lesion heterogeneity information helped improve the prognostic value of multivariable models for both OS and PFS.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

University of Western Australia.

Funding

AIQ Solutions.

Disclosure

R. Jeraj: Financial Interests, Personal, Officier: AIQ Solutions. O. Lokre: Financial Interests, Full or part-time Employment: AIQ Solutions. M. Dell'Oro: Financial Interests, Financially compensated role, AIQ Australia Pty Ltd in collaboration with UWA have established AIQ Research Fellows - full time research fellowships in medical imaging. Dr Dell’Oro holds one of these Fellowships: AIQ Solutions. R.J. Francis: Non-Financial Interests, Advisory Board, Author R.F. is a scientific advisory board member of AIQ Solutions: AIQ Solutions. T. Perk: Financial Interests, Full or part-time Employment: AIQ Solutions. All other authors have declared no conflicts of interest.