A large analysis of 9257 CT and MRI liver scans from 46 studies found that adding ancillary features to the Liver Imaging Reporting and Data System did not improve its diagnostic accuracy for hepatocellular carcinoma.
The findings suggest that radiologists may not gain additional diagnostic value from using ancillary features (AFs) beyond the major features already defined in Liver Imaging Reporting and Data System (LI-RADS).
LI-RADS is a standardized system for interpreting liver imaging in adults at risk for hepatocellular carcinoma (HCC). It categorizes liver lesions based on imaging findings from contrast-enhanced CT or MRI. While major features in LI-RADS have been extensively studied, the value of AFs—optional imaging findings such as fat in mass, restricted diffusion, or T2 hyperintensity—has remained unclear.
Researchers used individual participant data (IPD) from 7811 adults (6792 men, 1019 women; mean age, 58.7 years; standard deviation, 10.7) to compare three classification strategies. The first used only major features. The second included individual AFs. The third allowed AFs that favor malignancy or HCC to upgrade category LR-4 lesions to LR-5.
Lesions were confirmed as HCC, other malignancies, or benign using histopathology or composite imaging reference standards. Of the 9257 observations, 6034 were HCC, 755 were non-HCC malignancies, and 1265 were benign. Most observations were evaluated with MRI (8159), and the remainder with CT (1098).
Across all strategies, there were no significant differences in diagnostic accuracy. The area under the receiver operating characteristic curve (AUC), which measures how well a test distinguishes between disease and no disease, was not affected by the use of AFs. The P value for differences in AUC ranged from .65 to >.99.
There were also no differences in other diagnostic performance measures. The positive predictive value (PPV), sensitivity, and specificity of the LR-5 category for diagnosing HCC remained unchanged across strategies. P values ranged from .11 to >.99.
A secondary analysis limited to 9 low-risk-of-bias studies confirmed these findings. Allowing AFs to upgrade LR-4 lesions to LR-5 also did not improve diagnostic performance.
The proportion of LR-4 lesions upgraded to LR-5 based on AFs varied. Using AFs favoring malignancy, 5.2% to 76.0% of lesions were upgraded. Using AFs favoring HCC, upgrades ranged from 4% to 14.7%. Despite these reclassifications, diagnostic accuracy remained the same.
The findings suggest that LI-RADS major features alone are sufficient for accurate HCC diagnosis. Although AFs are associated with malignancy or benignity, they did not improve diagnostic outcomes when applied individually. The study also noted that AFs may introduce complexity and variability between readers without improving diagnostic utility.
Researchers used a bivariate mixed-effects model and adhered to strict IPD meta-analysis methods to ensure robustness. These results suggest that AFs may be unnecessary for improving LI-RADS diagnostic performance in clinical practice.
Full disclosures can be found in the published study.
Source: Radiology
