Markers for microvascular invasion in hepatocellular carcinoma: Where do we stand?


  • Potential conflict of interest: Nothing to report.

It is well known that hepatocellular carcinoma (HCC) commonly involves the local branches of portal and/or hepatic veins and causes a tumor thrombus even at a relatively early stage. Vascular invasion is classified as macrovascular invasion, which is grossly recognizable (mostly in large to medium vessels), or microvascular invasion (MVI), which can be identified only by microscopic observation (mainly in small vessels such as portal vein branches in portal tracts, central veins in noncancerous liver tissue, and venous vessels in the tumor capsule and/or noncapsular fibrous septa).

Although macroscopic vascular invasion in major vessels (and satellite nodules) is known to be a marker of poor outcomes after liver transplantation (LT) for HCC and is regarded as a contraindication for LT, the significance of MVI as a predictor of poor outcomes is still controversial. The controversy concerns the extent to which MVI (if it is identifiable before surgery) is a contraindication for LT, even though we know that only a minority of patients with MVI will experience HCC recurrence.


[18F]FDG, [18F]fludeoxyglucose; AFP, alpha-fetoprotein; CI, confidence interval; CT, computed tomography; DCP, des-gamma-carboxyprothrombin; FDG, fludeoxyglucose; HCC, hepatocellular carcinoma; HR, hazard ratio; LT, liver transplantation; MRI, magnetic resonance imaging; MVI, microvascular invasion; NM, not mentioned; NS, not significant; PET, positron emission tomography.


We searched the MEDLINE database (2002-2010) to determine the significance of MVI to the outcomes of LT for HCC; we used the keywords hepatocellular carcinoma, HCC, microvascular invasion, liver transplantation, and liver resection. We also performed a full manual search of the bibliographies of selected publications and included 4 additional publications from earlier years. Publications were included if they contained data on MVI and its relationship with tumor characteristics and/or prognostic data. The search resulted in a total of 48 relevant publications.

The publications were then ranked according to the classification proposed by the Oxford Centre for Evidenced-Based Medicine.


Question 1. Is There Any Correlation Between MVI and Tumor Characteristics?

Many published studies have found that the presence of MVI is closely related to the tumor size, number, and histological grade1-8 (Table 1). Esnaola et al.1 studied MVI in 245 HCC patients who underwent surgical resection and fulfilled the Milan criteria, and they found that 33% of their patients had histopathological evidence of MVI. Patients with tumors larger than 4 cm were 3 times more likely to have MVI than those with tumors measuring 4 cm or less. Patients with poorly differentiated or undifferentiated tumors were 6 times more likely to have MVI than those with well-differentiated tumors. In comparison with patients with well-differentiated tumors, patients with multiple tumors were 2 times more likely to have MVI, and patients with moderately differentiated tumors were 2.6 times more likely to have MVI. Others have also reported that the presence of MVI is closely related to a larger tumor size and a lower histological grade (Table 1).

Table 1. MVI and Tumor Characteristics
ReferenceSize (cm)nHistological Grade
Esnaola et al. (1)>4NSModerately to poorly differentiated
Shah et al. (2)>3.8>3NS
Parfitt et al. (3)>3>3Moderately to poorly differentiated
Löhe et al. (4)>5NSModerately to poorly differentiated
Jonas et al. (5)3-52-3Moderately to poorly differentiated
Bhattacharjya et al. (6)>43Moderately to poorly differentiated
Shirabe et al. (7)>2-4MultipleModerately to poorly differentiated
Pawlik et al. (8)>5MultipleHigh grade (moderately to poorly differentiated)

Relationship Between MVI and the Gross Type

The Liver Cancer Study Group of Japan proposed a gross classification scheme for resected HCC in 1997,9 and it has been widely used in Japan. In the Liver Cancer Study Group of Japan classification scheme, nodular tumors are divided into the following subclasses: (1) a simple nodular type (a distinctly nodular tumor that frequently has a capsule), (2) a simple nodular type with extranodular growth (a single nodular tumor with varying degrees of tumor growth beyond the tumor capsule), and (3) a confluent multinodular type (a confluence of several minute to small nodules; see Fig. 1).

Figure 1.

Gross classification scheme for HCC by the Liver Cancer Study Group of Japan. Reprinted with permission from General Rules for the Clinical and Pathological Study of Primary Liver Cancer.9 Copyright 2010, Kanehara & Co., Ltd. Tokyo, Japan.

Since the Liver Cancer Study Group of Japan classification was proposed, it has been reported that the rate of MVI is closely related to the gross type.10-12 In a study of 65 resected HCCs, Hui et al.10 reported that the MVI rates were 17% for the simple nodular type, 25% for the simple nodular type with extranodular growth, and 53% for the confluent multinodular type (P < 0.03 for the simple nodular type versus the simple nodular type with extranodular growth, P < 0.02 for the simple nodular type versus the confluent multinodular type). In a study of 110 resected HCCs, Sumie et al.11 also reported a significantly higher prevalence of MVI for the simple nodular type with extranodular growth (72%) and the confluent multinodular type (84%) in comparison with the simple nodular type (20%, P < 0.001). They stressed that the gross classification of the simple nodular type with extranodular growth or the confluent multinodular type was an independent predictor of MVI in their study [hazard ratio (HR) = 11.81, 95% confidence interval (CI) = 3.93-37.80, P < 0.001].

In a large series, Kaibori et al.13 confirmed that a large tumor size was a preoperative predictor of MVI. Unfortunately, the MVI group included patients with macrovascular invasion.

Similar results were obtained by Shirabe et al.14 for patients undergoing living donor LT. An independent predictor of poor recurrence-free survival was preoperative type 3 HCC (a contiguous multinodular type with a large tumor size, a poor histological grade, and a high incidence of MVI and multiple tumors).

Notably, tumor cell invasion of the portal vein is observed in 27% of cases with early HCC (ie, up to 2 cm) of a distinctly nodular type, and minute intrahepatic metastases in the vicinity of the tumor are present in 10%. These features are not observed in small HCCs of an indistinct nodular type.15


The majority of studies have shown a close correlation between MVI and tumor characteristics. Increases in the prevalence of MVI parallel increases in the tumor size, number, histological grade, and gross features. The last are based on the Japanese gross classification system, which is not widely applied in Western countries.9

In summary, because MVI is a microscopic feature that can be assessed only on the tissue level, MVI cannot be definitely confirmed during the pre-LT stage. However, to a certain extent, the likelihood of the presence or absence of MVI can be predicted during the pre-LT stage on the basis of the tumor size and number, which are assessable with imaging modalities. The histological grade is an additional predictive parameter on the tissue level that can be assessed by pre-LT biopsy, whereas the gross features can be examined only after LT. After LT, MVI is valuable because of its association with the tumor recurrence rate, which we address later.

Notably, 2 studies have reported that the tumor growth rate or the tumor doubling time is predictive of tumor recurrence after LT.16, 17 The association between these features and MVI has not been studied, but they merit further attention as additional important tumor characteristics that might be related to MVI.

Question 2. Are There Any Imaging Modalities for Detecting MVI ?

Although macrovascular invasion can be preoperatively detected by conventional imaging modalities such as computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound, the preoperative imaging determination of MVI is impossible because MVI is a microscopic parameter.

However, Kornberg et al.18 reported that increased [18F]fludeoxyglucose ([18F]FDG) uptake during positron emission tomography (PET) is predictive of MVI and tumor recurrence after LT. In their study, 42 patients who underwent LT for HCC were examined with PET preoperatively; 16 had positive PET scans, whereas 26 showed no increase in their fludeoxyglucose (FDG) uptake. Five of 6 patients with a poor tumor grade had a PET-positive status, and 14 of 17 patients with MVI (82.4%) had a positive PET finding. Eight of the 16 patients with a PET-positive status (50%) developed HCC recurrence, whereas only 1 of the 26 PET-negative patients did (P < 0.001). Furthermore, the PET-negative patients had a significantly better 3-year recurrence-free survival rate than the PET-positive patients. It has been shown that well-differentiated HCC cells exhibit an [18F]FDG metabolism similar to that of normal liver cells, whereas poorly differentiated tumor cells do not. The authors hypothesized that PET scan patterns may have a clinically relevant correlation with the presence of MVI because of the close relationship between MVI and a poor tumor grade. However, there is no evidence that FDG uptake is an independent predictor.

More recently, positive FDG PET results at the time of listing for LT have been found to reflect tumor behavior and to be strongly associated with dropout due to tumor progression. PET could be proposed as an additional tool for the pre-LT assessment of candidates with HCC.19

Hui et al.10 tested whether the Japanese gross classification system9 could be used with preoperative dynamic CT (10-mm interval slices). They found that classifications based on CT findings and classifications based on macroscopic observations were concurrent for 46% of their patients. The simple nodular type, the simple nodular type with extranodular growth, and the confluent multinodular type were identified at rates of 62%, 26%, and 40%, respectively.


So far, conventional imaging modalities have been ineffective for the preoperative detection of MVI. Only 1 study has shown the potential usefulness of PET scans in the prediction of MVI.18 The preliminary results of an MRI study have shown that the MRI features of HCC do not predict MVI in the transplanted liver.20 Although the gross classification scheme of the Liver Cancer Study Group of Japan9 is not popular in Western countries, because of the 40% identity rate for dynamic CT findings and the confluent multinodular type (for which there is an MVI prevalence rate of 53%-84%), gross subclassification by dynamic CT merits further study.11 This can also be recommended because of the potential value of PET scans and MVI and the scarcity of data.

Large multicenter studies are required to precisely assess the predictive value of imaging modalities for MVI.

Question 3. Are There Any Biochemical Markers for Predicting MVI?

Des-gamma-carboxyprothrombin (DCP), also known as protein induced by vitamin K absence or antagonist II, is an abnormal form of prothrombin, and it has been used as a good diagnostic biomarker for HCC, particularly in Japan.21-23 It has been reported that high serum levels of DCP reflect the invasiveness of HCC and are related to the histological vascular invasion of cancer cells. Koike et al.22 prospectively studied 227 patients with HCC who did not show portal venous invasion and who received percutaneous ethanol injection therapy and/or microwave coagulation therapy at the time of their first admission. The patients were followed for a mean of 19 months with ultrasonography every 3 months, with CT scans every 6 months, and with measurements of biochemical parameters such as DCP and alpha-fetoprotein (AFP) every month. A multivariate Cox regression analysis of various factors showing significant correlations with the later development of portal venous invasion revealed that the DCP level was the strongest predisposing factor (P < 0.001). Shirabe et al.7 studied 218 patients who had tumors without any extrahepatic metastases or vascular invasion (according to their preoperative evaluations) and underwent HCC resection. They found that the preoperative DCP level for the patients with MVI was significantly higher than the level for the patients without MVI (P < 0.049), and it was an independent predictor of MVI. In another recent study,14 the same authors showed that a preoperative serum DCP level exceeding 300 mAU/m was an independent predictor of poor recurrence-free survival in patients undergoing living donor LT.

In a study of 144 HCC patients treated with living donor LT, Fujiki et al.23 reported that the incidence of MVI was significantly higher for patients with a DCP level > 400 mAU/mL versus patients with a DCP level ≤ 400 AU/mL.

Eguchi et al.21 studied 179 patients without macroscopic venous invasion who underwent curative resection for HCC. High preoperative levels of DCP and the gross appearance of the tumor appeared to be strong predictive factors for MVI, even for tumors less than 5 cm in size. The gross features that were associated with MVI were extranodular growth, multinodularity, and infiltrative growth.

The value of DCP in predicting MVI recurrence should not be compared to the value of DCP in predicting HCC occurrence. Sterling et al.24 concluded that mild to moderate elevations of the total AFP and DCP levels (but not the AFP-L3 levels) occur frequently in patients with chronic hepatitis C and advanced cirrhosis, but they are related to factors other than HCC and are poor indicators of HCC. They recommended that these biomarkers not be part of routine surveillance protocols.

An increase in the AFP level greater than 15 μg/L/month during the wait for LT has been shown to be the most relevant preoperative prognostic factor for low overall and disease-free survival. An elevated AFP level was shown to be a predictor of a lower recurrence-free survival rate in a large series of 283 transplant patients (odds ratio = 2.88, 95% CI = 1.43-5.80, P = 0.003).25 An increase in the AFP level could be a preoperative pathological marker of tumor aggressiveness.26 In the past 2 decades, attempts have been made to detect circulating tumor cells in the peripheral blood of HCC patients through the examination of tumor-related markers.27-29 However, the results are still only probable because of the lack of truly reliable HCC-related markers.

It is likely that new blood markers will be proposed in the future. Recently, the angiopoietin 2 level has been shown to be a predictor of tumor invasiveness in patients with HCC.30


The significance of a high serum DCP level as a good biomarker for predicting MVI has been suggested mostly in Japanese studies. Hence, an international, multicenter study is required to assess the precise predictive value of DCP in a broader group of HCC populations.

Furthermore, other predictive parameters should be evaluated. Reports on factors in HCC tissue that are associated with MVI at protein and gene expression levels are emerging. Liver-intestine cadherin has been reported to be correlated with MVI in hepatitis B virus–related HCC.31 A genetic signature of microvascular invasiveness has also been reported in a genome-wide analysis.32 The potential of these factors as future biomarkers merits further studies.

The use of an artificial neural network to predict the tumor grade and MVI on the basis of noninvasive variables is notable.33 An artificial neural network and logistic regression models were based on a training group of 175 randomly chosen patients and were tested on 75 patients. In the testing group, the artificial neural network correctly identified 93.3% of the tumor grades (k = 0.81) and 91% of the MVI cases (k = 0.73). The logistic models correctly identified 81% of the tumor grades (k = 0.55) and 85% of the MVI cases (k = 0.57). In this study, many clinical, radiological, and pathological factors were taken into account, but the DCP level and the body mass index were not. Surprisingly, in a study of 138 consecutive patients who underwent surgery at Columbia University Medical Center from January 1, 2002 to January 9, 2008, Siegel et al.34 found that patients with a body mass index > 30 kg/m2 had MVI at rate of 40%, whereas patients with a body mass index < 25 kg/m2 had MVI at a rate of 14%.

Question 4. Is MVI Related to the LT Outcome?

The relationship between MVI and LT outcomes is still a matter of debate. The majority of studies favor the view of a relationship between MVI and poor posttransplant outcomes.1-11, 35-53 Many studies have reported a close correlation between the presence of MVI and a poor prognosis after LT; Bhattacharjya et al.6 performed a prospective study of 30 patients treated with LT for relatively small HCCs (range = 6-75 mm). They found that the tumor size and the presence of multifocal tumors did not influence survival, but MVI, which was more common with larger tumors (38% with tumors < 4 cm and 60% with tumors > 4 cm, P < 0.01), influenced survival. Cescon et al.25 reported a single-center series of 283 patients who underwent transplantation for HCC between 1997 and 2009; a Cox proportional hazards model showed that MVI was a predictor of lower recurrence-free survival (odds ratio = 4.82, 95% CI = 1.87-12.41, P = 0.001). The fact that the Milan criteria did not appear to be prognostic may not be surprising because 89% of the patients fulfilled the up-to-7 criteria at the time of orthotopic LT.

In a series of 51 patients undergoing LT for HCC, Choi et al.37 retrospectively analyzed 5 cases of incidental HCC with a mean size of 1.16 cm, no MVI, and good to moderate tumor differentiation. They found no recurrence of HCC in a mean follow-up period of 14 months. Lai et al.47 reported that only MVI and exceeding the University of California San Francisco criteria were independent risk factors for recurrence in a multivariate analysis. However, many have reported that a combination of a larger tumor size, a higher histological grade (less differentiation), and MVI is the strongest factor related to recurrence and a poor prognosis after LT. Jonas et al.5 found the presence of vascular invasion and the histological grade to be the only statistically significant independent predictors of poor survival after LT. In a study of 69 patients with HCC who underwent LT, Plessier et al.54 reported that MVI was significantly correlated with the presence of satellite nodules (P < 0.02) and a poor prognosis.

On the other hand, a few studies have not been able to correlate poor results with MVI, although macrovascular invasion has been related to bad outcomes. Lee et al.49 followed 38 patients after LT for HCC for a mean period of 17.7 months (range = 4-30 months), and they found that the number of tumor nodules and the presence of MVI did not affect tumor recurrence. Fan et al.55 studied 1078 patients who underwent orthotopic LT and found no correlation between MVI and overall or disease-free survival. Kornberg et al.46 documented MVI as a risk factor for tumor recurrence after LT. A poor tumor grade (HR = 21.8, 95% CI = 4.9-95.3, P < 0.001) and MVI (HR = 14.1, 95% CI = 1.4-147.1, P = 0.027) were identified as independent risk factors for reduced recurrence-free survival after LT.

Although the risk of recurrence is higher in patients with MVI, not all patients with MVI will experience recurrence. This has to be balanced with macrovascular invasion and satellite nodules (all detected by imaging techniques), which are contraindications for LT because of the high incidence of recurrence.

Table 2 summarizes the publications documenting several types of tumor features, biomarkers, and imaging modalities. Table 3 summarizes the publications specifically documenting MVI and several of these issues, and available correlations are included.

Table 2. Overview of Studies of Several Tumor Characteristics
ReferencePatients (n)Tumor Characteristics
AnalyzedMeeting Milan CriteriaSize (cm)nGradeAFP (ng/mL)DCP (mAU/mL)Radiological ImageGross Classification
  • *

    Confirmed histopathologically.

Esnaola et al. (1)245245++++
Shah et al. (2)154154+++
Parfitt et al. (3)7550*++++
Löhe et al. (4)9747*+++
Jonas et al. (5)120120+++
Bhattacharjya et al. (6)3030+++
Shirabe et al. (7)218NM+++++
Pawlik et al. (8)1073NM++++
Hui et al. (10)6565*+++
Sumie et al. (11)110110*++++++
Nakashima et al. (12)209209*+++
Kaibori et al. (13)434NM+++++
Shirabe et al. (14)11965++++++
Cucchetti et al. (16)62NM++++
Kornberg et al. (18)4220*+++++
Eguchi et al. (21)179NM+++++
Koike et al. (22)227NM++++
Fujiki et al. (23)14479+++++
Cescon et al. (25)283208*++++
Vibert et al. (26)15399++++
Ding et al. (31)255NM++++
Tanaka et al. (32)59NM++++++
Cucchetti et al. (33)250190*++++
Siegel et al. (34)138NM++
Roayaie et al. (35)131NM++++
Cucchetti et al. (36)204NM++++
Choi et al. (37)5NM++++
Dudek et al. (38)5548*+++
Plessier et al. (54)6969+++
Mazzaferro et al. (39)1556444*+++
Yao et al. (40)7046*++++
Coelho et al. (41)45NM++++
Park et al. (42)213176*++++
Vivarelli et al. (43)139113*++++
Silva et al. (44)257231++++
Suh et al. (45)190+++++
Kornberg et al. (46)6027*++++
Lai (47)8559*+++
D'Amico et al. (48)479272*++++
Lee et al. (49)3821*++
Bargellini et al. (53)330+++++
Fan et al. (55)1078394*++++
Table 3. Correlation Between MVI and Tumor Characteristics
ReferenceTumor Characteristics
Size (cm)nGradeAFP (ng/mL)DCP (mAU/ mL)Radiological ImageGross Classification
  • NOTE: A bolded plus sign indicates an independent predictor in the multivariate analysis, and a plus sign that is not bolded indicates a correlation between the factor and MVI but not an independent predictor. NS indicates no correlation, and a blank field indicates that data were not available.

  • *

    The preoperative tumor size.

  • Fifteen micrograms per liter per month.

Esnaola et al. (1)+ >4++ Poorly differentiated/ undifferentiatedNS   
Shah et al. (2)+ >3.6+ >3.8NS    
Parfitt et al. (3)+ >5      
Löhe et al. (4)+ >5NS+ Intermediate/ high grade    
Jonas et al. (5)+ >5++ Grade 2/3    
Bhattacharjya et al. (6)+ >4NS+ Moderately differentiated/ poorly differentiated    
Shirabe et al. (7)+NS+NS+  
Pawlik et al. (8)++ Multiple+ High grade+ ≥1000   
Hui et al. (10)      + Contiguous multinodular type
Sumie et al. (11)+NS+NSNS + Contiguous multinodular type and intrahepatic metastasis
Nakashima et al. (12)      + Simple nodular type with extranodular growth and contiguous multinodular type
Kaibori et al. (13)+ >5*NS+ Poorly differentiatedNS+ ≥200  
Shirabe et al. (14)      + Contiguous multinodular type
Kornberg et al. (18)NSNS+ Poorly differentiatedNS + PET -positive 
Eguchi et al. (21)+ >5+ ++ + Simple nodular type with extranodular growth, contiguous multinodular type, and infiltrative type
Fujiki et al. (23)NSNS + ≥800+ ≥400  
Vibert et al. (26)± >3NS + Rapid progression   
Cucchetti et al. (33)++ +   
Siegel et al. (34)NS  +   
Silva et al. (44)NSNS± Poorly differentiated+   

Other issues that should be taken into consideration are as follows:

  • 1The indication for LT in patients with HCC beyond the Milan criteria. Long-term survival is indeed achievable when MVI is absent; this emphasizes the need to find MVI biomarkers so that LT can be successfully performed for patients with HCC beyond the Milan criteria.50 The use of the up-to-7 criteria39 seems, however, to be a reasonable approach.51
  • 2The pre-LT treatment. Because of the long waiting time before LT, more patients are being treated preoperatively (locoregional therapy). Furthermore, patients with tumors exceeding the Milan criteria are also being treated so that their HCC can be down-staged.52 Locoregional tumor therapy has induced >50% tumor necrosis in two-thirds of cases. Even after down-staging, the overall survival and disease-free survival rates are lower than those for patients within the Milan criteria. However, in patients with HCC exceeding the Milan criteria, a complete response after transarterial chemoembolization (according to the amended Response Evaluation Criteria in Solid Tumors guidelines) has been associated with excellent posttransplant outcomes. The complete response was associated with lack of MVI in the expanded liver.53


  • 1MVI is an independent risk factor and/or one of the risk factors for HCC recurrence and poor outcomes after LT. However, a favorable prognosis can be expected when LT is performed for patients with small and well-differentiated HCC because MVI and HCC recurrence are rare in these cases.
  • 2MVI is most likely to be present in tumors larger than 3 cm, in tumors whose gross pattern is the nodular type with extranodular growth or the confluent multinodular type, and in tumors with a high histological grade (which shows a less differentiated pattern).
  • 3It is impossible to detect MVI preoperatively by conventional imaging modalities, and there is no widely recognized biomarker for predicting MVI. Because of the significance of MVI with respect to the recurrence of tumors and the prognosis, the possibility of predicting its presence by a comprehensive consideration of tumor characteristics (the size, histological grade, and number) and the potential role of imaging modalities should be further explored; before then, it should be better defined and graded. Further evaluations of the emerging roles of radiological, biological, and molecular profiling of HCC with an MVI signature are needed. Although the use of DCP is not popular in Western countries, its close relationship with the biological behavior of HCC suggests its potential as a biomarker of MVI, and it merits further study.
  • 4An artificial neural network may become useful for identifying the tumor grade and MVI on the basis of preoperative variables.