MRI Features for Predicting Microvascular Invasion and Postoperative Recurrence in Hepatocellular Carcinoma Without Peritumoral Hypointensity

Introduction

Hepatocellular carcinoma (HCC) is the fifth most common malignancy and the third leading cause of cancer-related deaths worldwide, with a relative 5-year survival rate of approximately 18%.^1,2 China has a high burden of liver cancer, with 45.3% of liver cancer cases and 47.1% of liver cancer deaths globally.³ Radical resection is the main curative treatment for HCC. However, more than half of the patients treated with R0 resection will develop recurrence,⁴ which has a strong impact on prognosis. Microvascular invasion (MVI), defined as the presence of clusters of tumor cells in vessels located in the peritumoral area of the liver,⁵ is a major risk factor for HCC recurrence after surgery. However, MVI can only be detected by microscopic examination of the surgical specimen, which limits its application, particularly in terms of preoperative therapeutic decision-making and determination of resection margins during hepatectomy. Therefore, it is of great value to explore noninvasive methods that can determine MVI status preoperatively.

Several noninvasive preoperative methods, including radiomics, deep learning, and gene expression signatures, have been used to predict MVI status and early tumor recurrence. Gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA)-enhanced MRI is widely used for the diagnosis and assessment of HCC. Assessment of tumor features on Gd-EOB-DTPA-enhanced MRI is easy to perform and does not require additional devices or expenses. Owing to the lack of organic anion-transporting polypeptide (OATP) transporter expression, most HCCs present hypointense signal compared to the surrounding liver parenchyma on hepatobiliary phase (HBP) imaging. In MVI-positive patients, tumor cells in the vessels can result in dysfunction of the OATP transporter in hepatocytes located in the peritumoral liver parenchyma, leading to peritumoral HBP hypointensity. Although peritumoral HBP hypointensity has been identified as an important imaging feature with high specificity for predicting MVI, its sensitivity is relatively poor and was found to be only 44.2% in a recent meta-analysis.^6,7 This finding indicates that more than half of patients with MVI do not show peritumoral HBP hypointensity. Given the importance of MVI in HCC management, it is imperative to identify complementary features that can predict MVI and tumor recurrence in patients without peritumoral HBP hypointensity. To the best of our knowledge, few studies have focused on this topic or its population.

Therefore, this study aimed to identify the MRI features of HCC that predict MVI and postoperative recurrence in patients without peritumoral HBP hypointensity.

Materials and Methods

Patients

This retrospective study was compliant with the ethical guidelines of the 1975 Declaration of Helsinki and approved by the ethics committee of the Second Affiliated Hospital of Guangzhou University of Chinese Medicine. The requirement for written informed consent was waived by the Institutional Review Board owing to the retrospective observational nature of the study, and all patient data are treated confidentially. The data of treatment-naive patients who underwent preoperative liver Gd-EOB-DTPA-enhanced MRI and curative hepatic resection for HCC at our institution between December 2017 and December 2021 were reviewed. The inclusion criteria were age >18 years and presence of a single histopathologically confirmed HCC lesion. The following exclusion criteria were applied: (1) more than a 2-month interval between Gd-EOB-DTPA-enhanced MRI and surgery, (2) suboptimal MRI quality, (3) pathologically confirmed macroscopic vascular invasion, (4) extrahepatic spread on preoperative work-up, (5) incomplete clinical and laboratory data, (6) lack of follow-up data, and (7) history of preoperative HCC-related treatment (chemotherapy, radiotherapy, transarterial chemoembolization, radiofrequency ablation, or immunosuppressive therapy).

MRI Protocol

Patients underwent upper abdominal MRI using two 3.0-T MRI vectors with an 8-channel Atlas Speeder phased-array coil (Titan; Toshiba Medical Systems, Tochigi, Japan) or 18-channel phased-array coil (Prisma; Siemens Healthcare, Erlangen, Germany). All patients fasted for at least 8 h before MRI examination. Axial T1-weighted, T2-weighted, and dynamic contrast-enhanced images were obtained in all the cases. Each patient received 0.1 mL/kg of gadoxetic acid disodium (Primovist; Bayer Vital GmbH, Berlin, Germany) at a flow rate of 2 mL/s to acquire Gd-EOB-DTPA-enhanced images. Arterial phase, portal phase, and HBP images were obtained 15 s, 60–70 s, and 20 min after contrast injection, respectively. The acquisition parameters are listed in Table S1.

Interpretation of Images

Two board-certified radiologists (X.H.L. and Y.M.Y., with 5 and 10 years of experience in liver MRI, respectively), who were blinded to the histopathological results, laboratory data, and follow-up information, independently reviewed all preoperative MRI scans. They were required to count the number, measure the longest diameter of each target observation, and classify the imaging features of each HCC lesion. Two weeks before the formal assessment, they were required to discuss the presentation of each MRI feature, and were trained in 20 cases. The definition of imaging features on Gd-EOB-DTPA-enhanced MRI has been described in other studies and is documented in Table S2.^8–10 The interobserver agreement was assessed after independent image review. In the event of disagreement between the two radiologists, a senior radiologist (Z.Y.C., with 15 years of experience in liver MRI) reviewed the images and discussed them with the other two radiologists, and a consensus was reached by a majority vote.

Clinical and Laboratory Data

Clinical (age, sex, etiology of chronic liver disease) and laboratory data (white blood cells [WBC], red blood cells [RBC], hemoglobin [HB], platelets [PLT], prothrombin time [PT], international normalized ratio of prothrombin [INR], activated partial thromboplastin time [APTT], fibrinogen [FIB], alanine aminotransferase [ALT], aspartate transaminase [AST], albumin [ALB], globulin [GLB], alkaline phosphatase [ALP], gamma-glutamyltransferase [GGT], total bilirubin [TBIL], creatinine [Cr]), and serum alpha-fetoprotein [AFP]) within 1 week before hepatectomy were reviewed. The Child-Turcotte-Pugh (CTP) stage and Model for End-Stage Liver Disease (MELD) score were calculated for each patient according to the American Association for the Study of Liver Disease (AASLD) guidelines.¹¹

Histological Evaluation

After surgical resection, formalin-fixed paraffin-embedded specimens were stained with hematoxylin and eosin and evaluated by a senior pathologist with 15 years of experience in liver pathology, according to the seven-point baseline sampling protocol. Histopathological reports included tumor number, tumor size, Edmondson grade, MVI status, and final diagnosis. MVI was defined as the presence of tumor cells in the endothelium-lined vascular lumen observed only by microscopy.

Follow-Up After Surgical Resection

All patients underwent clinical and radiological follow-up at 1 month after hepatectomy, at 3-month intervals for the first 2 years, and at 6-month intervals for up to 5 years thereafter. Radiological follow-up consisted of ultrasonography, dynamic-enhanced computed tomography, or Gd-EOB-DTPA-enhanced MRI. When recurrence was suspected clinically or radiologically, appropriate management was implemented according to our institutional protocol and the AASLD guidelines. Patients were censored in the event of local recurrence, loss to follow-up, or on March 1, 2023, whichever event occurred first. Intrahepatic recurrence-free survival (RFS) was defined as the time interval between hepatectomy and intrahepatic recurrence.

Statistical Analysis

Continuous variables were reported as the mean ± standard deviation or median (interquartile range [IQR]) and were compared between groups using Student’s t-test or Mann–Whitney U-test, as appropriate. Categorical variables were reported as frequency (percentage) and compared between groups using the chi-square test or Fisher’s exact test, as appropriate. Interobserver agreement regarding MRI features was assessed using the Cohen’s kappa (κ) value and graded based on the Landis and Koch scale (0.00, poor agreement; 0.01–0.20, slight agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and >0.80, nearly perfect agreement).¹²

The relationship between radiological variables and MVI status was assessed using univariate logistic regression analysis. Variables with a P-value <0.05 in univariate analysis were entered into multivariate logistic regression analysis with stepwise forward selection to identify independent risk factors. Prediction performance was assessed by measuring the area under the receiver operating characteristic curve (AUROC), odds ratio (OR) with 95% confidence interval (CI), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Subgroup analysis was performed for patients without peritumoral HBP hypointensity, using the same methods.

Postoperative intrahepatic RFS was estimated using the Kaplan–Meier method, and survival curves were compared using the Log rank test. The prognostic value of MRI features for postoperative intrahepatic RFS in the subgroup of patients without peritumoral HBP hypointensity was evaluated using univariate and multivariate Cox proportional hazards regression analyses and quantified using the hazard ratio (HR).

All statistical analyses were performed using SPSS (version 22.6; IBM Corp., Armonk, NY, USA) and MedCalc (version 16.8.4; Ostend, Belgium). Statistical significance was defined as a two-sided P-value <0.05.

Results

Patients

One hundred and thirty of 315 patients who underwent curative resection for HCC during the study period were enrolled (Figure 1). The median age was 55.0 years (IQR 47.0–63.0), 114 patients (87.7%) were male, the most common etiology of liver disease was chronic hepatitis B (84.6%), and 63 patients (48.5%) were MVI-positive. Edmondson grade III or IV was more common in the MVI-positive group than in the MVI-negative group. There were no other significant differences in the clinical or pathological characteristics between the two groups (Table 1).

Table 1 Clinical and Pathological Characteristics of MVI-Positive and MVI-Negative Cohorts

Figure 1 Flowchart showing the patient inclusion and exclusion criteria.

Interobserver Agreement Regarding MRI Features

The two radiologists showed moderate to excellent inter-observer agreement for the MRI features of HCC, with κ values in the range of 0.431–0.885 (Table 2). The best interobserver agreement was observed for satellite nodule (κ=0.885; 95% CI: 0.796–0.974). Moderate consistency was observed for peritumoral HBP hypointensity (κ=0.511; 95% CI: 0.339–0.682).

Table 2 The Interobserver Agreement of MRI Features

MRI Features of HCC That Identify MVI Status

In univariate analysis, corona enhancement (P=0.021) and peritumoral HBP hypointensity (P=0.021) were more common in the MVI-positive group than in the MVI-negative group (Table 3). In the multivariate analysis, corona enhancement (OR,3.019; 95% CI: 1.071–8.512; P=0.037) and peritumoral HBP hypointensity (OR, 3.019; 95% CI: 1.071–8.512; P=0.037) were independent predictors of MVI-positive status. The sensitivity, specificity, PPV, NPV, and AUROC of these two features for identifying MVI were 23.80%, 91.04%, 71.23%, 55.96%, and 0.574, respectively (representative images: Figure 2a and b).

Table 3 MRI Features of MVI-Positive and MVI-Negative HCCs

Figure 2 Representative MRI features associated with histopathological findings. HBP images (a) for a 29-year-old male patient showing peritumoral HBP hypointensity (arrow) and satellite nodule (arrowhead). Hepatocellular carcinoma with MVI-positive was confirmed by histopathology (b). HBP images (c) for a 41-year-old male patient showing a radiological capsule (arrow) without peritumoral HBP hypointensity. Hepatocellular carcinoma with MVI-positive was confirmed by histopathology (d). Abbreviations: HBP, hepatobiliary phase; MVI, microvascular invasion.

Characteristics of Patients Without Peritumoral HBP Hypointensity

The median age of the 109 patients without peritumoral HBP hypointensity was 56.0 years (IQR 49.0–64.0) and 94 (86.2%) were male. As in the entire study cohort, there were no significant differences in any of the clinical or pathological characteristics between the MVI-positive and MVI-negative groups in patients without peritumoral HBP hypointensity, except for Edmondson grades III and IV (Table S3).

In patients without peritumoral HBP hypointensity, the only significant difference in MRI features between the MVI-positive and MVI-negative groups was the presence of radiological capsule (Table 4). The sensitivity, specificity, PPV, NPV, and AUROC of the radiological capsule for identifying MVI-positive HCC in this subgroup were 89.58%, 26.23%, 48.86%, 76.19%, and 0.579, respectively (representative images: Figure 2c and d).

Table 4 MRI Features of MVI-Negative and MVI-Positive HCCs in Subgroup Without Peritumoral HBP Hypointensity

MRI Features for Predicting Intrahepatic Recurrence

During a median follow-up period of 11 months (range, 1–56 months; IQR, 4–20 months), intrahepatic recurrence was diagnosed in 51 patients (39.3%). The median time to intrahepatic recurrence after curative hepatic resection was 7 months (range, 1–47 months; IQR, 3–13 months). Kaplan–Meier curves showed that intrahepatic RFS was shorter in patients with peritumoral HBP hypointensity than in those without (Figure 3, P<0.001). After adjusting potential confounders of age, gender, tumor size, and other radiological features, peritumoral HBP hypointensity (HR: 2.381, 95% CI: 1.162–4.881, P=0.018) remained significantly associated with shorter intrahepatic RFS (Table S4).

Figure 3 Kaplan–Meier curves showing recurrence-free survival in patients with and without peritumoral hepatobiliary phase hypointensity.

In patients without peritumoral HBP hypointensity, subgroup analysis showed that tumor size >5 cm (P=0.001), non-smooth margins (P=0.040), mosaic sign (P=0.040), necrosis (P=0.009), and satellite nodule (P<0.001) were prognostic factors for intrahepatic RFS. However, only satellite nodule (HR: 3.324, 95% CI: 1.733–6.378, P<0.001) was identified as an independent prognostic factor for intrahepatic RFS in multivariate Cox regression analysis (Table 5).

Table 5 Univariate and Multivariate Cox Regression of MRI Features in Predicting Intrahepatic Recurrence in Subgroup Without Peritumoral HBP Hypointensity

As shown in Figure 4, intrahepatic RFS was similar in patients without peritumoral HBP hypointensity who had satellite nodule to that in patients with peritumoral HBP hypointensity (P=0.693). However, intrahepatic RFS was significantly shorter in both cohorts than that in patients without either of these MRI features (P<0.001) (representative images: Figure 5).

Figure 4 Kaplan–Meier curves showing recurrence-free survival in patients without peritumoral HBP hypointensity (PHBPH) and satellite nodule (SN), with SN but without PHBPH and with PHBPH.

Figure 5 Representative MRI features associated with recurrence-free survival. HBP images (a) for a 60-year-old female patient showing an irregularly hypointense tumor (arrow) but without peritumoral HBP hypointensity or a satellite nodule. Tumor recurrence has not been found after 28 months of follow-up. HBP images (b) for a 50-year-old male patient showing multiple satellite nodules (arrow) around the tumor without peritumoral HBP hypointensity. Tumor recurrence was detected 4 months after curative resection.

Two Stepwise Flowchart for Prediction of Intrahepatic Recurrence

To facilitate clinical use and improve performance, a two-step flowchart incorporating peritumoral HBP hypointensity and satellite nodule was developed to stratify intrahepatic recurrence risk (Figure 6). Patients without peritumoral HBP hypointensity and satellite nodule had a low risk of intrahepatic recurrence, whereas those at a high risk (47/130, 36.2%) had an increased risk of intrahepatic recurrence. Using the two-step flowchart, 26 (20%) of 130 patients were reassigned as high-risk for intrahepatic recurrence, which was a significant improvement when compared with peritumoral HBP hypointensity alone (P<0.001).

Figure 6 Two stepwise flowchart incorporating peritumoral hepatobiliary phase hypointensity and satellite nodule for stratification of the risk of intrahepatic recurrence. Recurrence-free survival was significantly shorter in patients with high-risk hepatocellular carcinoma than in those with low-risk hepatocellular carcinoma (hazard ratio 3.511, 95% confidence interval 1.995–6.181, P<0.001).

Discussion

The findings of this study confirm that peritumoral HBP hypointensity is an important MRI feature that can be used to preoperatively predict pathological MVI and shorter intrahepatic RFS and that the presence of a radiological capsule can help to predict MVI status in patients without peritumoral HBP hypointensity. However, only the presence of satellite nodule was identified as an independent risk factor for intrahepatic RFS in this subgroup after univariate and multivariate Cox regression analyses.

Previous studies have established that the presence of MVI is one of the key factors in early recurrence of HCC and poor outcomes after curative liver resection or liver transplantation.¹³ Preoperative assessment of MVI is important for accurate selection of candidates for the various treatment options. For example, studies in which MVI status was evaluated preoperatively found that the rate of early recurrence was lower after surgical resection than after radiofrequency ablation and that anatomical resection was superior to nonanatomical resection.^14–17 Therefore, the development of noninvasive methods for preoperative prediction of MVI status has become an important research topic in recent years. The features of HCC observed on Gd-EOB-DTPA-enhanced MRI were confirmed to be associated with MVI status. Our finding that corona enhancement and peritumoral HBP hypointensity are independent predictors of MVI is consistent with reports by Lee et al and Zhang et al^15,18 For MVI-positive HCCs, local tumor invasion or infiltration of tumor cells into vessels could impair the function of peritumoral hepatocytes, which would result in impaired OATP transporter expression and decreased uptake of Gd-EOB-DTPA in the peritumoral region, that is, peritumoral HBP hypointensity.¹⁹ However, when infiltration of tumor cells in vessels is limited, impairment of OATP transporter function is also limited, as is the likelihood of peritumoral HBP hypointensity. Furthermore, HCCs with coactivation of Wnt/β-catenin signaling and hepatocyte nuclear factor 4α could induce higher OATP expression and manifest as peritumoral HBP hyperenhancement.²⁰ Therefore, peritumoral HBP hypointensity would be a good indicator for ruling in MVI rather than ruling it out. Our finding that the sensitivity of peritumoral HBP hypointensity in identifying MVI was quite low (23.80%) supports this theory. Almost three-quarters of the MVI-positive HCCs were missed when using this feature alone. Therefore, there is a need to explore other MRI features to identify MVI-positive HCC in patients without peritumoral HBP hypointensity.

The presence of a radiological capsule is a MRI characteristic that has been controversial in terms of predicting MVI status. Some studies have found that it is a valuable predictor of MVI-positive HCC, while others have not.^21–24 Interestingly, in our study, a radiological capsule could predict MVI in patients without peritumoral HBP hypointensity, but not in the overall study cohort. A possible explanation for this finding is that peritumoral HBP hypointensity was significantly more common in patients with MVI-positive HCC than in MVI-negative HCC in those patients without a radiological capsule (50.0% vs 5.8%, P=0.015). After removal of patients with peritumoral HBP hypointensity, there was a marked decrease in the proportion of patients with MVI-positive HCC without a radiological capsule. Therefore, the absence of a tumor capsule in this subgroup strongly indicates MVI-negative status and could be used as a rule-out indicator. This is consistent with our finding that the presence of a radiological capsule has a relatively high NPV for predicting MVI-positive HCC in patients without peritumoral HBP hypointensity.

Peritumoral HBP hypointensity in HCC is not only associated with an increased risk of MVI but is also an established adverse prognostic imaging feature for early tumor recurrence after curative hepatic resection and liver transplantation.^19,25 In our study, intrahepatic RFS was significantly shorter in patients with peritumoral HBP hypointensity than in patients without this feature. However, given the relatively low frequency of peritumoral HBP hypointensity, it is necessary to explore other supplementary MRI features that could predict postoperative recurrence. Our study demonstrated that satellite nodule was an independent risk factor for intrahepatic RFS in patients without peritumoral HBP hypointensity. To the best of our knowledge, this is the first study to explore the risk of tumor recurrence in patients without peritumoral HBP hypointensity. The effect of satellite nodule on tumor recurrence has been demonstrated in previous studies. Mulé et al found that satellite nodule was independently associated with both early and overall tumor recurrence.²⁶ Similarly, Kim et al identified the presence of satellite nodule and peritumoral HBP hypointensity as independent factors associated with tumor recurrence after liver transplantation.²⁷ These studies strongly support our findings. Moreover, intrahepatic RFS was comparable between patients with peritumoral HBP hypointensity and those with satellite nodule but without peritumoral HBP hypointensity. Therefore, we established a two-step flowchart based on MRI features for the prediction of early hepatic recurrence, which showed good performance in risk assessment.

Interobserver variability is a major concern in terms of the reliability of the features seen on preoperative MRI images when used as a decision-making tool. In our study, the two radiologists yielded moderate to excellent agreement (κ=0.431–0.885) for MRI features of HCC, which is similar to the results reported by Lee et al (κ 0.64–0.88).²⁸ However, Min et al and Kim et al reported relatively lower κ values (0.38–0.47 and 0.53–0.59, respectively).^27,29 The good interobserver agreement in our study might be attributed to the discussion and case training on MRI features prior to formal assessment.

This study has several limitations. First, it had a single-center retrospective design and small sample size; therefore, there is a possibility of selection bias. Therefore, the results require confirmation in a larger multicenter prospective study. Second, the follow-up duration was relatively short and a proportion of patients were lost to follow-up, which may have diminished the strength of our results. However, most early intrahepatic HCCs occurred in the first year, similar to our data. Third, the main etiology of liver disease in our study was chronic hepatitis, particularly chronic hepatitis B. Further investigations in populations with different underlying causes are required.

Conclusion

In conclusion, we found peritumoral HBP hypointensity to be an important MRI feature for identifying MVI status and early intrahepatic recurrence but with low sensitivity. In patients without peritumoral HBP hypointensity, the presence of a radiological capsule may be useful for identifying the MVI status, whereas satellite nodule may be an independent risk factor for shorter intrahepatic RFS. The use of our two stepwise flowchart that incorporates peritumoral HBP hypointensity and satellite nodule could significantly improve the detection rate of high-risk HCC, which could assist clinicians in treatment decision-making.