The Course of LIRADS 3 and 4 Hepatic Abnormalities as Correlated With Explant Pathology: A Single Center Experience

Abstract

Background and aims

The Liver Reporting and Data System (LI-RADS) is the standard classification of imaging findings of hepatic abnormalities for hepatocellular carcinoma (HCC) surveillance. We aimed to study the course of LI-RADS 3 and 4 (LR-3 and LR-4) abnormalities through correlations with explant pathology.

Methods

A single center retrospective study of liver transplant recipients between January 2016 and September 2019 with HCC on explant pathology was conducted. Eligible patients were divided into three subgroups based on their LI-RADS classification: LR-3/4, LR-5 only, and combination of LR-3/4/5.

Results

There were 116 eligible patients with 99 LR-3/4 observations (60 LR-3 and 39 LR-4); the rest had LR-5 lesions. LR-4 more often than LR-3 observations progressed to LR-5 (36% vs 12%) and with shorter duration during follow-up (median 175 days and 196 days). Mean size growth of LR-3 and LR-4 abnormalities were 2.6 and 3.8 mm; median growth rates were 0.2 and 0.4 mm/month, respectively. Numbers of HCC lesions per explant, largest HCC lesion size, and cumulative size were higher in LR-3/4/5 subgroup than LR-5 subgroup (P = 0.007, 0.007 and 0.006, respectively); 68% of LR-3 and 82% of LR-4 abnormalities were confirmed HCC on explant (P = 0.09).

Conclusion

Compared to LR-3, more LR-4 abnormalities progressed to LR-5 (12% and 36%, respectively) in a shorter time and with faster growth rate. A high proportion of LR-3 and LR-4 lesions (68% and 82%, respectively) were confirmed HCC on explant, raising the question of whether excluding HCC based on radiologic criteria alone is adequate in those with LR-3/4 abnormalities.

Hepatocellular carcinoma (HCC) is the most common primary malignant tumor of the liver and is a leading indication for liver transplantation (LT) in the United States.^1,2 Due to its characteristic features on imaging, HCC rarely requires tissue diagnosis via liver biopsy.³ In 2018, the Liver Reporting and Data System (LI-RADS) classification system (LI-RADS v2018) was updated to standardize the terminology used to describe hepatic abnormalities and better reflect the absolute and relative probability of malignancy in patients at risk for HCC. Under this classification system, LI-RADS categories 3–5 reflect an increasing probability of underlying HCC, whereas those that are LI-RADS 1 and 2 are considered to be benign.⁴

When evaluating a focal abnormality, the differentiation between LI-RADS category 4 and 5 (LR-4 and LR-5) is essential, as this dramatically changes patient management.^5,6 While many LR-4 abnormalities are typically observed with serial imaging, LR-5 lesions are usually referred for treatment, such as locoregional therapy, resection or LT.⁷ Nevertheless, prior studies indicate that a large majority of LR-4 lesions and nearly half of LR-3 on imaging ultimately lead to confirmation of HCC.^8,9 At the outset, patients are known to differ in their background severity of underlying liver disease and available treatment options, such that for some, the difference in HCC probability between LR-5 and LR-3 or LR-4 lesions is small.⁶ The limitations of the LI-RADS classification system are highlighted by the observation that, while the vast majority of LR-5 nodules are ultimately proven to be HCC, the reverse is less likely and small or early HCCs are more likely to be improperly characterized as LR-3 and 4.^{8, 9, 10, 11}

To date, few studies have specifically focused on the course of LR-3 and LR-4 observations among patients at high risk for HCC, and current recommendations suggest a multidisciplinary approach, particularly of an LR-4 lesion and strategies would include a biopsy or follow-up imaging at short intervals.^12,13 A systematic review of LI-RADS categorization by CT and MRI has suggested that a more active management strategy for indeterminate observations is warranted.⁹ The objective of this study was to correlate, in LT recipients, the course of radiologically characterized LR-3 and LR-4 observations with explant findings in order to analyze the potential implication in such patients where tumor burden is an essential component of ideal management and listing for LT.

Materials and methods

Study Design and Data Source

This was a retrospective cohort study using data from patients' electronic medical health record (EMR) at a single, large LT center (University of Pennsylvania). After identifying patients who met the study-defined inclusion and exclusion criteria, the EMR was queried to extract all demographic, clinical, laboratory, pathology reports, and radiologic data in which original reports used. The radiologic interpretation pretransplant was experienced by radiologists at a major U.S. transplant center at a multidisciplinary conference but was not re-reviewed for this particular project. Subsequent to the transplant, the explant pathology was reviewed, and a tumor burden assessment and correlation with radiologic assessment was made. The protocol for this study was approved by the University of Pennsylvania's Institutional Review Board (IRB).

Study Population

The LI-RADS classification system has not been systematically used to describe hepatic observations until recently.⁴ Therefore, to ensure that study subjects had consistent use of LI-RADS classification in their radiological assessments, this study only included data from more recent years. All adult patients (≥18 years old) who underwent LT at the University of Pennsylvania between January 2016 and September 2019 with HCC on explant pathology were eligible for inclusion. Only individuals with pre-LT radiology reports using LI-RADS classification terminology of their respective hepatic observations were included. As the objective of this study was to evaluate the course of pre-transplant LR-3/4 abnormalities among patients with identified HCC on explant, only individuals with pre-LT imaging demonstrating ≥1 observation classified as LR-3 or higher were included. Similarly, the first imaging study of interest was defined as the date of the first imaging study demonstrating ≥1 observation(s) classified as LR-3 or higher. Subjects were then divided into three subgroups based on LI-RADS category on initial imaging: only LR-3 and/or LR-4 observation(s) present (LR-3/4 group), only definite HCC lesion(s) present (LR-5 group), or a combination of LR-3/LR-4 and LR-5 present (LR-3/4/5 group) and for the purpose of comparing progression of the radiologic abnormalities among the subgroups.

Exposures and Outcomes

Sex, race/ethnicity, age, body mass index (BMI), and Child-Pugh score were recorded at the time of first imaging. The highest AFP level between first imaging of interest and LT was recorded. Patients’ Model for End-stage Liver Disease Sodium (MELD-Na) score was measured at the time of LT admission.

For each patient, all subsequent imaging reports between the initial imaging demonstrating ≥1 observation(s) LR-3 or higher and LT were reviewed. The following end points were evaluated for LR-3 and LR-4 abnormalities identified pre-LT: (i) progression to a higher LI-RADS category or to LR-5 on subsequent pre-LT imaging, (ii) tumor growth during waitlisting, and (iii) concordance with explant pathology findings. Secondary end points included overall HCC burden on explant with regard to tumor size and number.

Statistical Analysis

For continuous variables, means, and standard deviations were calculated if normally distributed and medians (inter-quartile ranges, IQR) were calculated if non-normally distributed. Chi-square tests and Fisher-exact tests (for cell sizes <5) were used to test between group differences for categorical variables. One-way analysis of variance (ANOVA) and Kruskal–Wallis one-way ANOVA on ranks, for variables normally and nonnormally distributed, respectively, were employed to test differences among three comparison groups for continuous variables. Student's t-tests and Wilcoxon rank-sum tests, for variables normally and nonnormally distributed, respectively, were used to test differences between two subgroups in pair-wise comparisons. All P-values were reported as two-sided and a P-value <0.05 was considered statistically significant.¹⁴ All data analyses were performed using NCSS 8 software.¹⁵

Results

Between January 2016 to September 2019, 192 adult LT recipients with HCC noted on explant pathology were identified (Figure 1). Of these, 20 were excluded due to the absence of LI-RADS 3–5 observations pre-LT (had incidental HCC), while 56 were excluded due to inconsistent use of the LI-RADS classification system or discrepancy in LI-RADS categorization (different LI-RADS categorization by different radiologists). Thus, the final study population included 116 patients. Thirty patients (26%) were categorized as only having LR-3/4 abnormalities on initial imaging, 57 patients (49%) had only LR-5, and 29 patients (25%) had a combination of LR-3/4/5 observations on their first imaging study.

Figure 1.

Flow diagram demonstrating selection of study cohort.

The majority of patients were male (87%) and Caucasian (81%). Age at LT ranged from 45 to 74 years (mean age of 64.6 ± 5.6) (Table 1). The mean age of the LR-5 group was significantly higher than that of the LR-3/4 group (66 ± 4.6 vs. 63 ± 6.2, respectively; P = 0.02). The median laboratory MELD-Na score at LT of the LR-3/4 group was higher than the two groups with ≥1 LR-5 lesion(s) on account of decompensations (14 [IQR: 11–22] vs. 10 [IQR 7–13] in the LR-5 only group and 10 [IQR: 8–13] in the LR-3/4/5 group; P = 0.003). Differences in BMI, peak AFP levels, and Child-Pugh score between each group were not statistically significant (Table 1). Patients with ≥1 LR-5 lesion(s) on initial imaging were significantly more likely to have HCV (69% LR-5 only vs. 66% LR-3/4/5 vs. 40% LR-3/4; P = 0.03). Similarly, LR-5 lesions were found in 85% of patients with viral hepatitis (HBV and/or HCV) compared to only 57% in patients with other etiologies (P = 0.003). There were two patients who did not have cirrhosis; the patients were from LR-3/4/5 and LR-5 group. The LR-5 patient who developed HCC in a noncirrhotic liver had HCV, whereas the patient in the LR-3/4/5 group did not have either HBV or HCV.

Table 1.

Pretransplant Clinical Characteristics of Overall Study Cohort and of Subgroups.

Characteristic	All (n = 116)	LI-RADS 3/4a observation(s) only (n = 30)	LI-RADS 5 observation(s) only (n = 57)	Combination of LI-RADS 3/4/5 observations (n = 29)	P-values
Male, N (%)	101 (87%)	26 (87%)	50 (88%)	25 (86%)	1.0b
Race, N (%)
White	94 (81%)	26 (87%)	44 (77%)	24 (83%)	0.91b
Black	17 (15%)	3 (10%)	10 (18%)	4 (14%)
Asian	5 (4%)	1 (3%)	3 (5%)	1 (3%)
Age∗	64.6 ± 5.6	63 ± 6.2	66 ± 4.6	63.6 ± 6.1	0.02; 0.76; 0.06; 0.07c
BMI	27.8 ± 4.6	28.6 ± 4.5	27.3 ± 4.2	27.9 ± 5.3	0.49b
Peak AFP level Pre-LT (Median (IQR))	8.9 (3.6–31.8)	7.6 (3.9–13.4)	11.1 (3.1–44.2)	10 (4.9–44.5)	0.61b
Primary liver disease etiology, N (%)
HCV	70 (60%)	12 (40%)	39 (69%)	19 (66%)	0.01; 0.07; 0.81; 0.03
HBV	9 (8%)	2 (7%)	6 (10%)	1 (3%)	0.61; 1.0; 0.42; 0.61
EtOH	8 (7%)	4 (13%)	3 (5%)	1 (3%)	0.23; 0.35; 0.35; 0.36
AIH	3 (3%)	2 (7%)	1 (2%)	0 (0%)	0.27; 0.49; 1.0; 0.32
NASH	23 (20%)	8 (27%)	8 (14%)	7 (24%)	0.16; 1.0; 0.25; 0.29
Other	1 (1%)	1 (3%)	0 (0%)	0 (0%)	0.35; 1.0; 1.0; 0.51
Unknown	2 (2%)	1 (3%)	0 (0%)	1 (3%)	0.34; 1.0; 0.34; 0.26
Cirrhosis	114 (98%)	30 (100%)	56 (98%)	28 (97%)	–
Child-Pugh Score, n (%)
A	64 (56%)	12 (40%)	35 (64%)	17 (59%)	0.2b
B	35 (31%)	12 (40%)	16 (29%)	7 (24%)
C	14 (12%)	6 (20%)	4 (7%)	4 (14%)
MELD-Na Score (Median (IQR))	11 (8–15)	14 (11–22)	10 (7–13)	10 (8–13)	0.0003; 0.001; 0.96; 0.0025c

^aLR-3 and LR-4 observations without any intervention.

^bNot statistically significant in all pairwise comparisons and only P-value for overall difference listed.

^cWhen any pairwise comparisons was significant, the P-values for each comparisons are listed. The P-values reported signify the following comparisons: LR-3/4 vs. LR-5; LR-3/4 vs. LR-3/4/5; LR-5 vs. LR-3/4/5; overall difference among all three groups.

Course of LR-3/4 Abnormalities

Between initial imaging study and LT, the course of 99 LR-3/4 observations (n = 60 LR-3 and n = 39 LR-4) was evaluated among 59 patients; 23% of LR-3 and 21% of LR-4 abnormalities did not have sequential imaging prior to transplant (N = 22); 53% of LR-3 and 44% of LR-4 observations remained in the same LI-RADS category throughout waitlisting (Table 2). Twelve percent (7 out of 60) of initial LR-3 progressed to LR-4 on the last pre-LT imaging study, 36% (14 out of 39) of LR-4 progressed to LR-5, and 12% (7 out of 60) of LR-3 progressed to LR-5 (Figure 2). Mean size growth during waitlisting was 2.6 ± 4.2 mm for LR-3 and 3.8 ± 6.4 mm for LR-4 observations, which was not significantly different. Among LR-3 and LR-4 abnormalities with identified progression during waitlisting (N = 28), time from LR-3 or LR-4 progression to LR-5 was also not statistically different (Table 2). There was no statistical difference in median growth rate of LR-3 and LR-4 observations (0.24 (0–0.77) vs. 0.39 (0–0.95) mm per month, respectively). When further subgrouping, LR-3 and LR-4 observations according to their final LI-RADS category, size growth, growth rate, and time to progression were not significantly different among these subgroups (Table 3). For subset of patients with more than one lesion, we did not observe within patients a correlation among different lesions in term of LI-RADS progression (P = 0.25), progression to LR-4 (P = 0.23), or progression to LR-5 (P = 0.53). We treated these lesions as independent observations as patients-specific factors did not have statistically significant effect on lesion progression.

Table 2.

Natural History of LI-RADS 3 and 4 Observations Between First Imaging and Last Imaging Before LT.

	LR-3 (n = 60)	LR-4 (n = 39)	P-value
Imaging Modalities
MRI	55 (92%)	34 (87%)	0.70
CT scans	3 (5%)	0 (0%)	0.27
MRI + CT scans	2 (3%)	5 (13%)	0.11
Change during waitlisting, n (%)
No change/stable LR classification	32 (53%)	17 (44%)	0.41
Progressed, to LR-4a	7 (12%)	–	–
Progressed, to LR-5	7 (12%)	14 (36%)	0.009
No follow-up	14 (23%)	8 (21%)	0.93
Time observed (days), median (IQR)
Stable until LT	192 (123–279)	190 (102–435)	0.97
Progressed to LR-4a	312 (84–478)	–	–
Progressed to LR-5	191 (92–312)	167 (103–223)	0.26
Size Growtha (mm) (Median (range))	1.5 (-6–12)	3 (-13–21)	0.36
Growth Rate (mm per month) (Median (IQR))	0.2 (0–0.8)	0.4 (0–1)	0.56

^aObservations that were initially LR-3 and progressed to LR-4 before LT.

Figure 2.

Course of LR-3 and LR-4.

Table 3.

Disease Course of LR-3/4 Abnormalities According to Initial Imaging Subclassifications.

	Stable LR-3 (N = 32)	LR-3 to LR-4 (N = 7)	LR-3 to LR-5a(N = 7)	Stable LR-4 (N = 17)	LR-4 to LR-5a(N = 14)	P-values
Size Growth (mm) (Median (IQR))	1 (0–4.5)	7.5 (1–9)	5 (1–10)	0 (0–7)	4 (1–9)	0.1b
Growth Rate (mm per month)	0.2 (0–0.8)	0.45 (0.1–1.2)	0.3 (0.2–1.5)	0 (0–0.4)	0.8 (0.3–1)	0.28b
Duration in Days (Median (IQR))	196 (123–275)	380 (84–493)	191 (92–332)	174 (95–389)	175 (106–199)	0.78b

^aLR-3 observations that advanced to LR-5 without being categorized as LR-4 in between.

^bNot statistically significant in all pairwise comparisons and only P-value for overall difference listed.

LI-RADS and Explant Pathology

We found that 68% of LR-3 and 82% of LR-4 abnormalities were confirmed HCC on explant pathology (P = 0.09). The number of hepatic abnormalities identified on the first and last imaging studies prior to LT was highest in the LR-3/4/5 group compared to the LR-3/4 and LR-5 groups (P < 0.001). Notably, patients in the LR-3/4/5 group had a greater number of HCC lesions confirmed on explant pathology: median of 3 (2–4) in the LR-3/4/5 group vs 2 (1–3.3) in the LR-3/4 group and 2 (1–2.5) in the LR-5 group (Table 4). Median largest HCC tumor sizes were 22 (14–28), 20 (16–30) and 23 (20–32) mm in the LR-3/4 group, LR-5 group and LR-3/4/5 group respectively. Only the largest tumor sizes in the LR-5 and LR-3/4/5 groups were significantly different from one another (P = 0.007). The median cumulative sizes were 29 (19–46), 27 (19–42) and 45 (36–57) mm in LR-3/4 group, LR-5 group and LR-3/4/5 group respectively. There was no statistical difference between cumulative size in the LR-3/4 and LR-5 groups, but the differences between these groups and the LR-3/4/5 group were significant (P = 0.006 with LR-5 and P = 0.03 with LR-3/4 group). Other malignant lesions identified on explant included cholangiocarcinoma and hepatocholangiocarcinoma (N = 4 in LR-3/4 group, N = 4 in LR-5 group, and N = 2 in LR-3/4/5 group). No benign tumors were identified on explant for any of the patients in the LR-3/4 group. Patients in LR-5 group (N = 2) were found to have hemangiomas and the benign lesions found in patients from the LR-3/4/5 group (N = 3) included hepatic hemangioma, peribiliary gland hamartoma, and bile duct adenoma. In this study, out of 109 observations identified on the explant that were missed during the radiologic surveillances, 52 of them (47.7%) were retrospectively subcentimeter lesions that met the criteria for at least LR-3 category.

Table 4.

Explant Pathology Findings According to Initial Imaging Sub-group.

Characteristic	All (n = 116)	LI-RADS 3/4 observation(s) only (n = 30)	LI-RADS 5 observation(s) only (n = 57)	Combination of LI-RADS 3/4/5 observations (n = 29)	p-valuesa
Imaging Modalities
MRI + CT Scans	8 (7%)	5 (17%)	0 (0%)	3 (10%)	0.004; 0.7; 0.04; 0.003
MRI Only	102 (88%)	25 (83%)	51 (89%)	26 (90%)	0.23; 0.71; 1.0; 0.75
CT Scans Only	6 (5%)	0 (0%)	6 (11%)	0 (0%)	0.09; 1.0; 0.09; 0.03
Number of observation identified on imaging First identified
1	72 (62%)	20 (67%)	47 (83%)	5 (17%)	0.16; 0.0003; 0.0001; <1E-5
2	30 (26%)	7 (23%)	7 (12%)	16 (55%)	0.3; 0.025; 0.0001; 0.0009
3	11 (9%)	3 (10%)	3 (5%)	5 (17%)	0.41; 047; 0.11; 0.16
>3	3 (3%)	0	0	3 (10%)	1.0; 0.11; 0.04; 0.012
Most recent prior to LT
1	52 (45%)	19 (64%)	33 (58%)	0	0.07; <1E-5; <1E-5; <1E-5
2	38 (33%)	7 (23%)	17 (30%)	14 (48%)	0.15; 0.08; 0.15; 0.11
3	16 (14%)	3 (10%)	7 (12%)	6 (21%)	1.0; 0.3; 0.35; 0.48
>3	10 (8%)	1 (3%)	0	9 (31%)	0.34; 0.0006; 2E-6; <1E-6
Number of HCC per explant
1	44 (38%)	15 (50%)	28 (49%)	1 (3%)	0.94; 5E-5; <1E-5; <1E-5
2	29 (25%)	6 (20%)	16 (28%)	7 (24%)	0.57; 0.94; 0.8; 0.73
3	21 (18%)	3 (10%)	9 (16%)	9 (31%)	0.53; 0.06; 0.16; 0.11
>3	22 (19%)	6 (20%)	4 (7%)	12 (42%)	0.09; 0.14; 0.002; 0.007
Size of largest HCC lesion in mm. (Median (IQR))	22 (16–29)	22 (14–28)	20 (16–30)	23 (20–32)	0.91; 0.17; 0.007; 2E-5
Cumulative Size of HCC observations in mm. (Median (IQR))	33 (21–49)	29 (19–46)	27 (19–42)	45 (36–57)	0.76; 0.03; 0.006; 0.0005
Locoregional Therapy,n (%)
None	10 (8%)	9 (30%)	1 (2%)	0 (0%)	0.0002; 0.002; 1.0; 1.5E-5
TACE	97 (84%)	17 (57%)	53 (95%)	27 (95%)	0.0001; 0.002; 1.0; 7E-5
Ablation	6 (5%)	3 (10%)	2 (4%)	1 (3%)	0.33; 0.61; 1.0; 0.54
Other	2 (2%)	1 (3%)	0 (0%)	1 (3%)	0.34; 1.0; 0.34; 0.26
Explant Pathology, n (%)
Only HCC	101 (87%)	26 (87%)	51 (90%)	24 (83%)	0.73; 0.73; 0.5; 0.63
Other malignancyb	10 (9%)	4 (13%)	4 (7%)	2 (7%)	0.44; 0.67; 1.0; 0.68
Benign originc	5 (4%)	0 (0%)	2 (3%)	3 (10%)	0.11; 0.54; 0.33; 0.11

^aWhen any pairwise comparisons was significant, the P-values for each comparisons were listed. The numbers signify the following comparisons: LR-3/4 vs. LR-5; LR-3/4 vs. LR-3/4/5; LR-5 vs. LR-3/4/5; overall differences.

^bAdditional malignancies found were cholangiocarcinoma and hepatocholangiosarcoma.

^cAdditional benign tumors; Two patients in LR-5 group were found with hepatic hemagiomas and three patients from LR 3–5 group were found with hepatic hemangioma, peribiliary gland hamartoma, bile duct adenoma.

Discussion

The LI-RADS classification system is used to standardize the terminology used to describe hepatic abnormalities on radiology reports and better reflect the relative probability of malignancy in patients at risk for HCC [4]. While this classification system was first introduced in 2011, it has only recently become widely used in clinical practice.^4,16,17 Currently, patients with LR-5 lesions are referred for treatment, whereas LR-3/4 observations are typically observed with serial imaging even though LR-4 lesions have been classified as probable HCC and most often are not treated pretransplant, despite data being shown that the more complete tumor treatment pre-LT has lower risk of recurrence.^7,18,19

The objectives of our study were to evaluate the course of the LR-3/4 observations specifically and to correlate radiologic categorization with explant pathology. The study design was felt to be the most appropriate and confirmatory by using explant pathology as, in clinical practice, most of the LR-4 and particularly the LR-3 lesions are relegated to a follow up and not a biopsy. While similar studies have evaluated the diagnostic accuracy of the LI-RADs classification system based on lesional biopsy, follow-up imaging analyses, or treatment response, data directly comparing LR-3/4 observations to transplant explant pathology is limited.^{9,20, 21, 22} Additionally, this study uses a unique categorization schema that aims to compare patients with mixed lesional categorization (LR-3/4/5) with those with only LR-3/4 and only LR-5 observations.

This study notes that LR-3 observations were relatively more stable than LR-4 lesions, which had twice the rate of progressing to LR-5. Around a third of LR-4 lesions progressed to LR-5, as noted to similar studies on the course of LR-4 observations.^{23, 24, 25} There were no significant differences in the course of LR-3 and LR-4 observations in terms of observed growth in size, growth rate, and the duration for their progression. Notably, while 36% of LR-4 and 12% of LR-3 observations progressed to LR-5 on imaging, 82% of LR-4 and 68% of LR-3 lesions were confirmed HCC on explant pathology. Further, patients in the LR-3/4/5 group were shown to have greater tumor burden than patients in the LR-3/4 and LR-5 groups through higher numbers of HCC lesions and larger HCC tumor size (both median larger size and cumulative size), especially the cumulative size as has been noted by others.²⁶

Findings from this study suggest that a more active management strategy might be warranted for LR-3/4 observations, particularly for LR-4 lesions of which a vast majority were found to be HCC on explant. A study from Latin America reported similar results, where 50% of LR-3 and 89% of LR-4 lesions were confirmed HCC on explant.¹⁸ Two systemic reviews evaluating LIRADs diagnostic performance also observed high proportions of LR-3/4 abnormalities being HCC, where 64% and 74% of HCC lesions were in the LR-4 category.^9,22 Other studies that looked into LIRADs diagnostic accuracy have reported no difference in LIRADS category 4 and LIRADS category 5 criteria for HCC diagnosis.^18,20 Further, studies investigating LR-4 progression to malignancy observed such to be the case within 6 months.^23,24 Taken together, these data call into question the reliability of diagnosing HCC solely based on radiologic criteria and suggest that more rigorous management guidelines be pursued. While LI-RADS guidelines currently suggest only follow-up radiographic surveillance in LR-3 and multidisciplinary discussion for tailored workup which may include biopsy in LR-4, the high malignant transformation rate demonstrated suggests that routine biopsy should be more actively considered for LR-4 observations in order to properly strategize comprehensive management, while in LR-3, biopsy should also be considered as one of the alternative means of investigation.

Interestingly, 48% of HCC lesions found on explant were missed during imaging surveillance were subcentimeter lesions (<1 cm). These generally are called SHNHRs (subcentimeter hypervascular nodules at high risk for developing into hepatocellular carcinomas) or SAELs (sub-centimeter arterially enhancing and hepatobiliary hypointense lesions). These observations are often missed, and a majority of them eventually evolve to a malignant state.^27,28 This again raises a question of adequacy of solely depending on LI-RADS in diagnosing HCC, especially since size is an important component of the LI-RADS classification system.^{29, 30, 31, 32} The issue of under-diagnosed observations should not be ignored considering that according to LI-RADS management guidelines, subcentimeter lesions are at most classified as LR-4 and recommended for follow-up.^33,34

Demographic data in this study showed that patients presenting with LR-5 lesions at the outset were older. Further, patients with HCV and/or HBV infection had significantly higher rates of LR-5 on initial imaging than patients without these infections (85% vs. 57%; P = 0.003). The potential implications of these findings are that patients who are younger or have nonviral liver diseases and who are identified to have LR-3/4 observation(s) on imaging may warrant closer surveillance and/or early liver biopsy to rule out HCC.

Despite the unique study results and rationale, this study had certain limitations. First, it is a small sample size due to the strict requirements of precise and unequivocal use of LI-RADS to characterize hepatic observations and particularly limiting the correlation to those who had a liver transplant. Additionally, all patients with observation of hepatic lesions were not included due to a lack of a consensus reading by the tumor board radiologists, particularly since LI-RADS is an “operator-dependent” tool and may be somewhat subjective. Reported interobserver reliability for individual features is variable; in arterial phase hyperenhancement, the agreement has been high, but the agreement has been poor in washout appearance or pseudocapsule.³⁵ Thus, the interpretation of imaging studies may not be high in concordance and there may be discrepancy in assigning LI-RADS classification; yet this is a single center and more granular experience with relative homogeneity in reporting as opposed to a multicenter study with heterogeneity in LI-RADS characterization. Next, the initial imaging in this study is that on record at the institution, so it is possible that patients had initial imaging with LR-3/4 observations at other medical centers predating the images accessed in this study. Additionally, subcentimeter LR-3 observations may not have been included in the initial interpretation because the interpreting radiologist deemed they were LR-2 (e.g. vascular shunts, etc.). Finally, as it was not the focus of the study, we did not examine a relationship between the various categories of LIRADS observations and posttransplant outcomes although we have previously reported on our posttransplant outcomes with HCC over an extended period of time.³⁶ In 711 HCC patients transplanted at our center over an 8-year period, the recurrence rate was 13.5 percent.

In summary, the unique observations from this study on the course of LR-3/4 hepatic observations in the context of HCC emphasizes the need to closely monitor and evaluate them for HCC and suggest that a more active diagnosis and management strategy may be warranted for patients presenting with LR-3/4 LR characteristics. A biopsy of a lesion characterized as LR-4 rather than a follow-up strategy may facilitate earlier diagnosis and comprehensive treatment. Prospective trials, using a protocolized strategy that includes, standardized radiologic interpretation, biomarker assessment and agreed upon biopsy indications, who help us better understand and manage LR-3 and 4 lesions.

Credit Authorship Contribution Statement

K. Rajender Reddy: Conceptualization, Methodology. Panita Mettikanont, Neil Phillips, Anita Kalluri: Data Curation. Panita Mettikanont, Anita Kalluri: Writing - Review & Editing. Panita Mettikanont: Writing - Original Draft. Bao-Li Loza: Formal Analysis. Maarouf Hoteit, Abraham Shaked, Mark Rosen, Evan Siegelman, Emma Furth, Peter Abt, Kim Olthoff: Validation. K. Rajender Reddy, Therese Bittermann: Critical Revisions. K. Rajender Reddy: Supervision.

Conflicts of interest

All authors have none to declare.

Funding

None.

Data sharing

No additional data are available.