Vascular Invasion Prediction Via a Multimodal Model Supports Re-staging of Early-stage Hepatocellular Carcinoma

NCT05540925 | Completed DMC

• 3 other identifiers: EHBH-MVIPG372018ZX10723-204SHSLCZDZK02402
• Study Type: observational
• Target: 3,650
• Locations: 1 country
• Sites: 1

Brief Summary

Effective strategies for personalized treatment decisions for early-stage hepatocellular carcinoma (HCC) remain critically limited. Microvascular invasion (MVI) is a distinctive pathological hallmark of HCC invasiveness. Current evidence underscores that accurate preoperative prediction of MVI has the potential to support personalized selection among surgical resection, liver transplantation, and local ablation, as well as different approaches within each modality. However, despite numerous studies on MVI prediction, accuracy remains insufficient. Currently, radiomics and liquid biopsy are at the forefront of MVI prediction. Furthermore, combining both technologies may offer a more reliable prediction, but this integrated approach remains unexplored. This large-scale prospective cohort study (an observational study design). aims to develop a multimodal approach to predict MVI by integrating four data sources: clinicopathological variables, MRI-based radiomics features, MRI-based deep learning features, and MVI-related genomic alterations detected in circulating cell-free DNA (cfDNA). Individual models will first be developed for each data modality. Their predictive outputs will be calibrated into standardized risk probabilities and integrated via decision-level fusion to generate the final multimodal model. This system will be used to preoperatively stratify patients by predicted MVI risks. Based on this risk stratification, a re-staging system for early-stage HCC will be developed. Its prognostic value and clinical utility will be evaluated, with its ability to guide the selection of optimal resection margins to improve outcomes as a key illustrative example. The primary endpoints are 5-year recurrence-free survival (RFS), 5-year overall survival (OS), and recurrence patterns.

Conditions

Carcinoma, Hepatocellular; Hepatocellular Cancer; Liver Cancer

Keywords

hepatocellular carcinoma; re-staging; cell-free DNA; microvascular invasion; decision-making; liquid biopsy; radiomics; deep learning

Outcome Measures

Primary Outcomes (2)

• Recurrence-free survival (RFS)

  The time from surgery to the first diagnosis of recurrence or patient death without recurrence

  Between June 2016 and April 2026

• Overall survival (OS)

  The time from surgery to patient death from any cause or the last follow-up

  Between June 2016 and April 2026

Secondary Outcomes (1)

• Local recurrence

  Between June 2016 and April 2026

Study Arms (2)

The MVI low-risk group

We will develop a multimodal model to predict the preoperative risk of MVI in early-stage HCC patients. The model integrates four individual models: a clinicopathological model, an MRI-based radiomics model, an MRI-based deep learning image model, and a cfDNA genomic mutation model. The patient-level predicted probabilities of MVI from these models are calibrated into standardized MVI risk probabilities (ranging from 0.0 to 1.0) using Platt scaling or isotonic regression, and are then transformed to logits before being entered into an L2-regularized logistic stacking model. The fusion coefficients of the final multimodal system will be estimated exclusively from out-of-fold predictions. After all preprocessing, calibration, fusion and threshold-selection procedures have been locked, patients with a predicted MVI probability below the cutoff determined by the maximum Youden index (sensitivity+specificity-1) will be classified as low risk for MVI. Within this low-risk group, we assess t

Procedure: Curative-intent liver resection

The MVI high-risk group

We will develop a multimodal model to predict the preoperative risk of MVI in early-stage HCC patients. The model integrates four individual models: a clinicopathological model, an MRI-based radiomics model, an MRI-based deep learning image model, and a cfDNA genomic alteration model. The patient-level predicted probabilities of MVI from these models are calibrated into standardized MVI risk probabilities (ranging from 0.0 to 1.0) using Platt scaling or isotonic regression, and are then transformed to logits before being entered into an L2-regularized logistic stacking model. The fusion coefficients of the final multimodal system will be estimated exclusively from out-of-fold predictions. After all preprocessing, calibration, fusion and threshold-selection procedures have been locked, patients with a predicted MVI probability at or above the cutoff determined by the maximum Youden index (sensitivity+specificity-1) will be classified as high risk for MVI.. Within this high-risk group,

Procedure: Curative-intent liver resection

Interventions

Curative-intent liver resection PROCEDURE

All patients have undergone curative-intent resection for early-stage HCC. A wide resection margin is defined as ≥1 cm, and a narrow margin as \<1 cm. The prognostic comparison of wide resection margin versus narrow margin will be conducted among patients who were technically eligible for either approach, as determined by clinical guidelines and the assessment of an independent panel of surgical specialists.

The MVI high-risk group; The MVI low-risk group

Eligibility Criteria

• Age: 18 Years - 75 Years
• Sex: all
• Healthy Volunteers: No
• Age Groups: Adult (18-64), Older Adult (65+)
• Sampling Method: Non-Probability Sample

Study Population

China's National Center for Liver Cancer (NCLC)/Eastern Hepatobiliary Surgery Hospital (EHBH) maintains a prospectively collected database and biobank, which includes consecutive cases of surgically resected liver cancer between January 2015 and April 2026. We will identify study candidates from this database. A total of 3,000 eligible patients are planned to be enrolled. Their preoperative MRI data will be utilized to establish both radiomics and deep learning MVI-predictive models. We have also established a prospective cohort of 650 patients from four Chinese tertiary medical centers: Zhongda Hospital (Southeast University), Mengchao Hepatobiliary Hospital (Fujian Medical University), Sun Yat-Sen Memorial Hospital (Sun Yat-Sen University), and Eastern Hepatobiliary Surgery Hospital (EHBH). These patients had undergone HCC resection between January 2016 and June 2019. The paired tumor/non-tumor tissue or cell-free DNA (cfDNA) sequencing data of these patients are available.

Sponsors & Collaborators

• Eastern Hepatobiliary Surgery Hospital: lead

Study Sites (1)

Eastern Hepatobiliary Surgery Hospital, Naval Medical University,

Shanghai, Shanghai Municipality, 021, China

Location

Biospecimen

Retention: SAMPLES WITH DNA

Tissue sample of 150 HCC patients were used for WES/NGS sequencing to detect of a gene profile related to MVI to generate a gene panel for targeted sequencing of cfDNA. cfDNA sample were used for targeted NGS sequencing with the gene panel.

MeSH Terms

Conditions

Carcinoma, Hepatocellular; Liver Neoplasms

Condition Hierarchy (Ancestors)

Adenocarcinoma; Carcinoma; Neoplasms, Glandular and Epithelial; Neoplasms by Histologic Type; Neoplasms; Digestive System Neoplasms; Neoplasms by Site; Digestive System Diseases; Liver Diseases

Study Officials

• Feng Shen, MD, PhD

  Eastern Hepatobiliary Surgery Hospital, Naval Medical University

  PRINCIPAL INVESTIGATOR

Study Design

• Study Type: observational
• Observational Model: COHORT
• Time Perspective: PROSPECTIVE
• Sponsor Type: OTHER
• Responsible Party: PRINCIPAL INVESTIGATOR
• PI Title: Professor and Chief Surgeon

Study Record Dates

• First Submitted: September 12, 2022
• First Posted: September 15, 2022
• Study Start: January 1, 2022
• Primary Completion: April 30, 2026
• Study Completion: April 30, 2026
• Last Updated: September 28, 2026

Record last verified: 2026-09

Locations

CN (1)