A blood test found twice as many liver cancers as surveillance ultrasound, and all the small ones

Design

Cross-sectional, prospective, blinded, multicentre validation study with multiphasic MRI as reference standard (NCT03694600)

Population

1,268 evaluable adults with cirrhosis at high risk of hepatocellular carcinoma; 46 (3.6%) had cancer, 46% of lesions 2 cm or smaller

Primary outcome

Sensitivity and specificity of the multianalyte cell-free DNA blood test versus abdominal ultrasound for detecting hepatocellular carcinoma

Effect

Sensitivity 47.8% (95% CI 32.9-63.1) vs 28.3% (16.0-43.5); lesions 2 cm or less 28.6% (11.3-52.2) vs 0% (0.0-16.1); specificity 87.6% vs 93.9%

Six-monthly ultrasound is the surveillance standard for cirrhosis and it performs poorly, particularly for small nodules and particularly when patients do not attend. This blinded, prospective, multicentre validation study, published in May 2026, tested the multianalyte cell-free DNA blood test HelioLiver Dx against ultrasound in 1,268 adults with cirrhosis, using multiphasic MRI in everyone as the reference standard.

46 participants (3.6%) had hepatocellular carcinoma on MRI, and nearly half of those lesions were 2 cm or smaller. The blood test detected 47.8% of all cancers (95% CI 32.9-63.1) against 28.3% for ultrasound (16.0-43.5). For lesions of 2 cm or less the gap was starker: 28.6% for the blood test, and ultrasound found none at all (0%, 95% CI 0.0-16.1). Specificity went the other way - 87.6% for the blood test against 93.9% for ultrasound - so the test buys sensitivity with false positives, each of which becomes an MRI and a worried patient.

The honest headline is not that the blood test is good. It detects fewer than half of cancers, which is not a test you would accept in isolation. The finding is that ultrasound is worse than most clinicians assume, and specifically that it is close to useless for the small lesions that surveillance exists to catch.

So the immediate change is to how ultrasound is reported and discussed, not to what you order. Stop treating a clean surveillance ultrasound as a clean bill of health, particularly in an obese patient or one with a nodular liver where the scan was described as technically difficult - that is the patient for cross-sectional imaging. HelioLiver Dx is a US test; its availability and regulatory position in India are not addressed by this study and should not be assumed.

• Read 'technically difficult' on a surveillance ultrasound report as a failed test, not a normal one
• Have a low threshold for MRI or multiphasic CT in obese patients and in coarse, nodular livers
• Keep six-monthly intervals - none of this supports lengthening them
• Do not present a negative blood-based test as exclusion of cancer: over half were missed
• Alpha-fetoprotein alone remains inadequate; it was outperformed here too

The statistics, in plain English

Sensitivity of 47.8% means the test missed slightly more than half the cancers present; the confidence interval, 32.9 to 63.1, is wide because only 46 patients had cancer, so the true figure could be a third or nearly two-thirds. Ultrasound's 0% for small lesions has an upper confidence limit of 16.1%, so 'found none' is honest for this cohort but does not mean it never finds any. The specificity difference, 87.6% versus 93.9%, sounds small and is not: across 1,222 people without cancer it is roughly 75 extra false alarms, each generating an MRI.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.