Radiologic, Pathologic, and Deep Learning Predictors of Response to Immune Checkpoint Blockade in Renal Cell
Payal Kapur1,2,3, Alana Christie3, Vipul Jarmale4
1Department of Pathology, University of Texas Southwestern Medical Center, Dallas, TX, 75390.
Medrxiv : the Preprint Server for Health Sciences
|December 3, 2025
Summary
Assessing kidney cancer response to immunotherapy before surgery can guide treatment. Pathologic regression, quantified by deep learning, predicts better outcomes, differing from poor-prognosis coagulative necrosis.
Area of Science:
- Oncology
- Immunotherapy
- Radiology
- Pathology
- Machine Learning
Background:
- Assessing response to primary kidney tumor treatment in cytoreductive and neoadjuvant settings is crucial for adaptive postoperative strategies.
- Systematic analyses integrating radiology, pathology, and machine learning for response assessment are currently lacking.
Purpose of the Study:
- To develop and validate a quantitative framework for assessing post-immunotherapy response in renal cell carcinoma (RCC) using radiology, pathology, and deep learning.
- To correlate radiologic and pathologic response with clinical outcomes in patients receiving neoadjuvant or cytoreductive immunotherapy before nephrectomy.
Main Methods:
- Retrospective analysis of 99 advanced RCC patients treated with ICI-containing doublet therapy prior to nephrectomy.
- Central review of radiologic and pathologic features, correlating with freedom from next systemic therapy (FFNT) and metastasis-free survival (MFS).
- Development of deep learning (DL) models to objectively assess pathologic regression and quantify immune infiltrate.
Main Results:
- Radiologic tumor shrinkage (≥30%) and extent of ICI-induced pathologic regression (assessed by central review and DL) were significantly associated with prolonged FFNT.
- Deep learning models objectively validated pathologist-assessed regression and quantified immune infiltrate, showing significant association with outcomes.
- Multivariable analysis identified pathologic regression, DL-derived immune infiltrate, and tumor largest dimension as independent predictors of FFNT.
Conclusions:
- This study presents an integrated, quantitative framework for post-ICI response assessment in RCC, differentiating immune-mediated regression from coagulative necrosis.
- Findings suggest a complementary role for radiology and pathology in evaluating post-nephrectomy specimens after immunotherapy.
- Prospective validation could guide adaptive treatment strategies and clinical trial design for immunotherapy in kidney cancer and other malignancies.
