A newly developed and externally validated ARTE score predicted hepatic decompensation in patients with unresectable hepatocellular carcinoma (uHCC) receiving first-line atezolizumab plus bevacizumab, stratifying patients into groups with markedly different decompensation-free survival outcomes, according to a study published in Clinical Gastroenterology and Hepatology.
“HCC is one of the most common cancers worldwide and usually occurs in patients with cirrhosis,” explained Mark Russo, MD, medical director of liver transplantation and chief of hepatology at Atrium Health Wake Forest in Charlotte, and professor of medicine at Wake Forest School of Medicine. “When HCC occurs in patients with cirrhosis who have portal hypertension or decompensated disease it is unresectable. Thus, under these circumstances patients are considered for locoregional therapy and/or systemic therapy. However, patients may develop hepatic decompensation with locoregional therapy or systemic therapy.”
Study design and patient population
In a multicenter retrospective analysis of prospectively collected data from the ARTE database, researchers evaluated 453 patients with uHCC and Child-Pugh A cirrhosis who received first-line atezolizumab plus bevacizumab. An external validation cohort included 292 patients from the AB-real database. The primary objective was to identify predictors of hepatic decompensation and develop a clinically applicable risk score.
Decompensation events in the derivation cohort
During a median follow-up of 14 months in the derivation cohort, 74 patients (16.3%) developed hepatic decompensation, defined according to Baveno VII consensus as the development of clinically evident ascites, variceal bleeding or overt hepatic encephalopathy.
Ascites alone was the most frequent decompensation, accounting for 49.3% of decompensations. Among the decompensations, 75% were defined as pure decompensations (no radiological progression within 60 days).
ARTE score components and performance
Multivariable Cox regression identified three independent predictors of hepatic decompensation: neoplastic portal vein thrombosis, bilirubin levels of at least 2 mg/dL and platelet counts of 100 ×10³/mm³ or lower.
These variables were incorporated into the ARTE score, which assigns 1 point for neoplastic portal vein thrombosis, 2 points for bilirubin levels of at least 2 mg/dL, and 1 point for platelet counts of 100 ×10³/mm³ or lower, yielding a total score ranging from 0 to 4 points.
Patients were categorized as low risk (0-1 points; n = 360), intermediate risk (2 points; n = 49), or high risk (3-4 points; n = 44). Compared with the low-risk group, intermediate-risk patients had nearly twice the risk of hepatic decompensation, while high-risk patients had more than four times the risk. The model demonstrated good discrimination, with a Harrell C statistic of 0.7022. The area under the curve or AUC at 3 months was 0.76, at 6 months 0.76, and at 12 months 0.74. The ARTE score demonstrated better discrimination for hepatic decompensation than the albumin-bilirubin score.
Risk stratification and validation
Decompensation-free survival at 12 and 24 months was 87% and 83% for low-risk, 79% and 64% for intermediate-risk, and 57% and 56% for high-risk, respectively. The ARTE score retained prognostic value in the AB-real cohort.
The score remained predictive across multiple sensitivity analyses. In patients with Barcelona Clinic Liver Cancer stage C disease, the ARTE score remained significantly associated with hepatic decompensation. Similar findings were observed when variceal bleeding was excluded from the decompensation definition and among patients with hepatitis C virus infection who had achieved sustained virologic response. The score also independently predicted decompensation-free survival, overall survival, pure decompensation events unrelated to tumor progression and early decompensation occurring within the first 3 months of therapy. For early decompensation, the ARTE score was the only independent predictor identified.
External validation confirmed the score's prognostic value. In the AB-real cohort, 34 patients (11.6%) developed hepatic decompensation during a median follow-up of 10.7 months. The ARTE score remained an independent predictor of decompensation, with a model concordance statistic of 0.6489. Calibration analyses demonstrated close agreement between predicted and observed risks.
Study limitations
The authors noted several limitations, including the retrospective analysis of observational cohorts, potential residual confounding, lack of systematically available longitudinal albumin-bilirubin or Child-Pugh score data, and restriction of both derivation and validation cohorts to Western populations. They also cautioned that the score was developed specifically in patients treated with atezolizumab plus bevacizumab and requires validation in patients receiving other first-line immunotherapy regimens. In addition, the relative contribution of treatment-related toxicity versus disease-related mechanisms to hepatic decompensation may differ across therapeutic strategies, particularly between VEGF-based combinations and regimens that rely solely on immune checkpoint blockade.
“The ARTE score is a simple, biologically sound, and externally validated tool that accurately predicts hepatic decompensation in uHCC patients receiving first-line AB,” the researchers concluded, adding that its implementation in clinical practice can facilitate intensified monitoring, earlier initiation of supportive interventions and informed therapeutic decision-making for high-risk patients.
Dr. Russo also highlighted several considerations when interpreting the findings. “This study provides practical information about the risk of decompensation for our patients with CP-A cirrhosis and HCC. However, the study included patients with CP-A cirrhosis, so results are limited to this patient group and explains the relatively low rate of decompensation (16.3%), with ascites accounting for about half of the events.”
In addition, he said, most of the patients were male and more than half had a viral etiology of cirrhosis, which is not representative of the current etiology of cirrhosis in many areas.
“It would have been informative to present overall survival curves in patients with and without decompensation,” he added. “Nevertheless, the score incorporates readily available variables, platelet count, bilirubin, neoplastic portal vein thrombosis, that can be used at the bedside to inform our patients about the risks associated with A/B.”
Massimo Iavarone received consulting fees from AstraZeneca, EISAI, MSD, Roche and Roche Diagnostics; lecture fees from AstraZeneca, EISAI, Gilead, IPSEN, MSD and Roche; travel fees from Roche; and research funding (institution) from AstraZeneca, EISAI, Gilead and Roche. Numerous coauthors reported advisory, consulting, speaker, travel, or research relationships with pharmaceutical companies. Several authors reported no conflicts of interest. Study funding was provided in part by Ricerca Corrente.