Inter-observer reliability and predictive values of triphasic computed tomography for microvascular invasion in hepatocellular carcinoma

Abstract

Background

Hepatocellular carcinoma (HCC) is the most frequent primary liver tumor globally and a leading cause of mortality in cirrhotic patients. Our study aimed to estimate the diagnostic performance of triphasic CT and inter-observer reliability in the preoperative detection of microvascular invasion (MVI) in HCC. Two independent radiologists accomplished a retrospective analysis for 99 patients with HCC to assess the CT features for MVI in each lesion. Postoperative histopathology was considered the gold standard.

Results

Multivariate regression analysis revealed that incomplete or absent tumor capsules, presence of TTPV, and absence of hypodense halo were statistically significant independent predictors of MVI. There was excellent agreement among observers in evaluating peritumoral enhancement, identifying intratumoral arteries, hypodense halo, TTPV, and macrovascular invasion. Also, our results revealed moderate agreement in assessing the tumor margin and tumor capsule.

Conclusion

Triphasic CT features of MVI are reliable imaging predictors that may be helpful for standard preoperative interpretation of HCC.

Background

Hepatocellular carcinoma (HCC) is the most prevalent primary hepatic tumor globally and a main cause of mortality in cirrhotic patients [1]. The correct and early diagnosis of HCC is fundamental for ideal management and improved long-term patient survival [2]. HCC can be assertively diagnosed through specific imaging characteristics on computed tomography (CT) or magnetic resonance imaging (MRI) therefore, imaging plays a key role in both HCC diagnosis and treatment [3]. The updated Liver Imaging Reporting and Data System (LI-RADS) has established five major features for accurate HCC detection including enhancement pattern, growth rate, and size [4].

Hepatic resection proposes the highest chance of best survival for HCC patients. In the past decade, hepatic surgery has seen expanded indications, more intricate surgical techniques, and improved operative outcomes [5].

Vascular invasion, gross or microvascular, is a poor predictive factor in HCC [6]. Microvascular invasion (MVI) is the existence of micro-emboli of HCC deposits within liver vessels and is a critical risk factor for early recurrence occurring following curative treatment [7]. While it is feasible to detect gross vascular invasion through imaging studies, identifying MVI is often challenging and requires pathological assessment of surgical specimen [8, 9].

Previous imaging characteristics have been suggested as predictors of MVI, involving large tumor size, multifocality, irregular tumor margin or non-capsulation, and peritumoral parenchymal enhancement [10,11,12]. The efficacy of triphasic CT in preoperative detection of MVI has been introduced in limited studies with variable results [13,14,15,16] and still needs further validation. This study aimed to evaluate the inter-observer reliability and predictive values of triphasic CT imaging in detecting MVI in patients with HCC in correlation with postoperative pathological data.

Patients and methodology

Study population

This retrospective study has been granted by the local institutional review board with a corresponding waiver of informed consent. From January 2021 to March 2023, one hundred thirty patients with HCC underwent surgical resection and were initially enrolled. Thirty-one patients were excluded from the study. Of those, 28 had preoperative CT conducted outside our radiology unit and the other three had prior locoregional treatment. Finally, the study consisted of ninety-nine patients with HCC who underwent preoperative CT and laboratory assessment.

Clinical and laboratory data

Laboratory and clinical data were gathered for each patient. The clinical data included their age, sex, associated comorbidities, prior antiviral therapy, and the type of surgical resection (minor or major). The laboratory data included measurements of alpha-fetoprotein (AFP), serum creatinine (Cr), liver function tests (LFTs), complete blood count (CBC), and the international normalized ratio (INR).

Triphasic CT technique

All patients underwent a triphasic CT study on a 128 multidetector CT scanner (Toshiba Medical Systems Aquilion Prime ® scanner) within 2–3 weeks before the surgical resection. Patients were imaged in a craniocaudal direction with the following CT scan acquisition parameters: 120 KVp, 200–440 mAs, pitch of 0.8:1, slice thickness of 2 mm, reconstruction gap 1 mm, a field of view of 380 mm, matrix 512 × 512, and window width 150.

Non-ionic contrast medium (Omnipaque 350) was injected at a total dose of 100–120 mL with a flow rate of 3 mL/s. Triphasic CT includes three post-contrast phases: arterial phase acquired within 25–30 s after IV contrast agent injection, portal venous phase (PVP), and equilibrium phase (EP) acquired within 55 s and 3 min after IV contrast injection, respectively. Images were reconstructed utilizing a standard reconstruction algorithm in sagittal, and coronal reformats.

CT image analysis

CT image analysis was done by two independent radiologists (13 and 11 years of experience in hepatic imaging), and both radiologists were blinded to the postoperative pathological data about MVI.

Each observer separately revised CT studies to assess the following criteria for MVI: (a) The largest diameter; (b) The number of lesions; (c) The tumor margins categorized as smooth margins and non-smooth margins showing nodular tumors with smooth outline and nonnodular tumors in all imaging planes, respectively [10, 16]. The non-smooth margins were also classified into a single nodule with extranodular extension (SN EN) showing focal or crescent-like extension of nodules bulging to the hepatic parenchyma, a multinodular confluence (MN) showing multifocal outgrowths bulging into the parenchyma and infiltrative margin (d) Tumor capsule, defined as linear peripheral smooth hyperenhancement in PVP or EP and categorized as capsule complete, capsule incomplete, or no capsule [12]. (e) Peritumoral enhancement [16] is an obvious enhanced arterial part next to the tumor border, which afterward becomes isodense with the hepatic parenchyma in EP. If peritumoral enhancement is present, it is described as either irregular circumferential or wedge-shaped enhancement. (f) Intratumoral arteries were detected at the arterial phase. (g) Hypodense halo sign [13] defined as a hypoattenuating peritumoral ring of hepatic parenchyma in PVP or EP. (h) Two-trait predictor of venous invasion (TTPVI) [16] known as the detection of distinct two CT characteristics (the existence of intratumoral arteries and noncontinuous or absent hypodense halos) that can better predict MVI.(i) Hepatic capsular invasion [17] was detected by the existence of any of the following features: absent or incomplete tumor capsule; absence of hepatic parenchyma between the liver capsule and tumor, tumor growth outside the liver border or; interrupted hepatic capsule or noticeable subcapsular effusion accumulation. CT images were also assessed for macrovascular invasion, bile duct invasion, and intratumoral fat presence.

Surgical technique

All included patients underwent a preoperative 3D visualization analysis, liver volumetry was achieved for patients who required major liver resection [18]. The surgical resection was either minor or major according to Brisbane 2000 classification [19]. Major resections were carried out for big tumors or tumors adjacent to major liver vasculature, providing an adequate remnant liver [20].

Histopathological analysis

An expert pathologist with 15 years of expertise in hepatic pathology assessed the surgical specimens to detect MVI. The practical guidelines for the tissue diagnosis of primary liver cancer of Liver Cancer Pathology Group of China) (LCPGC) serve as the foundation for MVI diagnosis. Presences of nests of tumor cells in the hepatic vein, portal vein, and any other vessel lined by endothelium within the tumor capsule that were only discernible under a microscope were defined as MVI [21]. All specimens were allotted into positive MVI and negative MVI groups according to the MVI bi-tiered grading system.

Follow-up

Follow-up was directed one month following surgical resection and then every 3 months consuming laboratory (CBC, serum AFP, and LFTs) and radiological findings (abdominal ultrasonography and triphasic CT or MRI). The recurrence was diagnosed based on a newly developed hepatic focal lesion on triphasic CT and elevated AFP levels.

Statistical analysis

Data were evaluated with the Statistical Package of Social Science (SPSS) program for Windows (Standard version 24). First, the data was tested for normality with a one-sample Kolmogorov–Smirnov test. Numbers and percentages were used to describe qualitative data. The association among categorical variables was tested using the Chi-square test however Monte Carlo and Fisher exact tests were performed when the assumed cell counts less than 5. The continuous variables were expressed as median (Min–Max) for non-normal data and mean ± SD (standard deviation) for normally distributed data. The two groups were compared by independent t-test (parametric data) and Mann–Whitney test (nonparametric data). The agreement between the two observers was tested using the Kappa agreement. Significant variables identified on univariate analysis were participated in the logistic regression model consuming the forward wald statistical technique to expect the most important factors. The results were counted as significant when the p ≤ 0.05 for all statistical tests.

Results

Patients’ characteristics

The study group involved 99 patients with a mean age of 62.00 ± 9.41 (73 males, and 26 females). All patients had cirrhotic liver and were classified as Child–Pugh A. The postoperative histopathological results showed that 58 patients (58.6%) tested positive for MVI, while 41 patients (41.4%) tested negative for MVI (Table 1).

Table 1 Patients’ characteristics

Sixty-eight (86.9%) patients had received anti-HCV or HBV therapy with statistically significant association (p = 0.041) with negative MVI in 39 (95.1%) patients. Serum creatinine and alkaline phosphatase were statistically significantly higher in the MVI-negative than MVI-positive group (P = 0.018 and P = 0.009) respectively. There was no statistically significant association concerning other liver function tests, AFP, and MVI status of HCCs. Sixty-eight (68.7%) patients had no radiological recurrence in the first year following the operation with statistically significant association (p = 0.033) with MVI negative group in 33/41 (80.5%) patients.

Triphasic CT findings

Our results showed a statistically significant association between tumor margin (p = 0.001), tumor capsule (p ≤ 0.001), peritumoral enhancement (p = 0.037), intratumoral arteries (p = 0.008), hypodense halo (p ≤ 0.001), TTPV (p ≤ 0.001) and macrovascular invasion (p = 0.018) and MVI. Tumor margin was multinodular in 28 (48.3%) & infiltrative in 17 (29.3%) patients of 58 positive MVI group, while margin was smooth in 22 (53.7%) patients of 41 negative MVI group.

Twenty-six (44.8%) patients of 58 positive MVI had incomplete capsules, and 19 (32.8%) patients had absent capsules. On the other side, of 41 negative MVI patients, 26 (63.4%) patients had complete capsule. Peritumoral enhancement was wedge-shaped in 13 patients (13.1%); 11 of them were positive MVI in histopathological specimens. Irregular circumferential enhancement was found in 6 (6.1%) patients; five of them also had positive MVI. While peritumoral enhancement was absent in 80 (80.8%) patients, 38 patients (92.7%) of the MVI-negative group had absent peritumoral enhancement.

In 23.2% (23/99) of the patients, intratumoral arteries were absent, and 36.6% of these patients (15/23) showed negative MVI in histopathological reports. Hypodense halo was found in 18 (18.2%) patients, 17 of them showed negative MVI and was absent in 81(81.8%) patients, 57 of them were positive MVI in histopathological reports. TTPV was present in 66 (66.7%) patients, 47 of them showed positive MVI, while it was absent in 33 (33.3%) patients, 22 of them were negative MVI. Of 58 positive MVI patients 50 (86.2%) patients showed intratumoral arteries and 57 (98.3%) patients showed absent hypodense halo. Macrovascular invasion was identified in 15 (15.2%) patients, 13 of them showed positive MVI in histopathological specimens. The size of HCCs in all patients ranged from 2 to 20 cm. There was no significant association between other assessed triphasic CT parameters (Table 2). Demonstrative cases are illustrated at Figs. 1, 2, 3, 4.

Table 2 CT imaging features of MVI

Fig. 1

38-year-old female with solitary hepatocellular carcinoma (HCC) occupying segment VIII of the liver who underwent minor resection. A Axial non-contrast CT image shows a hypodense focal lesion at segment VIII. B Axial arterial phase CT image shows heterogeneous enhancement of the lesion with intratumoral arteries (arrow). C Axial equilibrium phase CT image shows washout of contrast and the lesion shows smooth margins, incomplete capsule (arrow), but no hypodense halo. TTPVI of positive intratumoral arteries and negative hypodense halo is indicative of positive MVI. D Histopathologic examination revealed frequent microvascular emboli in the portal tract (yellow arrow), magnification × 100

Fig. 2

A 73-year-old man with solitary hepatocellular carcinoma (HCC) at segment VI of the liver underwent minor resection. A Axial non-contrast CT image shows a hypodense focal lesion at segment VI. B Axial arterial phase CT image shows heterogeneous enhancement of the lesion with intratumoral arteries (arrow). C Axial equilibrium phase CT image shows washout of contrast and the lesion shows micronodular margins, incomplete capsule (arrow), but no hypodense halo. TTPVI of positive intratumoral arteries and negative hypodense halo is indicative of positive MVI. D Histopathologic examination revealed frequent microvascular emboli with protruding tumor cells into vessel lumen (yellow arrow), magnification × 100

Fig. 3

51-year-old female with solitary hepatocellular carcinoma (HCC) at segment VI of the liver who underwent minor resection. A Axial non-contrast CT image shows hypodense focal lesion at segment VI. B Axial arterial phase CT image shows heterogeneous enhancement of the lesion with intratumoral arteries (arrowhead). C, Axial PVP CT image shows evident hypodense halo (arrows). TTPVI of positive both intratumoral arteries and hypodense halo is indicative of absent MVI. D Axial equilibrium phase CT image shows washout of the contrast and the lesion shows smooth margins, and a complete capsule (arrow). E, F Histopathologic examination failed to detect any microvascular emboli neither in the capsule (E) nor in the portal tract (F), magnification × 100

Fig. 4

59-year-old female with solitary hepatocellular carcinoma (HCC) at segment VI of the liver who underwent minor resection. A Axial non-contrast CT image shows hypodense focal lesion at segment VI. B Axial arterial phase CT image shows heterogeneous enhancement of the lesion with intratumoral arteries (arrow). C Axial PVP CT image at different levels shows evident incomplete hypodense halo (arrows). D axial equilibrium phase CT image shows washout of contrast and the lesion shows smooth margins and complete capsule. E, F Histopathologic examination failed to detect any microvascular emboli neither in the capsule (E) nor in the portal tract (F), magnification × 100

Predictors for malignancy

Univariate analysis results were infiltrative margin of tumor (P = 0.013, OR 5.25), incomplete or absent tumor capsule (P = 0.002 and ≤ 0.001 respectively, OR 4.73 and 9.5 respectively), wedge-shaped peritumoral enhancement (P = 0.045, OR 4.97), presence of intratumoral arteries (P = 0.01, OR 3.6), absence of hypodense halo (P ≤ 0.001, OR 40.4), presence of TTPV (P ≤ 0.001, OR 4.9) and macrovascular invasion (P = 0.029, OR 5.6) were independent predictors for positive MVI. After multivariate analysis, the most significant predictor factors for positive MVI were incomplete and absent tumor capsule (P = 0.053 and 0.027 respectively), absence of hypodense halo (P = 0.002), and the existence of TTPV (P = 0.017) (Table 3).

Table 3 Multivariate logistic regression analysis for independent predictors of MVI

Inter-observer agreement

An excellent inter-observer agreement was detected in the evaluation of peritumoral enhancement, identification of intratumoral arteries, hypodense halo, TTPV, and macrovascular invasion. Furthermore, there was moderate inter-observer agreement regarding the tumor margin and tumor capsule (Table 4).

Table 4 Inter-observer agreement

Discussion

Recently MVI has aroused worldwide attention as a chief risk factor for poor survival and recurrence in patients with HCCs. It means early invasion of tumor vessels by tumor cells [22]. Detection of MVI in imaging studies is generally challenging [11]. So, we targeted to study the reliability and performance of some proposed CT features to MVI.

Our results revealed a significant association between Anti HBV and HCV therapy and the negativity of MVI in our results. This is explained by the effect of antiviral therapy in controlling liver inflammation; one of the primary contributors to develop HCC. Rationally it may affect the aggressiveness of the tumor. So, it can decrease the development of HCC as well as the post-surgical recurrence [23]. Chou et al. [10] showed no significant relation between chronic viral hepatitis and microvascular invasion, despite that, patients with chronic viral hepatitis were higher in number in the MVI-positive group compared to the MVI-negative group.

In this study, the non-smooth margin of the tumor was correlated with positive MVI, it has been further sub-classified into single nodule with extra nodular extension, multinodular margin & infiltrative margin with increasing incidence of MVI, respectively. That was in concordance with prior studies [10] [13, 16, 17]. Zhang et al. [17] reported that nodular borders in were gorgeous in blood vessels to predict MVI in a detailed evaluation of margin pathology. It has been previously reported by Chou et al. [10] that site of MVI occurs frequently at the location of extranodular involvement.

In the present study, the infiltrative margin has been added and had the highest incidence of MVI. Similarly, Renzulli et al. [16] determined that the infiltrative margin was the most pattern closely associated with MVI. It has been suggested that treatment of HCC having non-smooth margins could be remodeled by either extensive surgical resection or postoperative treatment to reach a complete response or elongated tumor-free survival [13].

Incomplete or absent tumor capsule is significantly correlated with MVI in the current study. Though, the relationship between tumor capsule and MVI stays debatable in previous studies [13]. Histologically, a fibrous capsule is comprised of two layers; inner fibrous-rich layer and an outer layer rich in small vessels and a newly formed bile duct [12]. Our results were in line with some authors who considered the fibrous capsule to be a favorable predictive factor since it precludes invasion of normal surrounding liver parenchyma [13, 24]. Also, Ariizumi et al. [25] showed a significant association between the incomplete capsule and MVI. On the contrary, Witjes et al. and Adachi et al. [26, 27] considered tumor capsule is predictive for vascular invasion, explained by the frequent invasion of the blood vessels within the fibrous capsule by tumor cells.

Irregular circumferential or wedge-shaped peritumoral enhancement was significantly related to MVI in this study. This is mostly attributed to peritumoral hemodynamic changes in the form of arterial hyperperfusion in areas with reduced portal flow secondary to minute portal branch occlusion by tumor thrombi [28]. This was in keeping with previous studies as Kim et al., and Miyata et al. [29, 30] using different imaging modalities and different CT angiographic technique.

Irregular circumferential peritumoral enhancement was significantly associated with MVI rather than wedge-shaped pattern in studies done by Yoneda et al. [12] and Nishie et al. [31] who stated that the size of peritumoral enhancement was found to be a significant risk factor for MVI by CT arterioportography, while Renzulli et al. [16] suggested assessment of peritumoral enhancement by noninvasive, commonly used imaging methods, such as CT or dynamic MRI.

The association between imaging features and HCC-precise “venous invasion” gene mark was named radiogenomic venous invasion in some reports or two-trait predictor of venous invasion (TTPVI) in others. It depends on the detection of intratumoral arteries and peritumoral hypodense halos in CT studies. It was suggested to have a strong association with histologic MVI [11, 16, 32].

Our results showed a significant association between the presence of TTPVI and MVI positivity in pathological reports. That was in agreement with Renzulli et al. [16] who considered it as one of three significant features to expect MVI on both CT and MRI with high positive predictive values (PPV) > 95%. Nevertheless, the identification of hypodense halo and internal arteries are challenging in small HCCs [11].

Macrovascular invasion reduces the therapeutic alternatives and signifies a contraindication to liver transplantation. While macrovascular invasion often coexists with lesions with an infiltrative appearance that refers to the permeative growth pattern throughout cirrhotic parenchyma, unsurprisingly, it should be associated with MVI as resulted in the present study [33].

In the current study, there was no significant association between MVI status and hepatic capsular invasion. This disagrees with Zhang et al. [17] who showed an association between this CT feature and MVI. This may be explained by common but less specific radiological criteria that were used to determine hepatic capsular invasion by authors.

Also, our results revealed no significant association concerning the enhancement pattern, presence of intratumoral fat, or bile duct invasion with MVI. This may be attributed to selection bias of patients that had already undergone liver resection.

Serum AFP is one of the most frequently used tumor markers for HCC, current study showed higher AFP in the MVI-positive compared to the MVI-negative group but with no detected significant difference between both studied groups. Eguchi et al. [34] stated that the AFP level could be employed as a predictor of MVI and early recurrence. Zhang et al. [17] used AFP and tumor size (> or < 5 cm) in stratified analysis of HCC for the prediction of MVI. However, our results were in line with Chou et al. [10]. This discrepancy in results could be exerted by possibility of the presence of non-AFP-secreting tumors.

Univariate analysis of this study revealed that infiltrative tumor margin, incomplete or absent tumor capsule, wedge-shaped peritumoral enhancement, existence of intratumoral arteries, lack of hypodense halo, presence of TTPV, and macrovascular invasion were independent predictors for MVI. Despite of that, multivariate analysis showed that incomplete and absent tumor capsule, the absence of hypodense halo, and the presence of TTPV are the most strong independent predictors for MVI.

Renzulli et al. [16] described TTPV involving the absence of hypodense halo is one of three worrisome features with a high predictive value for MVI. Other two features included non-smooth margin and peritumoral enhancement. Ueda et al. [35] reported that intact or complete capsule can avoid tumor spread.

Non-smooth or irregular margin was reported in prior studies [10, 13, 16] as a strong independent risk for MVI. Chou et al. [10] reported that a non-smooth margin was the single significant predictor for MVI in their multivariate analysis. The difference in results between this study and others may be attributed to tumor size that was non-significantly different between our studied groups. It was difficult to assess tumor margin in small liver tumors.

Similar to Renzulli et al. [16], the current study showed excellent inter-observer agreement in the assessment of peritumoral enhancement, identification of intratumoral arteries, hypodense halo, and TTPV. While moderate agreement was detected during assessment of tumor margin and tumor capsule. This also could be explained by variability in tumor size as; tumor margin and capsule are easier to be assessed in large tumors than in small ones.

This study had a few limitations. Firstly, the selection bias of surgical candidates besides retrospective design may cause inadequate expression of HCC radiologic characteristics, a prospective study is recommended in the future. Another limitation is the lack of incorporation of histologic features, e.g., grading of liver fibrosis. Thirdly, lack of adding of MRI examination, so, further combining CT and MRI in the future may be beneficial. Additionally, further quantitative or textural image analysis is recommended, may impact study results.

Conclusion

Incomplete or absent tumor capsule, presence of TTPV, and absence of hypodense halo are independent radiological predictors for MVI that may be helpful for standard preoperative interpretation of HCC.