A prospective cohort study of hepatic toxicity after stereotactic body radiation therapy for hepatocellular carcinoma.

Purpose: To build and validate multivariate normal tissue complication probability (NTCP) models for radiation-induced hepatic toxicity (RIHT) after stereotactic body radiation therapy (SBRT).

Methods: Eighty-five patients with hepatocellular carcinoma (HCC) in a phase II clinical trial were enroled. A progression of at least 1 or 2 points in the Child-Pugh (CP) score post-SBRT was classified as RIHT (≥1 or ≥2). NTCP models for RIHT (≥1 or ≥2) were developed using logistic regression. Nomograms for each model were formulated. The cut-off point of each independent dosimetric risk factor was obtained using receiver-operating characteristic (ROC) analysis. We used an independent cohort (101 patients) for model validation.

Results: Twenty (23.5%) and 12 (14.2%) patients experienced RIHT (≥1) and RIHT (≥2), respectively. V, VS, and pretreatment CP (pre-CP) were the optimal predictors for RIHT (≥1 and ≥2) modelling. V ≤33.1% and VS ≥416.2 mL for RIHT (≥1), and V ≤21.5% and VS ≥621.8 mL for RIHT (≥2), were the cut-off points. Four NTCP models and their nomograms were generated. These models and nomograms showed good prediction performance (area under the curve (AUC), 0.83-0.89). Our NTCP model (RIHT ≥2) based on V plus pre-CP performed well (AUC = 0.78) in a validation cohort.

Conclusion: V, VS, and pre-CP are crucial predictors for RIHT (≥1 and ≥2). Our NTCP models and nomograms were conducive to obtain individual constraints for patients with HCC.

Registration Number: ChiCTR-IIC-16008233.