From LI-RADS Classification to HCC Pathology: A Retrospective Single-Institution Analysis of Clinico-Pathological Features Affecting Oncological Outcomes after Curative Surgery

Background: The latest Liver Imaging Reporting and Data System (LI-RADS) classification by the American College of Radiology has been recently endorsed in the American Association for the Study of Liver Disease (AASLD) guidelines for Hepatocellular carcinoma (HCC) management. Although the LI-RADS protocol has been developed as a diagnostic algorithm, there is some evidence concerning a possible correlation between different LI-RADS classes and specific pathological features of HCC. We aimed to investigate such radiological/pathological correlation and the possible prognostic implication of LI-RADS on a retrospective cohort of HCC patients undergoing surgical resection. Methods: We performed a retrospective analysis of the pathological characteristics of resected HCC, exploring their distribution among different LI-RADS classes and analyzing the risk factors for recurrence-free, overall and cancer-specific survival Results: LI-RADS-5 (LR-5) nodules showed a higher prevalence of microvascular invasion (MVI), satellitosis and capsule infiltration, as well as higher median values of alpha-fetoprotein (αFP) compared to LI-RADS-3/4 (LR-3/4) nodules. MVI, αFP, satellitosis and margin-positive (R1) resection resulted as independent risk factors for recurrence-free survival, while LI-RADS class did not exert any significant impact. Focusing on overall survival, we identified patient age, Eastern Cooperative Oncology Group performance status (ECOG-PS), Model for End Stage Liver Disease (MELD) score, αFP, MVI, satellitosis and R1 resection as independent risk factors for survival, without any impact of LI-RADS classification. Last, MELD score, log10αFP, satellitosis and R1 resection resulted as independent risk factors for cancer-specific survival, while LI-RADS class did not exert any significant impact. Conclusions: Our results suggest an association of LR-5 class with unfavorable pathological characteristics of resected HCC; tumor histology and underlying patient characteristics such as age, ECOG-PS and liver disease severity exert a significant impact on postoperative oncological outcomes.


Introduction
Hepatocellular carcinoma (HCC) incidence is constantly rising worldwide, representing the third leading cause of cancer-related death and 85-90% of primary liver neoplasms [1,2].
Following the European Association for the Study of the Liver (EASL) and the American Association for the Study of Liver Disease (AASLD) guidelines, HCC diagnosis relies on triphasic contrast enhanced imaging [3,4]. Back in 2011, the American College of Radiology proposed a standardized diagnostic algorithm for liver imaging reporting, named Liver Imaging Reporting and Data System (LI-RADS), that was furtherly updated in 2013, 2014 and 2017, until the release of the latest version in 2018 [5].
Until this recent integration of LI-RADS protocol into clinical guidelines, HCC management was not supported by such probabilistic diagnostic tool: as a consequence, several LR-3/4 nodules underwent curative treatments such as percutaneous ablation (PA), surgical resection (SR) or liver transplantation (LT) [7], providing interesting subgroups for retrospective analysis.
Given this scenario, recent evidence supported a potential role of LI-RADS protocol on clinical decision making in the setting of HCC ablation vs. resection [8] and liver transplantation [9,10], suggesting an association of LI-RADS classes with specific pathological features of HCC nodules.
The aim of our study was to retrospectively assess the association of LR-3/4 and LR-5 subclasses with pathological characteristics of HCC treated by SR, investigating the risk factors for disease-free, overall and cancer-specific survival.

Materials and Methods
Study protocol followed the 1975 Declaration of Helsinki ethical guidelines, as revised in Brazil in 2013; all the subjects involved in the study gave their explicit informed consent for data collection and publication. Local ethical committees' review of the protocol deemed that formal approval was not required owing to the retrospective, observational and anonymous nature of this study.
Results are reported according to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) [11].

Study Design
The study enrolled all adult (age ≥ 18 years) HCC patients undergoing SR between 2006 and 2016, with a retrievable pre-treatment contrast-enhanced imaging study.
The most recent preoperative imaging (either computed tomography or magnetic resonance imaging) was retrospectively and blindly analyzed by three radiologists with 30-, 15-and 10-year experience in liver imaging, classifying all the nodules according to 2018 Li-RADS protocol, applying ancillary features when feasible.

Data Collection and Definition
All data were retrieved from a single university-affiliated, hepato-pancreato-biliary teaching center prospective database and anonymized prior to analysis.
Tumor characteristics such as number and size [mm] of HCC nodules, as well as alpha-fetoprotein (αFP) value [ng/m], Barcelona Clinic for Liver Cancer (BCLC) staging [7] and LI-RADS class [5] were also collected.
Pathological analysis included the definition of tumor grading according to Edmondson-Steiner's classification system [16], microvascular invasion (MVI) [17], satellitosis [18] and capsule infiltration [19], as well as the state of resection margins-margin-positive (R1) resection was defined as a <1 mm tumor distance from the cut surface.
Length of hospital stay [days] was collected, and postoperative complications were classified according to Clavien-Dindo classification [20].
The number of patients undergoing LT was also included. Last, the date of tumor recurrence and date and cause of patient death were recorded during follow-up.

Perioperative Management and Surgical Technique
Every case was evaluated during multidisciplinary team (MDT) meetings involving surgeons, hepatologists, diagnostic and interventional radiologists. Notably, these MDT meetings were attended before the blinded reclassification of the last preoperative imaging according to 2018 LI-RADS protocol, that therefore did not influence clinical decision making: in this setting, HCC diagnosis followed the noninvasive EASL criteria [3].
Impaired hepatic functional reserve (Child Pugh score C, intractable ascites, platelet count < 50.000/µL) and high operative risk (American Association of Anesthesiologists score > 3) were considered as contraindications to SR.
Liver resection was performed either open or laparoscopically, according to patients and tumor characteristics. Regardless of the chosen approach, all patients underwent intraoperative ultrasound for tumor and vascular mapping. Intermittent pedicle clamping was selectively used, with no longer than 15 min of clamping time and at least 5 min of release.
Parenchymal transection was performed using the CUSA ® Excel + Cavitron Ultrasonic Surgical Aspirator System (Integra, Ireland) and the Ultracision Harmonic scalpel (Ethicon Endo-Surgery, Cincinnati, OH, USA), while hemostasis and biliostasis on the liver cut surface were achieved using metallic clips, Hem-o-lock or non-absorbable sutures.
HCC diagnosis was confirmed by pathological evaluation of the resected specimens in all patients, that following SR underwent life-long surveillance for HCC recurrence following the EASL guidelines [3]. LT represented a leading treatment option in patients who experienced a transplantable recurrence [21].

Study Endpoints
The primary endpoint of our study was to investigate the correlations between clinicopathological characteristics of HCC patients and LI-RADS classes. As a secondary endpoint, we sought to analyze the risk factors affecting the oncological outcomes of SR.

Statistical Analysis
Continuous variables were described using median and interquartile range (IQR) while absolute numbers and proportions were reported for categorical variables. Baseline characteristics of LR-3/4 and LR-5 groups were compared using Mann-Whitney test and Chi-square test for continuous and categorical variables, respectively.
Uni-and multivariable Cox regression models were fitted for the time-to-event endpoints such as recurrence-free (RFS-time from resection to recurrence or death without recurrence) and overall survival (OS-time from resection to death), providing the hazard ratio (HR) and 95% confidence interval (95% CI) for identified risk factors.
Kaplan-Meier curves with the log-rank-test were also used to compare RFS and OS between LR-3/4 vs. LR-5 groups.
For the time to cancer-specific death, a competing regression analysis was performed considering death due to other causes as a competing event. The association of prognostic factors with this end-point was evaluated using cause-specific Cox models while Aalen-Johansen crude incidence curves with Gray test were estimated to compare LR-3/4 vs. LR-5 groups.

Results
The flowchart for patient selection is depicted in Figure 1. The starting population was composed of 389 HCC patients treated by SR between 2006 and 2016. Preoperative imaging was not retrievable for 136 patients that were therefore excluded from the analysis, while other 22 patients were excluded as classified as LR-TR-V, LR-TR-NV, LR-TR-E, LR-M and LR-TIV. Finally, we further excluded 45 patients that were lost after the first year of follow-up.

Patients and Tumor Characteristics
Clinical and demographic data of the whole study population are summarized in Table 1.

Survival Analyses
Uni-and multivariate analyses of risk factors for RFS, OS and cancer-specific survival are reported in Tables 3-5.

Discussion
Contrast-enhanced imaging represents the milestone for HCC diagnosis. According to current EASL guidelines, a nodule > 1 cm arising in a cirrhotic liver and presenting typical radiological hallmarks is eventually diagnosed as HCC. Such a binary approach has been challenged by the introduction of LI-RADS algorithm, which provides a probabilistic perspective to liver nodule diagnosis in at-risk patients, stratifying the probability of HCC (and overall malignancy) into specific classes.
Although the recent inclusion of LI-RADS protocol into AASLD guidelines represented a milestone towards the standardization of HCC diagnosis, there still is some concerns about the management of intermediate LI-RADS observations such as LR-3 (that potentially harbor an HCC focus in 38% of the cases) and LR-4 (with a 74% probability of HCC) nodules [6,22].
The prognostic value of LR-3 and LR-4 classes has been recently investigated in LT setting [9] through the blind application of LI-RADS categorization on a retrospective cohort of 245 transplanted HCC, highlighting a relatively high (40.9%) prevalence of such observations among all vital nodules detectable at last pre-LT imaging. Notably, inclusion of LR-3 and LR-4 categories (that should not be considered as HCC according to OPTN/UNOS policy [23,24]) into Metroticket 2.0 calculator [25] provided a higher prognostic accuracy, supporting the importance of such LI-RADS classes in the oncological outcomes of LT for HCC.
Similar conclusions have been suggested by another single-institution experience on a retrospective cohort of 72 patients undergoing LT [10], which highlighted a non-negligible prevalence of HCC (89%) and MVI (11%) in LR-4 observations, with not-statistically significant difference between LR-4 and LR-5 classes.
The oncological impact of LI-RADS classification has also been explored in the setting of PA and SR for single HCC [8]. After a blinded stratification of a highly selected (uninodular and treatment-naïve) retrospective cohort of 140 patients according to LI-RADS protocol, the Authors highlighted a significantly better outcome of SR over PA in LR-5 class, which was not observed in LR-3/4 subgroup of patients. These results were explained by a higher frequency of unfavorable pathological features in LR-5 classes, although not statistically significant.
The current study enrolled a relatively large cohort of 186 pathologically confirmed HCC treated by SR.
The first step of our analysis showed a higher prevalence of unfavorable histological findings in LR-5 class compared to LR-3/4 group, confirming a more aggressive pathological profiling of such class of observation and supporting the hypothesis that was raised in the above-mentioned study [8]. Moreover, it is important to highlight out how up to 22.6% of LR-3/4 observations in our cohort presented MVI, that has been already demonstrated as one of the most unfavorable histological features affecting worse oncological outcomes [26]. Another interesting finding was also represented by a non-negligible percentage of satellitosis (9.4%) and capsule infiltration (11.3%) even in LR-3/4 classes, supporting the need of cautious management of these intermediate-risk classes, as already suggested from other recently published studies [8][9][10]22].
The second step of our analysis sought to investigate the prognostic impact of several clinicopathological risk factors as well as the potential influence of LI-RADS classification on oncological outcomes of SR for HCC.
Interestingly, OS was significantly affected by impaired performance status and underlying liver function (as supported by the negative effect of higher MELD score and ECOG-PS 1) as well as being influenced by oncological characteristics such as the presence of MVI, satellitosis, higher log10αFP and R1 resection. Similarly to RFS, LI-RADS classification did not exert any statistically significant effect on OS.
Focusing on cancer-specific survival, we confirmed the detrimental effect of higher MELD score [30] as well as higher log10αFP, satellitosis and R1 resection on cancer-specific survival; once again, we did not observe any statistically significant association between LI-RADS class and cancer-specific outcomes.
In summary, our results confirmed the current evidence concerning the risk factors for cancer-specific outcomes after curative resection for HCC [31].
Despite the observed association of LR-5 class with several unfavorable pathological features, the absence of an independent correlation between LI-RADS classification and oncological results of SR could rely on the higher relative weight of tumor pathology and underlying tumorigenic liver asset in the determination of oncological results.
The main limitations of this study are represented by its retrospective nature, which may imply selection and indication biases, as well as the loss of many cases because of irretrievable preoperative imaging (196/349 patients; 56.16%) that caused a significant decrease in the study population.

Conclusions
In conclusion, the current study supported the hypothesis of a significant association of higher LI-RADS class with unfavorable pathological features such as MVI, satellitosis or capsule infiltration [8,32], confirming the detrimental effect of tumor pathology and margin-status on cancer-specific outcomes after SR for HCC [26][27][28][29][30][31]33].
Our results reinforce the role of preoperative radiological assessment as a predictor of tumor biology, a field that will find further application with the development and implementation of radiomics in the upcoming precision-medicine era [34,35].
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/diagnostics12010160/s1, Figure S1: Kaplan-Meier estimates of relapse-free survival, Figure S2: Kaplan-Meier estimates of overall survival, Figure S3: Aalen-Johansen crude incidence estimates of cancer-specific death (death for other causes as competing risk). Institutional Review Board Statement: Ethical review and approval were waived for this study, due to the retrospective and anonymous nature of the analysis.