Purpose Atezolizumab-Bevacizumab (AB) is first-line treatment for patients with hepatocellular carcinoma (HCC), and Bevacizumab is associated with increased gastrointestinal bleeding; hence IMBrave150 mandated esophago-gastroduodenoscopy (OGD) before therapy initiation. Universal pre-treatment OGD potentially delays treatment. Non-invasive scores have been developed to risk stratify patients for variceal screening to avoid unnecessary OGD. EVendo Score (ES) was developed with laboratory and clinical data, and demonstrated reliability in identifying patients with cirrhosis with gastroesophageal varices and varices needing treatment (VNT), but is not formally validated in patients with HCC. ES of <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$\le$</EquationSource> <EquationSource Format="MATHML"><math> <mo>≤</mo> </math></EquationSource> </InlineEquation> 3.90 is negative, suggesting low risk for VNT and that OGD can be deferred. We tested ES in a cohort of patients with HCC and underlying cirrhosis receiving AB, to evaluate if patients could be safely spared from OGD. Methods We retrospectively analysed data for 145 patients with HCC from National Cancer Centre (Singapore) and Asan Medical Centre (Korea) treated with AB from June 2018-July 2022 who underwent pre-treatment OGD. Biochemical data and clinical parameters were collected and ES calculated for each patient. Using pre-specified cut-offs, sensitivity, specificity, negative predictive value (NPV) and missed VNT rate were elucidated. Results ES of <InlineEquation ID="IEq2"> <EquationSource Format="TEX">$\le$</EquationSource> <EquationSource Format="MATHML"><math> <mo>≤</mo> </math></EquationSource> </InlineEquation> 3.90 identified patients without VNT with sensitivity of 67% (95% CI 59-76), specificity of 37% (95% CI 16-62), and NPV of 88% (95% CI 80-93). Missed VNT rate remained high (8.3%). Exploratory analysis was performed in our local cohort incorporating identified radiological features in combination with ES – sensitivity and NPV remained comparable, but missed VNT rate was lower (3.1%). Conclusion A low ES has high NPV for absence of VNT, but the NPV in our examined cohort is lower than previous validations, and there remains a significant missed VNT risk. Additional analysis should be performed with ES in conjunction with clinicoradiologic features before we can conclude its reliability as a screening tool. Further evaluation of cut-offs and application of this score in other populations should be explored.

Using the EVendo Score to Predict Varices and Varices Needing Treatment in Hepatocellular Carcinoma Patients Planned for Atezolizumab-Bevacizumab

期刊论文
发表日期：

2026年2月26日

查看全文

Suat Ying Lee,
Chiew Woon Lim,
Sejin Kim,
Umairah Binte Abdul Majeed,
Kennedy Yao Yi Ng,
Jason Pik-Eu Chang,
David Wai-Meng Tai,
Changhoon Yoo,
Joycelyn Jie Xin Lee

摘要

Purpose

Atezolizumab-Bevacizumab (AB) is first-line treatment for patients with hepatocellular carcinoma (HCC), and Bevacizumab is associated with increased gastrointestinal bleeding; hence IMBrave150 mandated esophago-gastroduodenoscopy (OGD) before therapy initiation. Universal pre-treatment OGD potentially delays treatment. Non-invasive scores have been developed to risk stratify patients for variceal screening to avoid unnecessary OGD. EVendo Score (ES) was developed with laboratory and clinical data, and demonstrated reliability in identifying patients with cirrhosis with gastroesophageal varices and varices needing treatment (VNT), but is not formally validated in patients with HCC. ES of $\le$ $\leq$ 3.90 is negative, suggesting low risk for VNT and that OGD can be deferred. We tested ES in a cohort of patients with HCC and underlying cirrhosis receiving AB, to evaluate if patients could be safely spared from OGD.

Methods

We retrospectively analysed data for 145 patients with HCC from National Cancer Centre (Singapore) and Asan Medical Centre (Korea) treated with AB from June 2018-July 2022 who underwent pre-treatment OGD. Biochemical data and clinical parameters were collected and ES calculated for each patient. Using pre-specified cut-offs, sensitivity, specificity, negative predictive value (NPV) and missed VNT rate were elucidated.

Results

ES of $\le$ $\leq$ 3.90 identified patients without VNT with sensitivity of 67% (95% CI 59-76), specificity of 37% (95% CI 16-62), and NPV of 88% (95% CI 80-93). Missed VNT rate remained high (8.3%). Exploratory analysis was performed in our local cohort incorporating identified radiological features in combination with ES – sensitivity and NPV remained comparable, but missed VNT rate was lower (3.1%).

Conclusion

A low ES has high NPV for absence of VNT, but the NPV in our examined cohort is lower than previous validations, and there remains a significant missed VNT risk. Additional analysis should be performed with ES in conjunction with clinicoradiologic features before we can conclude its reliability as a screening tool. Further evaluation of cut-offs and application of this score in other populations should be explored.