Radiologic, pathologic, and deep learning markers predict outcomes after post-immune checkpoint blockade nephrectomy in renal cell carcinoma

In 99 renal cell carcinoma (RCC) patients receiving immune checkpoint blockade before nephrectomy, ≥30% radiologic tumor shrinkage (p=0.0036) and greater pathologic regression (HR 0.97, CI 0.95-0.99) predicted longer time to next therapy.

• Deep learning–based quantitative pathologic assessment was concordant with central pathology review (HR 0.96; CI 0.93-0.99; p=0.0041); in exploratory multivariable Cox regression, pathologic regression, deep learning–derived immune infiltrate, and largest tumor dimension at nephrectomy remained associated with freedom from next systemic therapy.
• Coagulative tumor necrosis continued to be associated with poor outcomes.
• This was a single-institution retrospective cohort (n=99); deep learning models were not externally or prospectively validated in this study, so findings need prospective confirmation.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.