Accuracy and disagreement of computed tomography and magnetic resonance imaging for the diagnosis of small hepatocellular carcinoma and dysplastic nodules: Role of biopsy

Authors

  • Thomas Sersté,

    1. Assistance-Publique Hôpitaux de Paris, APHP Service d'Hépatologie, Hôpital Beaujon, Clichy, France
    2. Hepato-Gastroentérologie, Hôpital Saint-Pierre, Université Libre de Bruxelles, Bruxelles, Belgium
    Search for more papers by this author
  • Vincent Barrau,

    1. Assistance-Publique Hôpitaux de Paris, APHP, Hôpital Beaujon, Departement de Radiologie, Clichy, France
    Search for more papers by this author
  • Violaine Ozenne,

    1. Assistance-Publique Hôpitaux de Paris, APHP Service d'Hépatologie, Hôpital Beaujon, Clichy, France
    Search for more papers by this author
  • Marie-Pierre Vullierme,

    1. Assistance-Publique Hôpitaux de Paris, APHP, Hôpital Beaujon, Departement de Radiologie, Clichy, France
    Search for more papers by this author
  • Pierre Bedossa,

    1. Service d'Anatomo-Pathologie, Hôpital Beaujon, Clichy, France
    2. Université Paris Diderot, Sorbonne Paris Cité, INSERM Centre de Recherche Biomédicale Bichat Beaujon, CRB3 U773 75018 Paris, France
    Search for more papers by this author
  • Olivier Farges,

    1. Service de Chirurgie, Hôpital Beaujon, Clichy, France
    Search for more papers by this author
  • Dominique-Charles Valla,

    1. Assistance-Publique Hôpitaux de Paris, APHP Service d'Hépatologie, Hôpital Beaujon, Clichy, France
    2. Université Paris Diderot, Sorbonne Paris Cité, INSERM Centre de Recherche Biomédicale Bichat Beaujon, CRB3 U773 75018 Paris, France
    Search for more papers by this author
  • Valérie Vilgrain,

    1. Assistance-Publique Hôpitaux de Paris, APHP, Hôpital Beaujon, Departement de Radiologie, Clichy, France
    2. Université Paris Diderot, Sorbonne Paris Cité, INSERM Centre de Recherche Biomédicale Bichat Beaujon, CRB3 U773 75018 Paris, France
    Search for more papers by this author
  • Valérie Paradis,

    1. Service d'Anatomo-Pathologie, Hôpital Beaujon, Clichy, France
    2. Université Paris Diderot, Sorbonne Paris Cité, INSERM Centre de Recherche Biomédicale Bichat Beaujon, CRB3 U773 75018 Paris, France
    Search for more papers by this author
  • Françoise Degos

    Corresponding author
    1. Assistance-Publique Hôpitaux de Paris, APHP Service d'Hépatologie, Hôpital Beaujon, Clichy, France
    • Hopital Beaujon, Service d'Hepatologie, 100 Boulevard du Général Leclerc, 92118 Clichy, France
    Search for more papers by this author
    • Fax: (33)-01-47-30-94-40


  • Potential conflict of interest: Nothing to report.

Abstract

Liver macronodules, ranging from benign to low-grade or high-grade dysplastic nodules (LGDNs/HGDNs) and hepatocellular carcinoma (HCC), may develop during chronic liver diseases (CLDs). Current guidelines were recently updated and the noninvasive criteria for the diagnosis of small HCC are based on a single typical radiological pattern and nonconclusive coincidental findings with two techniques. This study aimed to assess the accuracy and disagreements of noninvasive multiphasic examinations for the diagnosis of HCC and dysplastic nodules (DNs) and the role of biopsy. Seventy-four consecutive patients with CLD with ultrasound-detected 1-2-cm nodules underwent, within 1 month, multiphasic computed tomography (CT), magnetic resonance imaging (MRI), and biopsy of the nodule. Median age was 60 years; 33 patients (45%) had hepatitis C virus, 20 (27%) had hepatitis B virus, and 13 (18%) patients had no cirrhosis. Biopsy revealed 47 HCCs, 6 HGDNs, 1 LGDNs, 1 cholangiocarcinoma, and 1 epithelioid hemangioendothelioma. There were no tumors in the other 18 patients. All patients (31 of 31; 100%) who had conclusive coincidental findings (i.e., arterial enhancement and washout) on both examinations had HCC or HGDN (sensitivity, 57%; specificity, 100%). All patients (51 of 51; 100%) who had conclusive findings on at least one of the two examinations had HCC or HGDN (sensitivity, 96%; specificity, 100%). There was a disagreement regarding imaging findings between CT and MRI in 21 of 74 (28%) patients and no washout on both examinations in 23 of 74 patients (31%). In these 44 patients, liver biopsy provided an initial accurate diagnosis. Conclusion: The noninvasive diagnosis of HCC or HGDN can be obtained if arterial enhancement and washout are found in a single dynamic imaging examination. These findings are frequently discordant on both CT and MRI, supporting the place of biopsy for the diagnosis of small HCCs. (HEPATOLOGY 2011)

Hepatocellular carcinoma (HCC) is the sixth most common malignant tumor in the world and its incidence has risen rapidly.1-3 Monitoring is recommended in the presence of cirrhosis and in areas endemic for chronic liver diseases (CLDs), such as hepatitis B virus (HBV) and/or hepatitis C virus (HCV).4-6 Cirrhosis is characterized by the development of a wide spectrum of hepatocellular nodules, ranging from benign to malignant.7, 8 Increased attention must be paid to small nodules (≤2 cm) that encompass both dysplastic nodules (DNs), considered to be premalignant, and overt HCC recognized as “small HCC.” Indeed, HCC may develop from DNs in a multistep process of liver carcinogenesis, both in patients with cirrhosis and in earlier stages of CLDs.9-11 Although detecting small HCC and premalignant DNs is critical for the outcome of patients, they can be difficult to distinguish from other liver nodules.12, 13 At present, the monitoring protocol for the early detection of HCC includes alpha-fetoprotein (AFP) levels and ultrasound (US). Based on the vascular pattern of the nodules, the American Association for the Study of Liver Diseases (AASLD) published guidelines in 2005 for the diagnosis of HCC 2 cm or smaller in the presence of cirrhosis and recommended performing two dynamic imaging techniques in this setting.14-16 A noninvasive diagnosis of HCC can be obtained when coincidental findings are observed and include hypervascularization on arterial-phase imaging and washout in the portal or delayed phase, in both imaging modalities. Indeed, hypervascularization in the arterial phase with portal or delayed washout is now considered to be a distinctive feature of malignancy.13 Very recently, AASLD guidelines were updated.15 US monitoring was recommended and the use of a diagnostic algorithm was suggested, including the detection of arterial hypervascularity and portal or delayed venous washout in only one dynamic radiological procedure, contrast-enhanced computed tomography (CT), or magnetic resonance imaging (MRI). When these features are not present, either the contrast-enhanced examination or biopsy is recommended.15

So far, the noninvasive criteria for HCC have only been validated for cirrhosis. However, HCC can occur before cirrhosis develops, and noninvasive diagnostic criteria could be evaluated in this context. Moreover, these criteria do not take into account premalignant DNs that should also be detected earlier.

Based on the previous AASLD guidelines for the diagnosis of HCC, we assessed the diagnostic accuracy of noninvasive techniques in US-detected nodules in 75 consecutive patients with CLD and/or cirrhosis. Sensitivity and specificity of the vascular pattern of HCC on enhanced multiphasic CT and MRI were assessed with biopsy for the diagnosis of DNs and small HCC for this purpose. Thus, the aim of this study was to answer the following questions: (1) What is the diagnostic accuracy of a single contrast-enhanced study?, (2) What is the frequency of disagreement between two contrast-enhanced imaging procedures (i.e., CT or MRI)?, and (3) What is the advantage and the role of biopsy in this context?

Abbreviations

AASLD, American Association for the Study of Liver Diseases; AFP, alpha-fetoprotein; CC, cholangiocarcinoma; CLD, chronic liver disease; CT, computed tomography; DNs, dysplastic nodules; EHE, epithelioid hemangioendothelioma; GPC-3, glypican-3; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HGDN, high-grade dysplastic nodule;IHC, immunohistochemical; LGDN, low-grade dysplastic nodule; MR, magnetic resonance; MRI, magnetic resonance imaging; RFA, radiofrequency ablation; US, ultrasound.

Patients and Methods

Between January 2005 and December 2010, all consecutive patients referred to our liver unit with cirrhosis or CLD and small nodules (diameter between 1 and 2 cm) newly detected by US and without previous HCC were included in the study. CLD was defined as a 2-fold increase in serum transaminases above normal for least 6 months before inclusion. This was a single-center, case-only, observational study. Subsequent to institutional approval and according to clinical practice in our center during the study period, all patients underwent an initial evaluation, including systematic contrast-enhanced multiphasic CT, MRI, and a liver biopsy of the nodule, all performed within 1 month. The Comité d'Evaluation de l'Ethique des Projets de Recherche Biomédicale du GHU Nord (Institutional Review Board of Northern Paris Hospitals, Paris 7 University, APHP, Paris, France) reviewed and approved this research study. During this first evaluation, patients underwent systematic biopsy of both the newly diagnosed nodule and nontumoral liver. If a patient had multiple US-detected nodules, the largest nodule was biopsied.

Patients with overt HCC or other malignancies were treated as appropriate (e.g., surgery, chemoembolization, radiofrequency, or liver transplant). Patients in whom no HCC was found underwent close radiologic follow-up, including enhanced CT or MRI 3 months after the first biopsy. If the nodule progressed or changed its vascular profile, a second biopsy was discussed. Evaluation of tumoral progression was based on the size of all detected nodules and defined as a 20% or greater increase in the sum of the diameters of each of the nodules reported in the first evaluation. The purpose of repeated biopsies was to minimize false-negative results.

Radiological Analysis.

All patients underwent multidetector, multiphasic CT before and after contrast medium administration (LightSpeed Plus; General Electric Healthcare, Waukesha, WI), including arterial, portal venous, and delayed phases. MRI images were obtained with a 1.5-T unit (Intera; Philips, Best, The Netherlands), using a phased array torso coil. All patients underwent transverse T1- and T2-weighted MRI and contrast-enhanced T1-weighted sequences during arterial, venous, and delayed imaging. CT and MRI results were read by two radiologists in consensus (V.B. and M-P.V.) who were blind to biopsy results. Nodules were classified according to lesion enhancement. Vascular pattern was qualified as “conclusive” for HCC if contrast washout occurred, defined as the presence of hypervascularity during the arterial phase followed up by a hypodense/hypointense appearance in later phases defining washout. Nodules in which an enhancement was found during the arterial phase without washout were qualified as “suspicious.”

Histological Analysis.

Percutaneous US-guided liver biopsies were performed using 18-gauge cutting needles. Specimens were routinely processed for systematic histological evaluation. All biopsies were routinely read by one pathologist (P.B.), then independently reviewed in a blinded manner by a second pathologist (V.P.) who was unaware of the previous pathological diagnosis and imaging results. Pathological diagnosis of DNs and HCC was based on international criteria for the classification of small liver nodules.17 Briefly, HCC was diagnosed when several of the following features were present, including increased cell density more than 2 times that of the surrounding tissue, increased nuclear/cytoplasm ratio, and irregular thin-trabecular and pseudoglandular patterns. Stromal invasion, which is recognized as a key feature of early HCC, was systematically screened.17 Diagnosis of high-grade dysplastic nodules (HGDNs) was made when nodules displayed architectural and/or cytologic atypias, but the atypias were insufficient for a diagnosis of overt HCC.18 Discordant diagnoses between both pathological readings were systematically reviewed in a common reading to reach a consensual diagnosis. When feasible, additional immunohistochemical (IHC) study, using glypican-3 (GPC-3) (1/100; BioMosaics Inc., Burlington, USA), was performed in difficult cases between HGDN and HCC.18 When positive GPC-3 staining was observed, the diagnosis of HCC was proposed. Grading of tumors was performed using Edmonson's classification.19 Nontumorous liver lesions were scored according to METAVIR.20

Analyses of Sensitivity and Specificity.

Analysis of the sensitivity and specificity of the vascular pattern of the nodules for the diagnosis of DNs and HCC with CT and MRI was performed in two sessions. The diagnostic “gold standard,” or the reference method for diagnosis, was the results of biopsy for both sessions. The first session calculated the accuracy of the noninvasive diagnosis of HCC and only took into account patients with biopsy-proven HCC. The aim of the second session was to define the accuracy of the noninvasive diagnosis of malignant and premalignant nodules together and, therefore, included patients who had both HCC and HGDN on biopsy.

Results

Seventy-five patients fulfilled the criteria for inclusion and underwent a multiphasic CT, MRI, and a liver biopsy of the nodule. One patient was withdrawn from the study because the studied nodule with washout on both examinations, without conclusive diagnosis on biopsy, underwent radiofrequency ablation (RFA). Clinical characteristics of the 74 patients are summarized in Table 1. Median age was 60 years (range, 38-88). Median AFP was 8 ng/mL (range, 1- 413); AFP was above 20 ng/mL in 19 patients (26%) and above 200 ng/mL in 2 patients. In 69 patients, histological analysis of nontumoral (i.e., adjacent) liver tissue was available. According to the METAVIR classification, there was stage F4 fibrosis (i.e., cirrhosis) in 56 patients (75%), F3 (numerous fibrous septa) in 3 (4%), and minimal/moderate fibrosis F1-F2 in 10 patients (14%). Cirrhosis was clinically or radiologically evident in 5 additional patients.

Table 1. Patient (N = 74) Characteristics at Enrollment
  1. Results are presented as number (n) or median and range.

  2. Abbreviation: NASH, nonalcoholic steatohepatitis.

Gender, male/female, n (%)58 (78)/16 (22)
Age (years)60 (38-88)
Cause of chronic liver disease 
 Hepatitis C virus, n (%)33 (45)
 Hepatitis B virus, n (%)20 (27)
 Alcohol, n (%)12 (16)
 NASH, n (%)4 (5)
Prothrombin time (% of normal)79 (29-118)
Platelets count (109/L)120 (46-300)
Total bilirubin (μmol/L)27 (4-100)
Serum albumin (g/L)34 (7-45)
Aspartate aminotransferase (UI/L)74 (12-390)
Alanine aminotransferase (UI/L)87 (14-333)
Number of nodules1 (1-3)
METAVIR fibrosis stage19 
 F13 (4)
 F27 (10)
 F33 (4)
 F456 (75)
Baseline AFP (ng/mL)8 (1-413)

Characteristics of the Nodules and Pathological Diagnosis.

A single nodule was detected by US in 59 patients (80%). Two nodules were found in 12 patients (16%), and 3 nodules were found in 3 patients (4%). The median size of the largest nodule was 18 mm (range, 10-20).

At the initial evaluation, HCC diagnosis was achieved on biopsy by both pathologists in 46 patients, and DNs were found in 8 patients. Additional IHC analysis performed in cases of DN showed GPC-3 consistent with the diagnosis of HCC in 1 of 8 DN cases, so that 47 patients were finally diagnosed with HCC and 7 patients with DN. Of these DNs, 6 patients had a high-grade dysplastic nodule (HGDN) (Fig. 1). Grading of the tumors was available in 44 of 47 HCC patients and showed the following: well-differentiated tumor in 25 of 44 patients (57%), moderately differentiated in 14 of 44 patients (31%), and poorly differentiated in 5 of 44 (11%).

Figure 1.

Pathological diagnosis of patients according to characteristics of the nodules.

In the group of 54 patients with biopsy-proven HCC or DN, biopsy of nontumoral livers revealed cirrhosis in 41 (80%), F3 fibrosis in 3 (5%), and F2 fibrosis in 6 patients (12%).

In the remaining cases, pathological diagnosis revealed a regenerative macronodule in 9 cases, cholangiocarcinoma (CC) in 1 case, and epithelioid hemangioendothelioma (EHE) in 1 case. In 9 patients, biopsy of the nodule showed features of CLD without any features of DN or HCC (Fig. 1).

The 2 patients with CC and EHE underwent surgery, and final pathologic analysis confirmed the biopsy diagnosis. According to our study protocol, the 25 remaining patients with no evidence of HCC were carefully followed up and underwent repeated imaging tests (e.g., contrast-enhanced multiphasic CT or MRI) for a median of 43 months (range, 6-80).

Among the 7 patients with DN on first biopsy, 6 corresponding to the 6 HGDNs showed vascular changes or tumoral progression over time during the radiological follow-up. These 6 patients underwent an appropriated treatment: transplantation (2 patients), chemoembolization (2 patients), liver segmentectomy (1 patient), and RFA (1 patient). In 4 of these 6 HGDN cases, a second histological analysis was obtained, as mentioned in the description of the study design, and showed an overt HCC after 5-18 months.

None of the remaining 18 patients without HCC or DN on the first evaluation developed radiological evidence of tumoral progression in the second evaluation or during follow-up (6-80 months), and none required a second biopsy according to the study protocol.

Diagnostic Accuracy of CT and MRI for the Diagnosis of HCC (DNs Not Included).

On CT, nodules were hypervascular in 51 of 74 cases (69%; suspicious diagnosis), including 40 of 74 (54%) that showed washout during the portal or late phase (conclusive diagnosis). On MRI, nodules were hypervascular in 57 of 74 cases (77%; suspicious), including 42 of 74 with washout during the portal or late phase (57%; conclusive). Sensitivity and specificity of conclusive imaging findings to diagnose HCC on CT were 74% and 81%, respectively, whereas they were 81% and 85% on MRI, respectively (Table 2).

Table 2. Diagnostic Accuracy of CT and MRI for 1-2-cm Hepatocellular Carcinoma*
  Patient NumberSensitivity (%)Specificity (%)Positive Predictive Value (%)Negative Predictive Value (%)
  • *

    DNs not included: N = 47.

  • At least one typical vascular pattern on CT and/or MRI.

  • Coincidental typical vascular pattern on both CT and MRI.

  • §

    Suspicious refers to arterial uptake of contrast medium.

  • Conclusive refers to arterial uptake of contrast medium and washout of contrast medium during portal or delayed phases.

CT scanSuspicious§5137/47 (79)13/27 (48)37/51 (73)13/26 (50)
Conclusive4035/47 (74)22/27 (81)35/40 (87)22/34 (65)
MRISuspicious5744/47 (94)14/27 (52)44/57 (77)14/17 (82)
Conclusive4238/47 (81)23/27 (85)38/42 (90)23/32 (72)
Arterial uptake+washout at CT and/or MRICriteria positive5146/47 (98)22/27 (81)46/51 (90)22/23 (96)
Arterial uptake+washout at CT and MRICriteria positive3127/47 (57)23/27 (85)27/31 (87)23/43 (53)

Sensitivity and specificity for a conclusive diagnosis of HCC (i.e., hypervascularization and wash-out during the portal or late phase) on at least one of the two imaging techniques was 98% and 81%, respectively. The presence of coincidental hypervascularization and washout during the portal or late phase in both examinations for the diagnosis of HCC had a sensitivity of 57% and a specificity of 85%.

Diagnostic Accuracy of CT and MRI for the Diagnosis of HCC and DNs.

Six of the seven DNs showed arterial enhancement on CT, whereas 4 of them were hypervascular on MRI. Washout was observed on CT in 5 HGDNs and on MRI in 4 HGDNs. There was coincidental arterial enhancement and washout during the portal or late phase on CT and MRI in 4 HGDNs (Table 3).

Table 3. Diagnostic Accuracy of CT and MRI for the Diagnosis of 1-2-cm Hepatocellular and HGDNs*
  Patient NumberSensitivity (%)Specificity (%)Positive Predictive Value (%)Negative Predictive Value (%)
  • *

    N = 53. The patient with LGDN was not included.

  • At least one typical vascular pattern on CT and/or MRI.

  • Coincidental typical vascular pattern on both CT and MRI.

  • §

    Suspicious refers arterial uptake of contrast medium.

  • Conclusive refers to arterial uptake of contrast medium and washout of contrast medium during portal or delayed phases.

CTscanSuspicious§5143/53 (81)13/21 (61)43/51 (84)13/23 (57)
Conclusive4040/53 (75)21/21 (100)40/40 (100)21/34 (59)
MRISuspicious5748/53 (90)12/21 (57)48/57 (84)12/17 (70)
Conclusive4242/53 (79)21/21 (100)42/42 (100)21/32 (66)
Arterial uptake+washout at CT and/or MRICriteria positive5151/53 (96)21/21 (100)51/51 (100)21/23 (91)
Arterial uptake+washout at CT and MRICriteria positive3131/53 (57)21/21 (100)31/31 (100)21/43 (49)

The patient with LGDN did not show arterial enhancement on CT or on MRI.

Sensitivity and specificity of hypervascularization and washout during the portal or late phase in the diagnosis of HCC or HGDN (LGDN excluded) with CT were 75% and 100%, respectively, whereas they were 79% and 100%, respectively, with MRI.

Washout did not occur in any of the nodules that were histologically classified as non-HCC.

Sensitivity and specificity of hypervascularization and wash-out during the portal or late phase in the diagnosis of HCC or HGDN on at least one of the two imaging techniques were 96% and 100%, respectively. The presence of coincidental hypervascularization and wash-out during the portal or late phase in both examinations for the diagnosis of HCC or HGDN had a sensitivity of 57% and a specificity of 100%.

Analysis of the Discordant Vascular Pattern Between CT and MRI.

In the complete series including 74 patients, there were 31 cases with coincidental arterial uptake and washout during the portal or delayed phase corresponding to 27 HCCs and 4 DNs. There were 21 (28%) discrepancies for washout between the two examinations, including 19 HCCs and 2 DNs. Washout was not identified on either CT or MRI for 23 patients (9 macronodules, 9 cases of fibrosis, 1 CC, 1 EHE, 1 HCC, and two DNs). The differences in vascular pattern between CT and MRI restricted to 53 biopsy-proven HCCs or HGDNs are reported in Table 4.

Table 4. Analysis of Concordance and Discordance Between the Two Imaging Techniques in the Diagnosis of Lesions of HCC and HGDN (N = 53)*
  CT Multiphasic Scan
 Arterial Uptake + Washout +Arterial Uptake + Washout −Arterial Uptake −
  • *

    The patient with LGDN was excluded from this analysis.

 Arterial uptake +3129
Washout +
MRIArterial uptake +600
Washout −
Arterial uptake −311

Imaging Techniques in non-HCC and non-DN Lesions.

There was no washout on CT or MRI in the 20 patients without HCC or DN on biopsy (including the 2 patients with CC and EHE). There was an arterial enhancement without washout on CT in 8 cases (40%) and in 9 cases on MRI (45%).

Discussion

For the aim of this study, we recruited a cohort of consecutive patients with CLD who were screened with US, and in whom US detected at least 1 nodule smaller than 2 cm.

First, and unlike other studies, 13% of our patients did not have cirrhosis based on clinical, radiological or histological criteria, but had various degrees of fibrosis, from F1 to F4, according to METAVIR staging. Indeed, most studies focusing on the noninvasive detection of HCC have been performed in patients with cirrhosis.14, 15 Yet, the prevalence of HCC in our study (64%) was similar to that reported by Forner et al.16

When using diagnostic criteria recommended by AASLD in 200514 for the diagnosis of HCC (i.e., arterial uptake and washout on portal or delayed phase), combination of CT and magnetic resonance (MR) (coincidental findings) yielded a sensitivity of 57% and a specificity of 85%. It has been already well shown that sensitivity drops when coincidental findings are required on two imaging modalities.16, 21, 22 But, conversely to these investigators, we observed a lower specificity, which could be related to the presence of HGDNs that behave like HCCs. Although the differential diagnosis between HGDN and early HCC may be difficult at pathology, we performed an independent pathological review with two pathologists, combined with an IHC analysis. Interestingly, most of these HGDNs rapidly progressed toward HCC in the follow-up, outlining the importance to identify and treat them as early as possible. Based on our results and as already stressed in the literature, HGDNs are considered as premalignant lesions.9-11 This is why we pooled HGDNs with HCCs (excluding the patient with LGDN) and evaluated the sensitivity and specificity of CT and MR using the classical imaging criteria based on lesion enhancement.

Our results show that coincidental findings on CT and MRI allow the diagnosis of HCC and HGDNs with a sensitivity of 57% and a specificity of 100%. Despite similar contrast biodistribution of the contrast agents, and similar acquisition timing between CT and MR, we still observed discrepancies in 39% in assessing tumor vascularity, leading to liver biopsy to solve doubtful cases in a high percentage of the cases, as reported by Leoni et al.21

As recently suggested by the 2011 updated AASLD guidelines, we also evaluated the sensitivity and specificity for a noninvasive diagnosis of HCC or HGDN with a single imaging technique (i.e., CT or MRI).21-23 Our results show that sensitivity ranged between 75% and 79%, with the highest sensitivity for MRI, as also observed by others.21 Interestingly, we had a specificity of 100% for HCC or HGDN with any of each single imaging modality.

When a first imaging modality does not provide a conclusive diagnosis, two different diagnostic strategies may be used; first, it was proposed that a sequential imaging strategy be employed, such as CT followed by MRI or the contrary.15 Khalili et al. have recently focused on the optimal imaging technique to diagnose small HCC and have shown that the best sequential diagnostic strategy consisted of MRI followed by CT (if MRI was inconclusive), aiming for a sensitivity of 74%.22 One could argue that such a strategy would still miss the diagnosis of HCC in 26% of the cases and is not cost effective.23 Alternatively, we recommend liver biopsy after a first inconclusive imaging examination. Biopsy remains important in our view, because the information provided both by examination of the nontumoral and tumoral liver should be considered from a prognostic point of view.24, 25 Indeed, the severity of the underlying liver disease frequently influences the therapeutic strategy, and molecular tissue markers of the nontumoral counterpart have been shown to be important prognostic factors for tumor progression and patient survival, as well.26 Moreover, tumor grading is considered to be an accurate predictive factor of survival.27 In addition, the growing interest in targeted therapies makes the accurate characterization of tumor proliferation crucial.28 Therefore, although biopsy may not be mandatory to reach a valid diagnosis in most cases, it provides essential information to evaluate prognosis and define the treatment strategy.

In conclusion, in patients with CLD or cirrhosis with hepatic nodules of 2 cm or less, this study confirms the high specificity and low sensitivity of combined contrast-enhanced CT and MRI for the diagnosis of HCC. We have also shown that the noninvasive diagnosis of HCC or HGDN could be achieved, in most cases, with 100% specificity by a single imaging modality. Because of the high number of disagreements between contrast-enhanced CT and MR, we do not recommend sequential imaging, and we suggest performing a biopsy when the first imaging is inconclusive. In expert hands, the biopsy of a nodule, and of the adjacent liver, provides an accurate diagnosis, helps in the therapeutic outcome, and can identify relevant prognostic factors, based on aspects of the adjacent liver.

Acknowledgements

The authors thank Nathalie Goutté, clinical research assistant, for her excellent technical support.

Ancillary