Diagnosing hepatocellular carcinoma (HCC) accurately is vital for patient survival, especially in high-risk populations in countries like India. Recently, researchers evaluated probability-based HCC categories using LI-RADS v2018 and KLCA-NCC v2022 criteria. Specifically, a landmark retrospective study analyzed how these different systems classify liver lesions on MRI and CT scans. Ultimately, this understanding helps clinical teams determine optimal treatment strategies.
Analyzing Probability-Based HCC Categories
This multicenter study evaluated 2,237 treatment-naïve patients at risk for HCC between January 2015 and June 2018. Specifically, these patients had 2,445 lesions, which four independent radiologists analyzed using both diagnostic criteria. Consequently, researchers compared the positive predictive value (PPV) and lesion distribution across the systems. The investigators used pathology or clinical follow-up as reference standards. Additionally, they employed generalized estimating equations to account for lesion clustering. Ultimately, this robust analysis revealed significant differences in performance on both MRI and CT.
Key Findings from MRI Evaluation
On MRI scans, the LI-RADS LR-5 category showed a higher PPV than the KLCA-NCC definite HCC category. Specifically, the PPV was 93.9% for LR-5 compared to 91.7% for the definite HCC category. However, the PPV was not different when comparing the intermediate categories. For instance, LR-4 and KLCA-NCC probable HCC categories showed statistically similar PPVs of 68.2% and 64.1%, respectively. In contrast, the proportion of actual HCCs differed significantly in the indeterminate groups. Indeed, the HCC proportion was 13.2% in LR-3, but it reached 40.8% in KLCA-NCC indeterminate nodules. Therefore, clinicians should exercise caution when comparing these specific categories directly.
Important CT Results and Clinical Takeaways
On CT scans, the results presented a slightly different diagnostic picture. For example, in 1,455 patients with 1,590 CT lesions, LR-5 and definite HCC showed an identical PPV of 92.9%. However, the PPV of the KLCA-NCC probable HCC category was significantly higher than the LR-4 category. Specifically, the probable HCC category achieved a PPV of 83.8%, whereas LR-4 achieved only 73.9%. Furthermore, the HCC proportion was 52.0% in LR-3 compared to 46.2% in the KLCA-NCC indeterminate category. Consequently, this indicates that the corresponding probability categories are not fully interchangeable across modalities. Ultimately, both guidelines provide effective risk stratification, but they require careful, context-specific application.
Frequently Asked Questions
Q1: Why did LR-5 show a higher PPV than KLCA-NCC definite HCC on MRI?
On MRI, the LI-RADS LR-5 criteria are slightly more conservative and specific than the KLCA-NCC definite HCC criteria, which successfully reduces false positives and elevates the positive predictive value.
Q2: Are the indeterminate categories of LI-RADS and KLCA-NCC interchangeable?
No, they are not interchangeable. For instance, on MRI, the proportion of actual HCCs was 13.2% in the LR-3 category but reached 40.8% in the KLCA-NCC indeterminate nodule category, indicating a substantial difference in risk level.
Q3: How do the systems perform differently on CT imaging?
On CT scans, both systems show an identical PPV of 92.9% for definite categories. However, the KLCA-NCC probable HCC category yields a significantly higher PPV (83.8%) compared to LI-RADS LR-4 (73.9%).
References
- Yoon JH et al. LI-RADS v2018 versus KLCA-NCC v2022: comparison of probability-based HCC categories. Eur Radiol. 2026 Jun 26. doi: 10.1007/s00330-026-12696-0. PMID: 42360478.
- Joo I et al. 2022 Korean Liver Cancer Association-National Cancer Center Korea Practice Guidelines for Imaging Diagnosis of Hepatocellular Carcinoma: What’s New? Korean J Radiol 2023;24(1):1-5. doi: 10.3349/kjr.2022.0837.
- Yoon JH et al. Comparison of Four Diagnostic Guidelines for Hepatocellular Carcinoma Using Gadoxetic Acid-enhanced Liver MRI. Radiology. 2024 Apr;311(1):e231580. doi: 10.1148/radiol.231580.
