Noninvasive model minimizes the risk for missing high-risk varices in unresectable HCC
Click Here to Manage Email Alerts
Key takeaways:
- The imaging-based model had a negative predictive value of 97.4%.
- Using the model, upper endoscopy could be avoided in 50 of 100 low-risk patients, while eight of 100 high-risk patients would be misclassified.
A model using both clinical and imaging data predicted the absence of high-risk varices in patients with unresectable hepatocellular carcinoma and may help avoid unnecessary upper endoscopy in low-risk patients, research showed.
“There are several first-line systemic therapies for patients with HCC, including atezolizumab and bevacizumab,” Neehar D. Parikh, MD, a transplant hepatologist and associate professor at Michigan Medicine, told Healio. “One of the logistical hurdles behind starting the atezolizumab-bevacizumab regimen is the need for variceal risk-stratification prior to therapy initiation. This is due to prior data showing some patients can have variceal bleeding related to bevacizumab therapy due to its VEGF effects.”
Parikh continued: “Obtaining an [esophagogastroduodenoscopy (EGD)] can be sometimes difficult to complete in an expeditious way and can delay initiation of therapy. Our team wanted to determine if we could develop a model so that low-risk patients could safely avoid EGD prior to initiating bevacizumab.”
In a retrospective cohort study, Parikh and colleagues included 707 adult patients (mean age, 64.6 years; 80.6% men; 74% white) with unresectable HCC and Child Pugh A5 to B7 cirrhosis from 21 sites in the North American Liver Cancer Consortium between 2007 and 2019. Eligibility requirements included completion of an EGD within 12 months of index imaging but before HCC treatment. At baseline, 25% of patients had high-risk varices.
The researchers randomly assigned patients 70:30 into training and validation sets, and selected the final model with a focus on maximizing negative predictive value and minimizing false negatives.
According to results published in Clinical Gastroenterology and Hepatology, a model using only clinical variables achieved a negative predictive value of 86.3%, a false negative rate of 26.3% and an area under curve of 0.687, whereas a model using both clinical and imaging variables achieved values of 97.4%, 7.7% and 0.694, respectively.
“We then validated the model in a subset of our cohort and the model showed favorable characteristics in categorizing patients as low-risk,” Parikh said.
By integrating imaging data into the model, EGD could be avoided in 50 of 100 low-risk patients, while just eight of 100 high-risk patients would be misclassified as low-risk, results showed.
“So, in a cohort of 100 patients with a population prevalence of 25% for high-risk varices, 38 EGDs would be avoided and two high-risk patients would be misclassified as low-risk,” Parikh noted.
“While the model has very promising results and we show that labs and imaging can accurately classify patient risk of having high-risk varices, we require further research to further validate the model for clinical use,” Parikh told Healio. “This is due to the retrospective nature of the data included and some of the biases we may see with that.”
He added, “We are currently developing a prospective validation study for this model to determine if it can be useful in clinical practice.”
Collapse
Click Here to Manage Email Alerts