Performance of Liver Imaging Reporting and Data System (LI-RADS) nonradiation treatment response algorithm version 2024 on magnetic resonance imaging for transarterial chemoembolization plus systemic therapy in hepatocellular carcinoma
Introduction
The Liver Imaging Reporting and Data System treatment response algorithm (LR-TRA) was developed to standardize the assessment and reporting of treatment response of hepatic observations, hepatocellular carcinoma (HCC) in particular, to locoregional therapy (LRT) (1). The algorithm was recently updated to version 2024 (v2024), and includes three categories for lesions treated with nonradiation-based modalities (LR-TR Nonviable, LR-TR Equivocal and LR-TR Viable), assigned on per-lesion bases (1).
Recent years have witnessed a shift in the treatment paradigm of HCC toward a combination of LRT and systemic therapy (2-4). For example, the phase III randomized clinical trial EMERALD-1 has demonstrated survival benefits of transarterial chemoembolization (TACE) combined with durvalumab and bevacizumab over TACE alone in patients with unresectable HCC (2). Similarly, the phase III randomized clinical trial LAUCH showed improved overall survival of lenvatinib combined with TACE over lenvatinib monotherapy in patients with advanced HCC (4). While per-patient treatment response assessment systems [e.g., the Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 and mRECIST] are widely used for LRT and systemic combination therapy in clinical trials, they do not provide granular per-lesion information regarding tumor viability for individual HCCs in the way that LR-TRA does (1,5,6). However, assessing the treatment response on a per-lesion level is clinically relevant because subsequent resection or locoregional treatment (e.g., ablation, radiation-based LRT) could be performed for the residual viable tumor.
Histopathological examination is currently the gold standard for assessing residual viable tumor. However, this information is available only after surgery rather than at the time of treatment decision-making. Fortunately, LR-TRA, particularly when applied to magnetic resonance imaging (MRI) (owing to the ability to obviate lipiodol effects and offer multiple imaging sequences) has shown satisfactory diagnostic accuracy (60–86%) for detecting residual viable tumor following LRT (7-9). However, the performances of MRI-based LR-TRA in assessing treatment response to LRT and systemic combination therapy remain to be elucidated (1), particularly given the potential confounding factors introduced by systemic agents’ biological effects on tumor microenvironment.
Therefore, this study aimed to investigate the diagnostic accuracy of LR-TRA computed tomography (CT)/MRI Nonradiation v2024 on MRI in detecting residual viable tumor in patients with HCC undergoing TACE plus systemic therapy, using postoperative histopathology as the gold standard. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1308/rc).
Methods
This single-center, retrospective, cohort study received approval from the Institutional Review Board of West China Hospital {approval No. [2022] 1993}. Informed consent requirements were waived due to the retrospective nature of the study. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.
Study population
Between July 2019 and November 2023, consecutive patients who fulfilled the following inclusion criteria were identified in an electronic database of tertiary-care referral hospital: (I) age ≥18 years; (II) HCC diagnosed either based on postoperative histopathologic examination or pretreatment imaging according to the Liver Imaging Reporting and Data System (LI-RADS) v2018 [in cases of pathologic complete response (pCR)]; (III) received downstaging TACE plus systemic therapy and subsequent curative-intent surgical resection; (IV) underwent contrast-enhanced MRI examination after combination therapy and within 2 months before surgical resection; (V) had complete documentation of residual tumor proportion on pathology reports. Patients were excluded if the MRIs were of insufficient quality (e.g., severe motion artifact). Patient inclusion and exclusion are illustrated in Figure 1.
Whether and when to perform surgical resection were discussed at the multidisciplinary tumor boards while adhering to patient preferences. Generally, curative-intent surgical resection would be considered when all of the following were met: (I) treatment response to downstaging combination therapy was complete response, partial response, or stable disease (with reduced tumor size) according to both RECIST 1.1 and mRECIST for at least 2 months (5,6); (II) adequate hematologic and organ function; (III) absence of severe or persistent (i.e., >2 weeks) adverse events from downstaging therapy as per the Common Terminology Criteria for Adverse Events version 5.0; (IV) absence of serious comorbidities; (V) absence of clinically significant portal hypertension [i.e., splenomegaly (longest diameter on axial images >12 cm) coupled with platelet count <100×103/µL]; and (VI) R0 resection anticipated achievable with sufficient remnant liver volume (≥40% for cirrhotic patients or ≥30% for non-cirrhotic patients).
Clinical (e.g., patient demographics, underlying etiology of chronic liver diseases) and key laboratory results (e.g., serum alpha-fetoprotein) within 2 weeks before surgical resection were collected.
MRI techniques
MRI examinations were conducted using several 3.0-T or 1.5-T MR systems. Either extracellular or hepatobiliary contrast agent was used based on multidisciplinary tumor board discussions or at the discretion of the radiologists. MRI acquisition parameters and vendors are described in Table S1.
Image analysis
A study coordinator with five years of experience in liver MRI who did not participate in image analysis annotated the size and location of the treated lesions on the contrast-enhanced MRIs acquired after downstaging combination therapy and within two months before surgical resection (i.e., the index MRIs) with reference to the postoperative histopathologic reports to guarantee rigid radiology-pathology correlation. Thereafter, three abdominal radiologists (L.S., H.J., and Y.Q., with 3, 7, and 10 years of experience in liver MRI, respectively) independently reviewed the index entire MRI datasets. These readers were aware that all patients had HCC but were blinded to the remaining clinical and pathological information. Disagreement for imaging features was resolved using the majority interpretation.
On a per-lesion basis, the following imaging features were evaluated: (I) tumor size and number; (II) LR-TRA CT/MRI Nonradiation v2024 features, including masslike enhancement in general [i.e., enhancing area (any degree, any phase) that occupies space], masslike arterial phase hyperenhancement, masslike washout, diffusion restriction, and mild-moderate T2 hyperintensity; and (III) LR-TRA CT/MRI Nonradiation v2024 categories (i.e., nonviable, equivocal, and viable) (1). Notably, despite not being included in LR-TRA CT/MRI Nonradiation v2024, masslike arterial phase hyperenhancement and washout were still evaluated because they were the key features for tumor viability in LR-TRA CT/MRI v2018 and that using masslike enhancement (in general) to replace these two features represented one of the most important changes in LR-TRA CT/MRI Nonradiation v2024.
For the patient-level analyses, an imaging feature would be considered present if shown for any of the treated lesions; otherwise, absent. Imaging feature definitions are detailed in Figures S1-S6. When investigating the performances of LR-TR categories for the diagnosis of residual tumor, the LR-TR Equivocal category was grouped into the LR-TR Viable category (9).
Histopathologic examination
Residual viable tumor proportion retrieved from clinical histopathologic reports was used as the reference standard to evaluate treatment response. According to West China Hospital protocol, all histopathologic analyses were conducted by two liver pathologists in consensus, who had access to all clinical and imaging information.
To ensure reliable histopathologic assessment of residual tumors, standardized multiple sampling was applied for all resected tumors. Specifically, complete sampling was performed for small tumors (i.e., ≤3 cm). For large tumors >3 cm, complete sampling was performed for the section at the maximum tumor diameter, and selective sampling for suspicious viable tumor area was performed for the remaining sections at the pathologists’ discretions. Thereafter, the residual viable tumor proportion was calculated as the average of all analyzed tumor samples. For the patient-level analyses, the residual viable tumor proportion was determined as the maximum of all analyzed tumors in patients with multiple tumors.
The residual viable tumor proportion was further dichotomized as pCR (defined as the absence of any residual viable tumor) and major pathologic response (MPR; defined as residual viable tumor proportion ≤10%). Both pCR and MPR were evaluated as endpoints in this analysis. While pCR has been widely evaluated for assessing treatment response (7-11), accumulating evidence suggests that MPR may provide equivalent or potentially superior prognostic significance (12-14).
Statistical analysis
Assessment of inter-reader agreement
Inter-reader agreement was evaluated with the intraclass correlation coefficient for continuous variables (i.e., tumor size), the Fleiss κ value for binary features (i.e., LR-TRA features), and the weighted Fleiss κ value for categorical variables (i.e., tumor number and LR-TR category) or for highly skewed binary imaging features (i.e., those with frequency >90.0 % or <10.0 %) against the paradoxes of kappa. Agreement was categorized as slight (intraclass correlation coefficient or κ, 0–0.20), fair (intraclass correlation coefficient or κ, 0.21–0.40), moderate (intraclass correlation coefficient or κ, 0.41–0.60), substantial (intraclass correlation coefficient or κ, 0.61–0.80), or almost perfect (intraclass correlation coefficient or κ, 0.81–1.00) (15).
Diagnostic test
The diagnostic performances of individual imaging features as well as the dichotomized LR-TR category for evaluating treatment response were assessed based on per-reader and consensus interpretations. Sensitivities, specificities, positive predicting values (PPVs), negative predicting values (NPVs), and accuracies were calculated for detecting any (i.e., non-pCR) as well as >10% residual viable tumor (non-MPR), respectively.
The statistical analyses were performed using MedCalc (version 19.0.7; MedCalc Software) and open-source R software (version 4.1.1; R Foundation for Statistical Computing). A two-tailed P<0.05 was considered statistically significant.
Results
Patients
Patient characteristics are summarized in Table 1. The study included 51 patients [median age, 56 years; interquartile range (IQR), 50.3–63.8 years; 45 males] with 63 HCCs (median size, 5.5 cm; IQR, 2.2–10.0 cm). Chronic hepatitis B viral infection was the predominant etiology of chronic liver diseases (96.1%, 49/51), and cirrhosis was present in 23 (45.1%) of 51 patients. Systemic therapy regimens included antiangiogenic agents in 50 (98%) of 51 patients, and immune checkpoint inhibitors in 37 (72.5%) of 51 patients. Treatment details are summarized in Table S2.
Table 1
| Characteristics | Values |
|---|---|
| Patient level (n=51) | |
| Age (years) | 56 [50.3–63.8] |
| Male | 45 (88.2) |
| Cirrhosis | 23 (45.1) |
| Etiology of chronic liver disease | |
| Chronic hepatitis B† | 44 (86.3) |
| Chronic hepatitis C† | 2 (3.9) |
| Unknown etiology† | 5 (9.8) |
| Systemic therapy | |
| Antiangiogenic agents | 50 (98.0) |
| Immune checkpoint inhibitors | 37 (72.5) |
| AFP (ng/mL) | 11 [3.3–163.75] |
| PIVKA-II (mAU/mL) | 91 [29.3–468.8] |
| AST (IU/L) | 33 [23.5–44.5] |
| ALT (IU/L) | 29 [20–44.5] |
| TBIL (μmol/L) | 12.9±5.6 |
| PT (s) | 11.4±0.9 |
| INR | 1.04±0.08 |
| PLT (109/L) | 122 [90.3–198.5] |
| Interval between initial TACE and MR (days) | 95 [51–181] |
| Prior curative-intent surgical resection | 6 (11.8) |
| Lesion level (n=63) | |
| Tumor size (cm) | 5.5 [2.2–10.0] |
| Tumor number | |
| Solitary | 40 (78.4) |
| 2–3 | 11 (21.6) |
| Over 3 | 0 (0.0) |
| Presence of masslike arterial phase hyperenhancement | 31 (49.2) |
| Presence of masslike washout | 21 (33.3) |
| Presence of masslike enhancement | 31 (49.2) |
| Presence of diffusion restriction | 58 (92.1) |
| Presence of mild-moderate T2 hyperintensity | 62 (98.4) |
| LR-TR Viable category | 31 (49.2) |
| pCR | 15 (23.8) |
| MPR | 33 (52.4) |
Data are presented as median [interquartile range], n (%) or mean ± standard deviation. AFP, alpha-fetoprotein; ALT, alanine aminotransferase; AST, aspartate aminotransferase; INR, international normalized ratio; LR-TR, Liver Imaging Reporting and Data System treatment response; MPR, major pathologic response; MR, magnetic resonance; pCR, pathologic complete response; PIVKA-II, protein induced by vitamin K absence or antagonist-II; PLT, platelet count; PT, prothrombin time; TACE, transarterial chemoembolization; TBIL, total bilirubin.
On pathology, 23.8% (15/63) and 52.4% (33/63) of the treated lesions were categorized as pCR and MPR, respectively.
Diagnostic performances
The diagnostic performances of LR-TRA v2024 features and categories based on consensus interpretations are summarized in Table 2, while those based on per-reader interpretations in Table S3. Representative cases demonstrating LR-TR Viable and LR-TR Nonviable are shown in Figures 2-4.
Table 2
| Consensus | Masslike | LR-TR | ||||
|---|---|---|---|---|---|---|
| Arterial phase hyperenhancement† | Washout† | Enhancement† | Nonviable category‡ | Viable category† | ||
| Any/absence of any residual viable tumor (non-pCR/pCR) | ||||||
| Per-lesion (n=63) | ||||||
| Sensitivity | 31/48 (64.6) [49.5, 77.8] | 21/48 (43.8) [29.5, 58.8] | 31/48 (64.6) [49.5, 77.8] | 15/15 (100.0) [78.2, 100] | 31/48 (64.6) [49.5, 77.8] | |
| Specificity | 15/15 (100.0) [78.2, 100] | 15/15 (100.0) [78.2, 100] | 15/15 (100.0) [78.2, 100] | 31/48 (64.6) [49.5, 77.8] | 15/15 (100.0) [78.2, 100] | |
| PPV | 31/31 (100.0) [NA, NA] | 21/21 (100.0) [NA, NA] | 31/31 (100.0) [NA, NA] | 15/32 (46.9) [37.6, 56.4] | 31/31 (100.0) [NA, NA] | |
| NPV | 15/32 (46.9) [37.6, 56.4] | 15/42 (35.7) [30.2, 41.6] | 15/32 (46.9) [37.6, 56.4] | 31/31 (100.0) [NA, NA] | 15/32 (46.9) [37.6, 56.4] | |
| Accuracy | 46/63 (73.0) [60.3, 83.4] | 36/63 (57.1) [44.0, 69.5] | 46/63 (73.0) [60.3, 83.4] | 46/63 (73.0) [60.3, 83.4] | 46/63 (73.0) [60.3, 83.4] | |
| Per-patient (n=51) | ||||||
| Sensitivity | 29/41 (70.7) [54.5, 83.9] | 19/41 (46.3) [30.7, 62.6] | 29/41 (70.7) [54.5, 83.9] | 10/10 (100.0) [69.2, 100] | 29/41 (70.7) [54.5, 83.9] | |
| Specificity | 10/10 (100.0) [69.2, 100] | 10/10 (100.0) [69.2, 100] | 10/10 (100.0) [69.2, 100] | 29/41 (70.7) [54.5, 83.9] | 10/10 (100.0) [69.2, 100] | |
| PPV | 29/29 (100.0) [NA, NA] | 19/19 (100.0) [NA, NA] | 29/29 (100.0) [NA, NA] | 10/22 (45.5) [34.1, 57.3] | 29/29 (100.0) [NA, NA] | |
| NPV | 10/22 (45.5) [34.1, 57.3] | 10/32 (31.3) [25.5, 37.7] | 10/22 (45.5) [34.1, 57.3] | 29/29 (100.0) [NA, NA] | 10/22 (45.5) [34.1, 57.3] | |
| Accuracy | 39/51 (76.5) [62.5, 87.2] | 29/51 (56.9) [42.2, 70.7] | 39/51 (76.5) [62.5, 87.2] | 39/51 (76.5) [62.5, 87.2] | 39/51 (76.5) [62.5, 87.2] | |
| >10%/≤10% residual viable tumor (non-MPR/MPR) | ||||||
| Per-lesion (n=63) | ||||||
| Sensitivity | 25/30 (83.3) [65.3, 94.4] | 19/30 (63.3) [43.9, 80.1] | 25/30 (83.3) [65.3, 94.4] | 27/33 (81.8) [64.5, 93.0] | 25/30 (83.3) [65.3, 94.4] | |
| Specificity | 27/33 (81.8) [64.5, 93.0] | 31/33 (93.9) [79.8, 99.3] | 27/33 (81.8) [64.5, 93.0] | 25/30 (83.3) [65.3, 94.4] | 27/33 (81.8) [64.5, 93.0] | |
| PPV | 25/31 (80.6) [66.5, 89.7] | 19/21 (90.5) [70.7, 97.4] | 25/31 (80.6) [66.5, 89.7] | 27/32 (84.4) [70.5, 92.4] | 25/31 (80.6) [66.5, 89.7] | |
| NPV | 27/32 (84.4) [70.5, 92.4] | 31/42 (73.8) [63.6, 82.0] | 27/32 (84.4) [70.5, 92.4] | 25/31 (80.6) [66.5, 89.7] | 27/32 (84.4) [70.5, 92.4] | |
| Accuracy | 52/63 (82.5) [70.9, 90.9] | 50/63 (79.4) [67.3, 88.5] | 52/63 (82.5) [70.9, 90.9] | 52/63 (82.5) [70.9, 90.9] | 52/63 (82.5) [70.9, 90.9] | |
| Per-patient (n=51) | ||||||
| Sensitivity | 24/27 (88.9) [70.8, 97.6] | 17/27 (63.0) [42.4, 80.6] | 24/27 (88.9) [70.8, 97.6] | 19/24 (79.2) [57.8, 92.9] | 24/27 (88.9) [70.8, 97.6] | |
| Specificity | 19/24 (79.2) [57.8, 92.9] | 22/24 (91.7) [73.0, 99.0] | 19/24 (79.2) [57.8, 92.9] | 24/27 (88.9) [70.8, 97.6] | 19/24 (79.2) [57.8, 92.9] | |
| PPV | 24/29 (82.8) [68.5, 91.4] | 17/19 (89.5) [68.6, 97.1] | 24/29 (82.8) [68.5, 91.4] | 19/22 (86.4) [68.1, 94.9] | 24/29 (82.8) [68.5, 91.4] | |
| NPV | 19/22 (86.4) [68.1, 94.9] | 22/32 (68.8) [57.0, 78.5] | 19/22 (86.4) [68.1, 94.9] | 24/29 (82.8) [68.5, 91.4] | 19/22 (86.4) [68.1, 94.9] | |
| Accuracy | 43/51 (84.3) [71.4, 93.0] | 39/51 (76.5) [62.5, 87.2] | 43/51 (84.3) [71.4, 93.0] | 43/51 (84.3) [71.4, 93.0] | 43/51 (84.3) [71.4, 93.0] | |
Data are numbers of lesions and patients, with diagnostic values as percentages in parentheses and 95% confidence intervals in square brackets. †, the diagnostic performances of characteristics were calculated for the detection of any residual viable tumor and >10% residual viable tumor; ‡, the diagnostic performances of characteristic were calculated for the detection of absence of any residual viable tumor and ≤10% residual viable tumor. LR-TR, Liver Imaging Reporting and Data System treatment response; MPR, major pathologic response; NA, not available; NPV, negative predicting value; pCR, pathologic complete response; PPV, positive predicting value.
LR-TRA v2024 imaging features
Among the 63 treated HCCs, masslike arterial phase hyperenhancement, masslike washout, masslike enhancement in general, diffusion restriction, and mild-moderate T2 hyperintensity were observed for 31 (49.2%), 21 (33.3%), 31 (49.2%), 58 (92.1%), and 62 (98.4%) lesions, respectively.
For the detection of any residual viable tumor (i.e., non-pCR), the per-lesion sensitivity, specificity, PPV, NPV and accuracy of masslike enhancement were 64.6% [95% confidence interval (CI): 49.5–77.8%], 100% (95% CI: 78.2–100%), 100% (95% CI: cannot be calculated), 46.9% (95% CI: 37.6–56.4%), and 73% (95% CI: 60.3–83.4%), respectively. The per-patient sensitivity, specificity, PPV, NPV and accuracy were 70.7% (95% CI: 54.5–83.9%), 100% (95% CI: 69.2–100%), 100% (95% CI: cannot be calculated), 45.5% (95% CI: 34.1–57.3%), and 76.5% (95% CI: 62.5–87.2%), respectively.
For the detection of >10% residual viable tumor (non-MPR), the per-lesion sensitivity, specificity, PPV, NPV and accuracy of masslike enhancement were 83.3% (95% CI: 65.3–94.4%), 81.8% (95% CI: 64.5–93%), 80.6% (95% CI: 66.5–89.7%), 84.4% (95% CI: 70.5–92.4%), and 82.5% (95% CI: 70.9–90.9%), respectively. The per-patient sensitivity, specificity, PPV, NPV and accuracy of masslike enhancement were 88.9% (95% CI: 70.8–97.6%), 79.2% (95% CI: 57.8–92.9%), 82.8% (95% CI: 68.5–91.4%), 86.4% (95% CI: 68.1–94.9%), and 84.3% (95% CI: 71.4–93%), respectively.
LR-TRA v2024 category
Among the 63 treated HCCs, 31 (49.2%), 0 (0%), and 32 (50.8%) were categorized based on the consensus interpretations as LR-TR Viable, Equivocal, and Nonviable, respectively. After applying the ancillary features (i.e., diffusion restriction and mild-moderate T2 hyperintensity), no lesion rated as LR-TR Equivocal was upgraded to TR-TR Viable. Therefore, the diagnostic measures of the LR-TR Viable category were the same as those of “masslike enhancement” as described above.
Inter-reader agreement
The per-reader and consensus interpretations, as well as inter-reader agreements of all evaluated features are summarized in Table 3.
Table 3
| Imaging feature | Frequency | The inter-reader agreement | |||||
|---|---|---|---|---|---|---|---|
| Reader 1 | Reader 2 | Reader 3 | Consensus | ICC or κ (95% CI) | Agreement | ||
| Categorical variable | |||||||
| Mild-moderate T2 hyperintensity†, n/N (%) | 0.79 (0.68, 0.86) | Substantial | |||||
| Presence | 58/63 (92.1) | 60/63 (95.2) | 60/63 (95.2) | 62/63 (98.4) | |||
| Absence | 5/63 (7.9) | 3/63 (4.8) | 3/63 (4.8) | 1/63 (1.6) | |||
| Diffusion restriction†, n/N (%) | 0.66 (0.55, 0.79) | Substantial | |||||
| Presence | 57/63 (90.5) | 56/63 (88.9) | 49/63 (77.8) | 58/63 (92.1) | |||
| Absence | 6/63 (9.5) | 7/63 (11.1) | 14/63 (22.2) | 5/63 (7.9) | |||
| Tumor number†, n/N (%) | 1.00 (NA, NA) | Almost perfect | |||||
| Solitary | 40/63 (78.4) | 40/63 (78.4) | 40/63 (78.4) | 40/63 (78.4) | |||
| 2–3 | 11/63 (21.6) | 11/63 (21.6) | 11/63 (21.6) | 11/63 (21.6) | |||
| Over 3 | 0/63 (0.0) | 0/63 (0.0) | 0/63 (0.0) | 0/63 (0.0) | |||
| Treatment-specific expected enhancement†, n/N (%) | 0.05 (−0.11, 0.19) | Slight | |||||
| Presence | 44/63 (69.8) | 17/63 (27.0) | 48/63 (76.2) | 41/63 (65.1) | |||
| Absence | 19/63 (30.2) | 46/63 (73.0) | 15/63 (23.8) | 22/63 (34.9) | |||
| Masslike arterial phase hyperenhancement†, n/N (%) | 0.44 (0.27, 0.59) | Moderate | |||||
| Presence | 44/63 (69.8) | 29/63 (46.0) | 34/63 (54.0) | 31/63 (49.2) | |||
| Absence | 19/63 (30.2) | 34/63 (54.0) | 29/63 (46.0) | 32/63 (50.8) | |||
| Masslike washout†, n/N (%) | 0.46 (0.27, 0.62) | Moderate | |||||
| Presence | 26/63 (41.3) | 18/63 (28.6) | 22/63 (34.9) | 21/63 (33.3) | |||
| Absence | 37/63 (58.7) | 45/63 (71.4) | 41/63 (65.1) | 42/63 (66.7) | |||
| Masslike enhancement†, n/N (%) | 0.44 (0.27, 0.59) | Moderate | |||||
| Presence | 44/63 (69.8) | 29/63 (46.0) | 34/63 (54.0) | 31/63 (49.2) | |||
| Absence | 19/63 (30.2) | 34/63 (54.0) | 29/63 (46.0) | 32/63 (50.8) | |||
| LR-TR category†, n/N (%) | 0.39 (0.25, 0.54) | Fair | |||||
| Nonviable | 19/63 (30.2) | 34/63 (54.0) | 29/63 (46.0) | 32/63 (50.8) | |||
| Equivocal | 0/63 (0.0) | 0/63 (0.0) | 3/63 (4.8.0) | 0/63 (0.0) | |||
| Viable | 44/63 (69.8) | 29/63 (46.0) | 31/63 (49.2) | 31/63 (49.2) | |||
| Continuous variable | |||||||
| Tumor size (cm)‡, median [IQR]/mean ± SD | 5.6 [2.0, 10.4] | 5.4 [2.0, 9.3] | 6.4±4.2 | 5.5 [2.2, 10.0] | 0.98 (0.98, 0.99) | Almost perfect | |
| Viable or equivocally viable disease size (cm)‡, median [interquartile range] | 1.8 [0, 4.0] | 0 [0, 2.1] | 1.1 [0, 3.1] | 0 [0, 3] | 0.87 (0.78, 0.92) | Almost perfect | |
†, the inter-reader agreement between the three readers was conducted by computing the Fleiss κ values as well as weighted Fleiss κ values; ‡, the inter-reader agreement between the three readers was conducted by computing the ICCs. Agreement was set as slight (ICC or κ, 0–0.20), fair (ICC or κ, 0.21–0.40), moderate (ICC or κ, 0.41–0.60), substantial (ICC or κ, 0.61–0.80), or almost perfect (ICC or κ, 0.81–1.00). CI, confidence interval; ICC, intraclass correlation coefficient; IQR, interquartile range; LR-TR, Liver Imaging Reporting and Data System treatment response; NA, not available; SD, standard deviation.
Across readers, masslike arterial phase hyperenhancement was observed in 46.0–69.8% of lesions [consensus: 49.2% (31/63), Fleiss κ =0.44, 95% CI: 0.27–0.59], masslike washout in 28.6–41.3% of lesions [consensus: 33.3% (21/63), Fleiss κ =0.46, 95% CI: 0.27–0.62], masslike enhancement in general in 46.0–69.8% of lesions [consensus: 49.2% (31/63), Fleiss κ =0.44, 95% CI: 0.27–0.59], diffusion restriction in 77.8–90.5% of lesions [consensus: 92.1% (58/63), weighted Fleiss κ =0.66, 95% CI: 0.55–0.79], and mild-moderate T2 hyperintensity in 92.1–95.2% of lesions [consensus: 98.4% (62/63), weighted Fleiss κ =0.79, 95% CI: 0.68–0.86]. The LR-TR Viable category was assigned in 46.0–69.8% of lesions [consensus: 49.2% (31/63), weighted Fleiss κ =0.39, 95% CI: 0.25–0.54].
Discussion
Our study demonstrated that in patients who received TACE plus systemic therapy and subsequent resection for HCC, MRI-based LR-TRA v2024 the LR-TR Viable category showed a PPV of 100% (per-lesion and per-patient analyses) as well as NPVs of 45.5% (per-patient analysis) and 46.9% (per-lesion analysis) based on consensus interpretations for detecting any (i.e., non-pCR) residual viable tumor. For detecting >10% residual viable tumor, the LR-TR Viable category showed PPVs of 80.6% (per-lesion analysis) and 82.8% (per-patient analysis) as well as NPVs of 84.4% (per-lesion analysis) and 86.4% (per-patient analysis) based on consensus interpretations.
pCR, representing microscopic complete tumor necrosis, has been used as an objective reference criterion in many studies (7-11). LR-TRA v2024 defines posttreatment viability as the presence of live tumor cells within or along the margin of a treated lesion. In our study, the PPV of LR-TR Viable category was 100% (on both lesion- and patient-levels) for detecting any residual viable tumor. This highlights the potential of LR-TRA CT/MRI Nonradiation v2024 in detecting pathologically viable tumors, and thereby indicates that all LR-TR Viable lesions may warrant further treatment. However, the NPVs of LR-TR Viable category were only 46.9% on per-lesion level and 45.5% on per-patient level, reflecting the limited sensitivity of MRI in detecting microscopic or small foci of residual viable tumors. Therefore, caution is needed for the interpretation of LR-TRA nonviability (in other words, “clinical complete response”) in routine clinical practice. For instance, subsequent curative-intent treatment such as resection, ablation or transplantation may be more radical and potentially beneficial than the “watch and wait” strategy for patients with LR-TR nonviable lesions. Notably, the accuracies of LR-TR Viable category ranged between 73% and 76.5% for detecting any residual viable tumor, which were slightly higher than a previous study (accuracies, 60–71%) using LR-TRA v2018 on patients treated with TACE (8). This improvement may be attributed in part to the biological effect of systemic agent’s on tumor microenvironment (16) or the modified definition of enhancement to indicate viability. The definitional change is from presence of “arterial phase hyperenhancement (i.e., masslike arterial phase hyperenhancement), washout appearance (i.e., masslike washout), or enhancement similar to pretreatment” in v2018 to the broader “enhancing area (any degree, any phase) that occupies space” (i.e., masslike enhancement in general) in v2024. Furthermore, v2024 formally incorporates two ancillary features favoring viability: “diffusion restriction (any degree) and mild-moderate T2 hyperintensity”, potentially enhancing performances (1,17). These preliminary data suggest the potential of LR-TRA in effectively evaluating treatment response to TACE plus systemic therapy.
MPR, referring to residual viable tumor proportion ≤10%, has been strongly associated with posttreatment prognosis (12-14). The LR-TR Viable category showed a PPV of 80.6% on per-lesion level and 82.8% on per-patient level as well as an NPV of 84.4% on per-lesion level and 86.4% on per-patient level for detecting >10% residual viable tumor. These results underscored the potential of LR-TRA as a promising decision-making tool. For example, for LR-TR nonviable lesions, there is ~85% probability that the proportion of residual viable tumor is less than 10%. Therefore, drug dose reduction or local treatment interval extension may be considered to balance the antitumor benefits and safety. Additionally, the accuracies of LR-TR Viable category for detecting >10% residual viable tumor were 82.5–84.3%, which were higher than that of LR-TR Viable category for detecting any residual viable tumor (non-pCR). This indicates that although imaging is not sensitive in depicting microscopic residual tumor, it performs well when there is a certain amount (>10%) of live tumor cells in the treated lesion.
Moderate and fair inter-reader agreement were observed for the “masslike arterial phase hyperenhancement enhancement”, “masslike washout”, “masslike enhancement” in general and LR-TR category (Fleiss κ =0.44, 0.46, 0.44 and 0.39, respectively), likely in part explained by the complexity of lesion components (e.g., varying degrees of lesion necrosis and residual viable tumors) after receiving TACE plus systemic therapy. Our results are slightly lower than previous studies, in which LR-TR category generally showed moderate concordance between readers (Fleiss κ =0.55) (8). After eliminating the kappa paradox, inter-reader agreement was substantial for “diffusion restriction” (weighted Fleiss κ =0.66) and “mild-moderate T2 hyperintensity” (weighted Fleiss κ =0.79). However, inter-reader agreement was particularly low for the “treatment-specific expected enhancement” (Fleiss κ =0.05). We speculate that it is attributed to the relatively vague definition and is highly dependent on the readers’ expertise in interpreting posttreatment images. In this regard, further detailed clarification, examples, and illustrations may help to improve inter-reader agreement.
Our study has several limitations. Firstly, it was a single-center retrospective study with a relatively small sample size. Secondly, the included patients with HCCs were managed with various types of systemic combination therapy, potentially leading to different posttreatment appearances, and potentially impacting the interpretation of “treatment-specific expected enhancement”. Thirdly, the study cohort comprised patients who had all undergone surgical resection following successful downstaging. While this approach may raise a potential risk of selection bias through the exclusion of non-surgical candidates, it remains the sole methodology currently available for obtaining histopathologically confirmed residual tumor information. Lastly, the use of microscopic tumor complete necrosis (i.e., pCR) as the reference standard may have introduced bias in performance results, as LR-TRA was designed to detect gross viable tumors given the inherent inability of imaging to detect microscopic or small foci of residual tumor (18). However, our study suggested that MPR might be a potential alternative for pCR.
Conclusions
In conclusion, our results demonstrated that LR-TRA can be applied to evaluated the treatment response of HCC to TACE plus systemic therapy. The positive predictive value for the detection of any (i.e., non-pCR) residual viable tumor was 100%, and the accuracies for the detection of >10% (i.e., non-MPR) residual viable tumor were 82.5–84.3%. Therefore, LR-TRA may be a promising decision-making tool to select subsequent resection or locoregional treatment (e.g., ablation, radiation-based LRT) for patients receiving TACE plus systemic therapy for HCC.
Acknowledgments
None.
Footnote
Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1308/rc
Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1308/dss
Funding: This study was supported by
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1308/coif). V.C. is a consultant for Bayer. M.R.B. is a principal investigator on grants paid to his institution from the National Cancer Institute, Siemens Healthineers, Bayer Healthcare, NGM Biopharmaceuticals, Madrigal Pharmaceuticals, Carmot Therapeutics, and Corcept Therapeutics. H.J. is a stock owner of Kanghong Technology Co., Ltd. The other authors have no conflicts of interest to declare.
Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study received approval from the Institutional Review Board of West China Hospital {approval No. [2022] 1993}. Informed consent requirements were waived due to the retrospective nature of the study. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.
Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.
References
- American College of Radiology Web site. LI-RADS® CT/MRI Nonradiation TRA v2024. 2024. Accessed 15 May 2024. Available online: https://www.acr.org/Clinical-Resources/Clinical-Tools-and-Reference/Reporting-and-Data-Systems/LI-RADS
- Lencioni R, Kudo M, Erinjeri J, Qin S, Ren Z, Chan S, Arai Y, Heo J, Mai A, Escobar J, Chuken YAL, Yoon JH, Tak WY, Suttichaimongkol T, Bouattour M, Lin SM, Żotkiewicz M, Udoye S, Cohen G, Sangro B. EMERALD-1: A phase 3, randomized, placebo-controlled study of transarterial chemoembolization combined with durvalumab with or without bevacizumab in participants with unresectable hepatocellular carcinoma eligible for embolization. J Clin Oncol 2024;42:LBA432.
- Kudo M, Ueshima K, Ikeda M, Torimura T, Tanabe N, Aikata H, et al. Final Results of TACTICS: A Randomized, Prospective Trial Comparing Transarterial Chemoembolization Plus Sorafenib to Transarterial Chemoembolization Alone in Patients with Unresectable Hepatocellular Carcinoma. Liver Cancer 2022;11:354-67. [Crossref] [PubMed]
- Peng Z, Fan W, Zhu B, Wang G, Sun J, Xiao C, Huang F, Tang R, Cheng Y, Huang Z, Liang Y, Fan H, Qiao L, Li F, Zhuang W, Peng B, Wang J, Li J, Kuang M. Lenvatinib Combined With Transarterial Chemoembolization as First-Line Treatment for Advanced Hepatocellular Carcinoma: A Phase III, Randomized Clinical Trial (LAUNCH). J Clin Oncol 2023;41:117-27. [Crossref] [PubMed]
- Llovet JM, Lencioni R. mRECIST for HCC: Performance and novel refinements. J Hepatol 2020;72:288-306. [Crossref] [PubMed]
- Eisenhauer EA, Therasse P, Bogaerts J, Schwartz LH, Sargent D, Ford R, Dancey J, Arbuck S, Gwyther S, Mooney M, Rubinstein L, Shankar L, Dodd L, Kaplan R, Lacombe D, Verweij J. New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1). Eur J Cancer 2009;45:228-47. [Crossref] [PubMed]
- Park S, Joo I, Lee DH, Bae JS, Yoo J, Kim SW, Lee JM. Diagnostic Performance of LI-RADS Treatment Response Algorithm for Hepatocellular Carcinoma: Adding Ancillary Features to MRI Compared with Enhancement Patterns at CT and MRI. Radiology 2020;296:554-61. [Crossref] [PubMed]
- Shropshire EL, Chaudhry M, Miller CM, Allen BC, Bozdogan E, Cardona DM, King LY, Janas GL, Do RK, Kim CY, Ronald J, Bashir MR. LI-RADS Treatment Response Algorithm: Performance and Diagnostic Accuracy. Radiology 2019;292:226-34. [Crossref] [PubMed]
- Chaudhry M, McGinty KA, Mervak B, Lerebours R, Li C, Shropshire E, Ronald J, Commander L, Hertel J, Luo S, Bashir MR, Burke LMB. The LI-RADS Version 2018 MRI Treatment Response Algorithm: Evaluation of Ablated Hepatocellular Carcinoma. Radiology 2020;294:320-6. [Crossref] [PubMed]
- Seo N, Kim MS, Park MS, Choi JY, Do RKG, Han K, Kim MJ. Evaluation of treatment response in hepatocellular carcinoma in the explanted liver with Liver Imaging Reporting and Data System version 2017. Eur Radiol 2020;30:261-71. [Crossref] [PubMed]
- Cools KS, Moon AM, Burke LMB, McGinty KA, Strassle PD, Gerber DA. Validation of the Liver Imaging Reporting and Data System Treatment Response Criteria After Thermal Ablation for Hepatocellular Carcinoma. Liver Transpl 2020;26:203-14. [Crossref] [PubMed]
- Allard MA, Sebagh M, Ruiz A, Guettier C, Paule B, Vibert E, Cunha AS, Cherqui D, Samuel D, Bismuth H, Castaing D, Adam R. Does pathological response after transarterial chemoembolization for hepatocellular carcinoma in cirrhotic patients with cirrhosis predict outcome after liver resection or transplantation? J Hepatol 2015;63:83-92. [Crossref] [PubMed]
- Travis WD, Dacic S, Wistuba I, Sholl L, Adusumilli P, Bubendorf L, et al. IASLC Multidisciplinary Recommendations for Pathologic Assessment of Lung Cancer Resection Specimens After Neoadjuvant Therapy. J Thorac Oncol 2020;15:709-40. [Crossref] [PubMed]
- General Office of National Health Commission. Standard for diagnosis and treatment of primary liver cancer (2022 edition). Journal of Clinical Hepatology 2022;38:288-303.
- Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159-74.
- Llovet JM, Montal R, Sia D, Finn RS. Molecular therapies and precision medicine for hepatocellular carcinoma. Nat Rev Clin Oncol 2018;15:599-616. [Crossref] [PubMed]
- Zhou S, Zhou G, Shen Y, Xia T, Zhao B, Sun Z, Gao L, Li B, Wang W, Zhang S, Opara NC, Chen X, Ju S, Wang YC. LI-RADS Nonradiation Treatment Response Algorithm Version 2024: Diagnostic Performance and Impact of Ancillary Features. AJR Am J Roentgenol 2025;224:e2432035. [Crossref] [PubMed]
- Kielar A, Fowler KJ, Lewis S, Yaghmai V, Miller FH, Yarmohammadi H, Kim C, Chernyak V, Yokoo T, Meyer J, Newton I, Do RK. Locoregional therapies for hepatocellular carcinoma and the new LI-RADS treatment response algorithm. Abdom Radiol (NY) 2018;43:218-30. [Crossref] [PubMed]
