2074 - Optimal Central Hepatobiliary Tract Volumetric Expansion to Predict Adverse Events after MRI-Guided Liver SBRT: An NTCP Analysis

Purpose/Objective(s):

To identify the optimal portal vein (PV) expansion margin for predicting hepatobiliary events (HBE) in patients receiving MRI-guided stereotactic body radiation therapy (MRgSBRT) for central liver tumors.

Materials/Methods:

We retrospectively reviewed clinical and dosimetric data for patients with primary and secondary liver cancer treated with MRgSBRT from 2019-2025 at a single institution. Online adaptation was performed selectively to meet organs at risk constraints (e.g., mucosal V33Gy < 1 cc). Patients were treated in deep inspiratory breath hold (DIBH) with real-time gating. Planning target volumes (PTV) were generated using an isometric 3mm expansion from the gross tumor volume. The central hepatobiliary tract was delineated by uniform expansion off the PV (+ 5, 10, 15, 20, 25mm). Central lesions were defined by direct PV overlap, whereas peripheral lesions were located >25mm from the PV. HBEs included biliary tract infection, stricture formation, fistulization, ulcer, or unplanned stent interventions. We used univariate logistic regression and NTCP modeling to evaluate the relationship between HBEs and PV expansion dosimetry and calculated F1 scores for each expansion NTCP model performance. The Lyman-Kutcher-Burman variant of NTCP modeling was used, with model parameters from the RADBIOMOD program (n=0.86, m=0.12, TD50=39).

Results:

We analyzed 67 MRgSBRT plans from 65 patients. 39 (58.2%) were male and 34 (50.7%) were non-Hispanic White. Median age was 65 (IQR 55-72). 40 (59.7%) patients had primary liver cancer (29 cholangiocarcinoma, 11 hepatocellular carcinoma) and 27 had secondary liver cancer. Baseline Child-Pugh scores were A for 60 patients (89.5%), B for 5 patients (7.5%), and C for 2 patients (3%). 28 (41.8%) patients had biliary stents pre-MRgSBRT. Median PTV was 48.3cc (IQR 20.6-90.1). The most common prescription was 50 Gy in 5 fractions (76.1%). 24 HBEs (36%) occurred at a median of 3.1 months post-MRgSBRT (IQR 1.6–3.7), of which there were 3 CTCAE grade 2 (12.5%), 20 CTCAE grade 3 (83.3%) and 1 CTCAE grade 4 (4.2%). The majority of HBEs were hepatobiliary infections (62.5%). Across all PV expansions, targets directly overlapping the PV—or within a 5mm expansion margin—had the highest odds of an HBE (direct PV overlap – odds ratio [OR] 10.4, 95% CI: 3.0-36.1, p<0.01; 5mm PV expansion overlap - OR 10.7, 95% CI: 2.8-41.5, p<0.01). Median follow-up after MRgSBRT was 13.3 months (IQR 4.3-25.5). The 1- and 2-year freedom from local progression were 66.4% (95% CI 50.3-78.3%) and 59.3% (95% CI 42.2-72.9%). The NTCP model demonstrated the highest F1 scores for PV overlap (0.745) and PV+15mm expansion (0.733).

Conclusion:

High-grade HBE risk appears to be greatest within 5mm of the portal vein; however, a 15mm expansion margin may serve as an appropriate surrogate for NTCP modeling of hepatobiliary adverse events after MRgSBRT for central liver tumors.