Qualitative CT Features as Predictors of Microscopic Perinephric Fat Invasion in Renal Cell Carcinoma: Towards a Radiomics-Informed Staging Framework
Year: 2026
Type: preprint
Abstract: Purpose Up to 14% of cT1 renal cell carcinomas (RCC) treated with partial nephrectomy are upstaged to pT3a owing to microscopic perinephric fat invasion (PFI) not visible on preoperative computed tomography (CT). We aimed to identify reproducible qualitative CT predictors of microscopic PFI. Methods Consecutive patients undergoing partial nephrectomy for cT1 RCC without overt local invasion (2009–2020) were enrolled in a prospective database. Pre-specified CT features were scored by a blinded primary radiologist and re-scored by two further blinded readers. Associations with PFI were modelled using Firth penalised logistic regression and inter-reader agreement with Fleiss’ κ. A two-feature nomogram was internally validated by bootstrap. Results Of 135 patients, 117 formed the imaging cohort; 10 (8.5%) had PFI. Tumour margin irregularity (odds ratio [OR] 5.20, 95% confidence interval [CI] 1.45–22.6) and perinephric neovascularity (OR 4.84, 95% CI 1.31–18.0) predicted PFI on univariable analysis, whereas fat stranding did not (OR 0.95). Three-reader agreement was substantial (κ 0.67–0.74). After mutual adjustment, neither feature remained significant (margin irregularity OR 3.56, 95% CI 0.78–17.8; neovascularity OR 2.41, 95% CI 0.55–10.9). The nomogram achieved an optimism-corrected area under the curve of 0.71 (apparent 0.73, 95% CI 0.54–0.90). Conclusion Margin irregularity and neovascularity are candidate CT predictors of microscopic PFI in cT1 RCC, whereas peritumoural fat stranding is not. Tumour margin irregularity, the strongest univariable predictor with substantial inter-reader agreement, is a practical marker to report when staging cT1 RCC and a biologically anchored target for future radiomics models, pending external validation.
Source: Research Square
Language: English
Subfield: Oncology
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7225781564