News

Article

Radiomic Biomarkers Predict Response in Lung and Liver Uveal Melanoma Metastases

Author(s):

Fact checked by:

Key Takeaways

  • Radiomic biomarkers can enhance treatment response prediction in metastatic uveal melanoma, addressing RECIST criteria limitations.
  • The study of radiomic biomarkers aligns with Project Optimus, emphasizing dose optimization and early-phase drug development.
SHOW MORE

Radiomic biomarkers can predict treatment response through assessments of tumor growth in patients with metastatic uveal melanoma treated with roginolisib.

Radiomic Biomarkers in Uveal Melanoma Metastases | Image Credit: © ReisMedia - stock.adobe.com

Radiomic Biomarkers in Uveal Melanoma

Metastases | Image Credit: © ReisMedia

- stock.adobe.com

Radiomic biomarkers can be leveraged to predict treatment response through assessments of tumor growth in patients with metastatic uveal melanoma treated with roginolisib (IOA-244), according to findings from a study that were presented at the 2024 SITC Annual Meeting.1

“Radiomic feature analysis, complementary to volumetric analyses, demonstrates considerable promise for early-stage drug development by providing additional objective measures of treatment response,” lead study author Nikhil Sindhwani, PhD, of Radiomics.bio in Liège, Belgium, and coauthors, wrote in a poster of the findings.

A Project Optimus–Informed Study Rationale

The authors explained that the assessment of tumor response to treatment has traditionally been conducted via RECIST criteria. However, RECIST criteria have several limitations, including selective lesion assessment, operator dependence, and one-dimensional measurements. Therefore, the oncology field needs more sophisticated methods for evaluating tumor response, especially in early-phase clinical trials and the management of metastatic disease.

Project Optimus is an FDA Oncology Center of Excellence initiative that aims to reform dose-optimization and -selection paradigms in early-phase oncology drug development to allow for the selection of drug doses that maximize drug efficacy, safety, and tolerability.2 Its goals are to:

  • Share expectations for dose finding and optimization with the oncology community
  • Create opportunities for drug developers to meet with the FDA early in their development programs to discuss dose finding and optimization prior to conducting registrational clinical trials
  • Institute dose-finding and -optimization strategies that use both nonclinical and clinical data

The goals of this study aligned with those of Project Optimus.1 Investigators explored the use of radiomic biomarkers in screening and early follow-up scans, aiming to provide a nuanced and more comprehensive picture of treatment response that may improve clinical decision-making and drug development.

Study Design and Patient Characteristics

This study used data from a phase 1 trial (NCT04328844) that evaluated the novel PI3Kδ inhibitor roginolisib in patients with metastatic uveal melanoma. Evaluable metastatic lesions were stratified by location (lung vs liver), tracked across 4 time points (screening, week 8, week 16, and week 24), and categorized by size-specific thresholds (10-34 mm vs 35-49 mm vs 50-100 mm). The authors noted that lesions measuring 50 mm to 100 mm (“progressive” lesions) have different volume variation sensitivities compared with smaller lesions (“responsive” lesions).

The investigators explored radiomic biomarkers by extracting 120 quantitative features from CT scans across the 4 time points. Delta features, defined as the difference between measurements at screening and week 8, were used to identify early treatment response. Appearing lesions were excluded from this analysis because there are no screening values for comparisons between these lesions.

The study cohort included 23 patients, 11 male and 12 female, who were recruited from 3 centers. The median age was 54 years, and the median overall survival with roginolisib was 14.1 months. Patients had metastases across various organs; this analysis focused on lung and liver metastases because of insufficient statistical power for other organ sites.

Univariate Radiomic Analysis

A univariate analysis was used as an initial screening method to find individual radiomic features potentially predictive of treatment response. Investigators explored whether individual features at screening and early changes were significantly different from responsive and progressive lesions at week 24.

“This preliminary step helped identify promising features for subsequent multivariate analyses,” the authors wrote.

In liver lesions, this univariate analysis identified 5 baseline features (4 texture, 1 shape) and 6 delta features (1 texture, 5 shape). In lung lesions, 16 identified features were found to be significantly different between the progressive and responsive lesions, including 3 screening features and 13 delta features characterized by shape, intensity, and texture.

“The higher number of significant delta features, particularly in lung lesions, suggests that early changes in radiomic characteristics may be more informative than baseline measurements for predicting treatment response,” the authors explained.

Multivariate Radiomic Analysis

This multivariate analysis employed generalized linear models (GLMs) to differentiate between progressive and responsive lesions. GLMs created with early changes in radiomic delta features demonstrated superior performance compared with GLMs created with only baseline characteristics, with higher area under the curve (AUC) values for both liver (mean delta AUC = 0.86 [standard deviation (SD), 0.08], sensitivity = 0.95 [SD, 0.06], specificity = 0.49 [SD, 0.09]; mean baseline AUC = 0.68 [SD, 0.05], sensitivity = 0.97 [SD, 0.03], specificity = 0.13 [SD, 0.08]) and lung (mean delta AUC = 0.83 [SD, 0.04], sensitivity = 0.62 [SD, 0.07], specificity = 0.84 [SD, 0.09]; mean baseline AUC = 0.64 [SD, 0.10], sensitivity = 0.38 [SD, 0.12], specificity = 0.76 [SD, 0.06]).

“This suggests that early temporal changes in radiomic features may possess greater predictive value for treatment response assessment,” the authors noted.

Of the delta features, the change in minor axis diameter screening between baseline and week 8 was identified as the most influential predictor of treatment response, indicating that early lesion dimension reduction may indicate ultimate treatment response.

Longitudinal Feature Evaluation

Investigators conducted a longitudinal evaluation to analyze changes in radiomic features over time in responsive vs progressive lesions across 3 key effects: time variation, response category, and the interaction between the two. Lesions were only included in this evaluation if they were consistently present throughout the observation period.

A total of 145 progressive and 45 responsive lung lesions were included from 10 patients. Among 118 features that passed variance filtering, 89 demonstrated significant time effects, 18 demonstrated response effects, and 98 demonstrated interaction effects (P < .000423). A correlation reduction revealed 13 unique informative features.

A total of 59 progressive and 127 responsive liver lesions were included from 16 patients. Among 107 features that passed variance filtering, 63 demonstrated significant time effects, 59 demonstrated response effects, and 84 demonstrated interaction effects (P < .000467). A correlation reduction revealed 17 unique informative features.

This longitudinal analysis also identified distinct organ-specific response patterns. Liver lesions demonstrated radiographic and shape changes by week 8, whereas lung lesions underwent these changes later, by week 24.

“These temporal and organ-specific differences in response patterns suggests that radiomic features can capture unique aspects of treatment response, potentially due to variations in tissue microenvironment, vascularization, or drug penetration between organs,” the authors concluded.

References

  1. Sindhwani N, Rasooli A, Mawaz MH, et al. Baseline and early follow-up radiomics biomarkersfor tumor growth in uveal melanoma patients treated with roginolisib (IOA-244). Presented at: 2024 SITC Annual Meeting. November 6-10, 2024; Houston, Texas. Abstract 131.
  2. Project Optimus. FDA. Updated September 9, 2024. Accessed November 7, 2024. https://www.fda.gov/about-fda/oncology-center-excellence/project-optimus
Related Videos
Stephanie Graff, MD, and Chandler Park, FACP
Mariya Rozenblit, MD, assistant professor, medicine (medical oncology), Yale School of Medicine
Maxwell Lloyd, MD, clinical fellow, medicine, Department of Medicine, Beth Israel Deaconess Medical Center
Neil Iyengar, MD, and Chandler Park, MD, FACP
Azka Ali, MD, medical oncologist, Cleveland Clinic Taussig Cancer Institute
Rena Callahan, MD, and Chandler Park, MD, FACP
Hope S. Rugo, MD, FASCO, Winterhof Family Endowed Professor in Breast Cancer, professor, Department of Medicine (Hematology/Oncology), director, Breast Oncology and Clinical Trials Education; medical director, Cancer Infusion Services; the University of California San Francisco Helen Diller Family Comprehensive Cancer Center
Virginia Kaklamani, MD, DSc, professor, medicine, Division of Hematology-Medical Oncology, The University of Texas (UT) Health Science Center San Antonio; leader, breast cancer program, Mays Cancer Center, UT Health San Antonio MD Anderson Cancer Center
Aditya Bardia, MD, MPH, FASCO, professor, Department of Medicine, Division of Hematology/Oncology, director, Translational Research Integration, UCLA Health Jonsson Comprehensive Cancer Center
John M. Burke, MD