EASL and AASLD recommendations for the diagnosis of HCC to the test of daily practice.

Aims: To evaluate the diagnostic performance of CT, MRI and CEUS alone and in combination, for the diagnosis of HCC between 10 and 30 mm, in a large population of cirrhotic patients. Patients and methods: In a multicentre prospective trial, 442 patients have been enrolled. Within a month, CEUS, CT and MRI were performed for all patients. A composite algorithm was defined to obtain the more accurate gold standard. Results: A total of 544 nodules in 381 patients have been retained for the performance analysis. Eighty ‐ two percent of the patients were male, mean age was 62 years. For the 10 ‐ 20 mm nodules (n=342), the sensitivity (Se) and specificity (Sp) for the diagnosis of HCC were, respectively, 70.6% and 83.2% for MRI, 67.9% and 76.8% for CT and 39.6% and 92.9% for CEUS. For the 20 ‐ 30 mm nodules (n=202), the Se and Sp were, respectively, 72.3% and 89.4% for MRI, 71.6% and 93.6% for CT and 52.9% and 91.5% for CEUS. The best combination for the 10 ‐ 20 mm nodules was MRI + CT (Se: 55.1%, Sp: 100.0%). After a first inconclusive technique, CEUS as second image technique allowed the highest specificity with only a slight drop of sensitivity for 10 ‐ 20 mm nodules and the highest sensitivity and specificity for 20 ‐ 30 mm nodules. Conclusion: This large multicentre study validates the EASL/AASLD recommendations in daily practice. Specificity using CT or MRI in 10 ‐ 20 mm HCC was low, but we do not recommend combined imaging at first as sensitivity would be very low. The best sequential approach combined MRI and CEUS.

Contrast ‐ enhanced ultrasound; hepatocellular carcinoma; magnetic resonance imaging; multidetector computed tomography

Abbreviations

AASLD American Association for the Study of the Liver

AISF Italian Association for the Study of the Liver

APASL Asian Pacific Association for the Study of the Liver

CEUS contrast ‐ enhanced ultrasound

CT computed tomography

EASL European Association for the Study of the Liver

EORTC European Organisation for Research and Treatment of Cancer

HCC hepatocellular carcinoma

MRI magnetic resonance imaging

NICE British National Institute for Health and Clinical Excellence

Key points

EASL ‐ AASLD recommendations are validated in clinical practice in a large prospective multicentric study

MRI is the best examination to diagnose HCC using EASL ‐ AASLD criteria but its specificity is not excellent for 10 ‐ 20 mm nodules

Diagnostic performances of EASL ‐ AASLD criteria are similar in patients with history of previously treated HCC than in the whole population

CEUS could have a role in a two ‐ step algorithm for diagnosing non ‐ invasively HCC

.

In a cirrhotic liver, a probabilistic diagnosis of hepatocellular carcinoma (HCC) is possible, based on imaging in accordance with international recommendations.[1] , [2] , [3] Since the possibility of curative treatment and prognosis of the disease is related to the size of the lesion,[4] it is desirable to diagnose small HCCs that could benefit from curative treatments.[5] Because a typical feature of HCC in imaging is based on vascular changes from a regenerative nodule to HCC,[6] and as these changes are gradual, it is harder to characterize small HCCs using imaging criteria than large ones.

In the same way, the choice of different imaging techniques to detect typical HCC features, and therefore non ‐ invasive HCC diagnosis is still debated.

The updated European (EASL) and American Association for the Study of the Liver (AASLD) recommendations have endorsed the sequential application of computed tomography (CT) scans and magnetic resonance imaging (MRI) to diagnose HCC. A single imaging technique is now considered adequate for the diagnosis of HCC larger than 1 cm in size.[1] , [2] But EASL ‐ EORTC guidelines mention that two concordant imaging techniques are still recommended to diagnose nodules between 10 and 20 mm in non ‐ expert centres. Moreover, the most recent European and North American guidelines dropped contrast ‐ enhanced ultrasound (CEUS), since this imaging modality is considered inaccurate in distinguishing HCC from intrahepatic cholangiocarcinoma.[1] , [2] Yet, CEUS is still part of the HCC diagnostic algorithm recommended by the Asian Pacific Association for the Study of the Liver (APASL)[3] and the Japanese Society of hepatology[7] , [8] and has recently been reintroduced by the German Roentgen society (www.drg.de),[9] the British National Institute for Health and Clinical Excellence (NICE) (www.nice.org.uk) and the Italian association for the study of the liver (AISF).[10]

Indeed, these recommendations are based more on expert opinions than controlled studies. Only a few single ‐ centre studies[11] , [12] , [13] , [14] , [15] , [16] , [17] have prospectively validated these recommendations. Moreover, apart from the studies by Khalili et al., Sangiovani et al. and Furlan et al. that included, respectively, only 101, 67 and 96 nodules, none has compared the diagnostic accuracy of three contrast ‐ enhanced examinations (CEUS, CT and MRI). Finally, in the Furlan et al. series only nine patients had both CT and MRI.

Therefore, the present multicentre prospective trial aims to evaluate the diagnostic performance of CT, MRI and CEUS alone and in combination, for the diagnosis of HCC between 10 and 30 mm, in a large population of patients with cirrhosis.

This study was approved by the Ethics committee CPP Ouest II and registered as # NCT00848952. Informed consent was obtained from all the patients. This study was supported by a national institutional grant (PHRC 2008 ‐ 01).

From April 2010 to April 2013, 442 patients were prospectively enrolled in 16 expert centres. The inclusion criteria were:

Detection of de novo liver nodules in patients with compensated cirrhosis under surveillance; Nodule size between 10 to 30 mm in the examination technique used, leading to inclusion (US, CT or MRI); more than one nodule could be present but none of them should exceed 30 mm; the patient could be naïve of HCC or have already been treated for an HCC by surgery or thermal ablation. In these cases, the new nodule should be more than 2 cm from the resection or ablation area. Previous treatment by transarterial chemoembolization during the last 5 years was an exclusion criterion.

To evaluate the proportion of patients included through a first CT or MRI and the proportion of patients with HCC in this sub ‐ population, a retrospective analysis has been conducted in the three main centres.

A total of 684 nodules were identified in 442 patients. A total of 140 nodules were excluded for several reasons: (i) 64 nodules with obvious characteristic images of pseudolesions or benign non ‐ hepatocellular lesions (arteriovenous shunt, haemangioma, focal fatty sparing in fatty liver) found after the first examination performed for this study; (ii) 69 nodules because they were smaller than 10 mm or larger than 30 mm (n=2); (iii) in 7 nodules, a definitive diagnosis could not be reached (nodule, which turned out to be an HCC at 9 or 12 months follow ‐ up but without data for 3 and 6 months follow ‐ up). The diagnostic performance of CT, MRI and CEUS was therefore calculated on 544 nodules in 381 patients. A flow chart of this study is shown in Figure [NaN] .

After finding the nodule(s), the three imaging techniques (CEUS, MDCT, MRI) had to be performed within a month. There was no specific order for performing the three examinations.

Description and technical parameters of each technique are reported in Appendix [NaN] .

For each modality (US, CT or MRI), the image analysis performed by the radiologist was done unblinded to the screening technique (the size and the location of the nodule were reported on request) but was blinded to the other examinations performed in the study.

After contrast medium injection, enhancement in the different vascular phases was notified. Enhancement patterns are reported in the Appendix [NaN] .

Biopsy of all the target nodules was not mandatory for HCC diagnosis during the study. In particular, a liver biopsy was performed according the EASL ‐ AASLD guidelines when typical hallmarks of HCC were not met on CT or MRI.

Technical details of the biopsy and histological analysis are reported in the Appendix [NaN] .

Gold standard was the main limitation of all the previous studies published regarding the performance of imaging for the diagnosis of HCC.[11] , [12] , [13] , [14] , [15] , [16] , [17] Three criteria were assumed to build this algorithm: (i) The only absolute reference would be histology after surgical resection (or transplantation) but this was not realistic. (ii) Histology on biopsy is a perfect gold standard when positive but not when negative. (iii) The high specificity of the typical HCC hallmark using two different contrast ‐ enhanced imaging techniques make it possible to use imaging modalities as reference methods in patients with no histological proof..[10] , [17] , [18] The goal of the algorithm was to classify lesions as “HCC” or “not HCC” (Figure [NaN] ). We did not aim to get all the diagnoses of non ‐ HCC lesions. This step ‐ by ‐ step algorithm was built to obtain high specificity.

Hepatocellular carcinoma diagnosis was obtained firstly by histology if the nodule was treated by surgery (resection or transplantation), or if a biopsy was performed within 6 months of the date of the last imaging examination performed.

If the diagnosis of HCC was not provided by histology the policy was decided according to the imaging pattern. At least two positive examinations were needed to diagnose a nodule as a HCC.[19] If the nodule remained negative for the HCC diagnosis, we took into account an AFP serum level of more than 200 ng/mL. For a solitary nodule, the nodule was considered to be a HCC regardless of its size and its imaging pattern, whether hypervascular or hypovascular. If there were two or three nodules, only nodules larger than 2 cm were considered to be HCC.[20]

If the diagnosis of HCC remained negative at this stage, the nodule was monitored at 3, 6 and 12 months. If any of the following changes were seen within 6 months, the diagnosis of HCC was made: (i) increase in size more than 30% in maximal diameter (mean growing size=74%) or appearance of hyperarterial enhancement or washout. The remaining nodules were considered to be non ‐ HCC nodules. To ensure a high specificity for the diagnosis of HCC, we did not take into account, the nodules (n=6) that displayed changes suggestive of HCC at 12 months. Figure [NaN] and Table [NaN] shows the number of HCC diagnosed at each stage and according to each imaging pattern.

Detailed version of the Statistical analysis section is reported in the Appendix [NaN] .

To evaluate the performance of the three imaging techniques for the diagnosis of HCC, the diagnosis was considered positive if the nodule displayed the typical features of HCC. If the nodule was not seen in an examination, the technique was considered as negative for the diagnosis of HCC. We obtained exact 95% CIs for sensitivity and specificity from the binomial distribution. Cochran's Q ‐ tests were performed to compare sensitivities and specificities among the three examinations and McNemar test was applied to test the performance between techniques pairwise.

Diagnostic performance of a combination of two of the three imaging techniques (CEUS+CT; CEUS+MRI; CT+MRI) was evaluated. In these situations, to be positive, the diagnosis of HCC needed both techniques to display a typical feature of HCC. For this calculation, patients who had two imaging tests as reference standard for the diagnosis of HCC were excluded.

Diagnostic performance of a sequential combination of two of the three imaging techniques (CEUS+CT; CEUS+MRI; CT+MRI) was evaluated following two ways. Firstly, the diagnosis of HCC was made with typical imaging features of HCC on at least one imaging modality. Secondly, the performance of the second imaging technique was calculated on the population that was negative (no typical imaging features of HCC) at the first examination.

Characteristics of the population

1 A patient could have multiple aetiologies.

2 60 missing data. High AFP serum means above the upper normal limit.

Three hundred and eighty ‐ one out of 442 (86%) patients with 544 nodules were finally included. There were 314 (82%) males and 67 (18%) females, with a mean age of 62±10 years. Among them, 39 (10%) with 57 nodules had a history of HCC already treated.

The causes of cirrhosis were mainly alcohol (58% of the cases), hepatitis C virus (32%) and metabolic syndrome (19%). The mean size of the nodules was 18.1±5.7 mm; 342 nodules were ≥10 and <20 mm, while 202 were ≥20 and ≤30 mm; 128 (33%) of the patients had more than one nodule.

The final diagnosis was HCC for 342 lesions (63%), and 295 (77.4%) patients had at least one HCC. One hundred and eighty ‐ seven out of 342 (55%) nodules between 10 and 20 mm and 155/202 (77%) nodules between 20 and 30 mm were HCCs, 67 patients (18%) had HCC and non ‐ HCC nodules.

Of the 544 nodules, 145 (27%), 107 (20%) and 131 (24%) were hypervascular in the arterial phase but did not exhibit washout on MRI, CT and CEUS respectively. Of those, 53 (37%), 44 (41%) and 96 (73%), respectively, were HCC.

Forty ‐ nine nodules were resected. Among them, 36 were HCC. Other nodules were: regenerative nodules (n=5), dysplastic nodules (n=4), cholangiocarcinoma (n=1), toxocara canis granuloma (n=1), post ‐ RF nodules (n=2).

Two hundred and forty ‐ one nodules were biopsied. Among them 159 were HCCs and 29 non ‐ HCCs. Non ‐ HCC nodules were: regenerative nodules (n=7), dysplastic nodules (n=12), cholangiocarcinoma (n=2), hemangioma (n=3), metastasis (n=1), focal nodular hyperplasia (n=1), tumoral necrosis (n=1) and no specified carcinoma (n=2). In 53 nodules, histology was considered inconclusive (cirrhotic liver or non ‐ tumoral).

The proportions of nodules found through CT or MRI in the three main centres were 21% [range 7% ‐ 33%] and 14% [range 8.5% ‐ 19%] respectively. Among these nodules, the percentage of HCCs were, respectively, 58% [range 40% ‐ 75%] and 52% [range 38.5% ‐ 75%].

The three imaging techniques were performed in a mean time of 17±20 days. Performances of each imaging technique according to nodule size are shown in Table [NaN] .

Performance of arterial enhancement followed by washout for each examination for the diagnosis of HCC

3 ND, not done because Cochran's Q test was not significant.

For all nodules (10 ‐ 30 mm), MRI, CT and CEUS sensitivity for HCC diagnosis was 71.4% [66.2; 76.1], 69.6% [64.4; 74.4] and 45.6% [40.3; 51.1] respectively; and specificity was 84.7% [78.9; 89.3], 80.7% [74.6; 85.9] and 92.6% [88.1; 95.8] respectively.

Whatever the size of the nodules (10 ‐ 20 or 20 ‐ 30 mm), sensitivity was similar using CT and MRI. CEUS sensitivity was lower than that of CT or MRI, for 10 ‐ 20 mm nodules (P<.0001) and for 20 ‐ 30 mm nodules (P<.0015).

For the larger nodule group (20 ‐ 30 mm), specificity was high and almost equivalent for the different techniques (89.4% to 93.6%) but with smaller nodules, CEUS specificity was significantly higher than that of CT and MRI (92.9% vs 76.8% and 83.2%, respectively, P<.0016).

As some nodules have been detected firstly by CT or MRI and could not have been seen on ultrasound, we also calculated the sensitivity of CEUS taking in account only the nodules visible at ultrasonography. This lead to an increase in sensitivity of CEUS from 39.9% to 45.7% for the 10 ‐ 20 mm nodules and from 52.4% to 58.2% for the 20 ‐ 30 mm nodules (Table [NaN] ).

Because resection is the only perfect gold standard, we have studied the performance of the three imaging techniques in the sub ‐ population of nodules that have been resected. We have found a sensitivity close to that of the whole population for 10 ‐ 20 mm nodules and slightly better for the 20 ‐ 30 mm nodules. Specificity was very similar for 20 ‐ 30 mm nodules and higher in nodules between 10 and 20 mm, especially for MRI. The results are reported in Table [NaN] .

Performance of arterial enhancement followed by washout for each examination for the diagnosis of HCC in patients who had surgical resection as method of reference

In the sub ‐ group of patients previously treated for HCC, the performances of MRI and CT were not clearly different from those in the whole population. The sensitivity and the specificity were a little bit lower for MRI and CT ( about 5% and 10% respectively) and appeared to be increased for CEUS. But because of the low number of patients (especially for 10 ‐ 20 mm nodules) this was not statistically significant. These results are detailed in the Table [NaN] .

We evaluated the performance of a combination of two out of three of the techniques for nodules between 10 and 20 mm, following EASL and AASLD 2005 recommendations and those from EASL 2012 for non ‐ expert centres. Combination was positive when both techniques showed the hallmark pattern of HCC. For this statistical analysis, patients whom had two imaging tests as reference standard for the diagnosis of HCC were excluded, because in this situation, 100% specificity would be systematically reached. Results for the different combinations are reported in Table [NaN] .

performances of the combination of examinations for the diagnosis of HCC with the exclusion of patients who had two imaging tests as reference standard for the diagnosis of HCC. Combination was positive when both techniques showed the typical pattern of HCC

• 4 ‐ , could not be calculated because Sp=100%.
• 5 ND, not done because Cochran's Q test was not significant.
• 6 Performances when two (whatever of the technique) or the three examinations were positive.

Whatever the size of the nodule and the different imaging combinations, sensitivity was weak (between 27.8% and 56.1%). The best sensitivity was obtained by combining MRI and CT. On the other hand, the different combinations had extremely high specificities (99.4% to 100%).

We also evaluated the benefit of sequential strategies taking into account the first performed examination. Table [NaN] shows the performance of a sequential combination of two of the three imaging techniques (CEUS+CT; CEUS+MRI; CT+MRI). Sequential combination was positive when either one or the other technique showed the typical pattern of HCC. In this situation, the sensitivity was increased but with a slight decrease in specificity.

Performances of the different sequential combinations of examinations. Combination was positive when either one or the other technique showed the typical pattern of HCC

Table [NaN] shows the performance of the second imaging technique calculated on the population that did not display typical features of HCC at the first examination. In 10 ‐ 20 mm nodules, CEUS used as the second examination had the highest specificity with only a slight drop of sensitivity. In the 20 ‐ 30 mm nodules, CEUS in second line lead to the highest sensitivity and specificity whatever the first inconclusive imaging technique (CT or MRI).

Performances of the second examination for the nodules not diagnosed as HCC by a first one regarding the first and the second examination performed

Non ‐ invasive strategy for HCC diagnosis, as recommended by EASL and AASLD guidelines, is widely used in clinical practice but has rarely been evaluated.[11] , [12] , [13] , [14] , [15] , [16] , [17] This is especially the case for the updated version published in 2012, which proposes a sequential strategy with a single imaging technique. Moreover, CEUS has been dropped out of the most recent EASL guidelines.[3] , [16]

Our study is the only large, multicentre, prospective one to have evaluated firstly, the diagnostic performance of each imaging technique that could be used for non ‐ invasive diagnosis of HCC, and secondly, the different algorithms proposed by the EASL and AASLD guidelines in 2005[18] and its recent update in 2012.[2] In this study, CT and MRI sensitivity in HCC diagnosis based on hypervascularity and washout were 70% and 71% (p: ns) respectively. These sensitivities were not decreased in nodules smaller than 20 mm. These results are in line with previous publications for nodules smaller than 20 mm (44% ‐ 75% for CT and 65% ‐ 80% for MRI).[11] , [12] , [13] , [15] , [21] Yet, sensitivities for nodules between 20 and 30 mm, reported in our results, were lower than those reported in the literature. Indeed the studies[14] , [15] , [16] , [17] conventionally define cut ‐ off at 20 mm, with two groups: smaller and larger than 20 mm with no upper limit. Thus, imaging performance for nodules larger than 20 mm is probably raised by the largest nodules. Whatever the size, we found that MRI was not significantly more sensitive than CT.[16] , [22]

The specificity for both CT and MRI found in our study was lower than expected and differed according to size stratification. Although it was good for 20 ‐ 30 mm nodules (>89%), it fell for the smaller nodules (10 ‐ 20 mm) to 77% and 83% for CT and MRI respectively. Recent studies have already reported low specificities for MRI. Sersté et al.[21] found a specificity of 85% for MRI and Di Martino et al.[15] reported a specificity of 91% for MRI in 10 ‐ 20 mm nodules. This issue is critical, as a significant number of false ‐ positive cases will be considered to be HCC and consequently treated. It has been already shown that some of these false ‐ positive cases are high ‐ grade dysplastic nodules.[21] We hypothesize that the main cause of this lower specificity could be related to overinterpretation of washout. The findings of these imaging techniques could have been assessed differently by radiologists since examinations were read on ‐ site and not centrally. The poor correlation of washout readings has already been reported by Furlan et al.,[14] Davenport et al.[23] and Sofue et al.[24] Indeed, centralized reading could have overcome this limitation but our results would not have reflected the daily practice.

Another point that could be highlighted is the almost similar performances of EASL ‐ AASLD criteria for nodules in already treated patients in comparison to whole population, which reinforces the usefulness of these criteria in all clinical situations. Nevertheless, we had less than 60 patients in this category and our results should be further confirmed.

Finally for the 42 nodules, which were resected, the results were close to that observed in the whole population. This confirms that the composite gold standard was appropriate in this study and that diagnostic performance was not altered in surgically confirmed patients. Therefore, we should notice that including only nodules submitted to resection would probably increase the number of true ‐ positive cases because in the vast majority of the cases only nodules with a high likelihood of HCC are resected.

In the current guidelines, regarding the similar sensitivities of CT and MRI for nodules between 10 and 30 mm, the highest specificity of MRI indicates the use of this technique as first line modality.

We also looked at the diagnostic performance of CEUS. Our results show low sensitivity for any lesion size. Inclusion of nodules seen firstly by CT or MRI and that were potentially non ‐ visible at ultrasound is not enough to explain this, since taking into account only nodules seen on ultrasound increased only moderately (10%) the sensitivity of CEUS. On the contrary, the specificity of CEUS was quite high (92.9%), and significantly higher than CT and MRI for the 10 ‐ 20 mm nodules. The role of CEUS has been largely debated since it was removed from the AASLD and EASL/EORTC guidelines. The main reason for its removal is the report of similar features for HCC and intrahepatic cholangiocarcinoma using CEUS.[25] Yet, recently, because of its high positive predictive value >95% for HCC when displaying a typical pattern,[7] , [13] CEUS has been reincorporated in the guidelines of some European countries (Germany, UK and Italy) ([9] , [10] www.drg.de, www.nice.org.uk). In this study, the highest specificity of CEUS is in agreement with the recent studies by Manini et al.[16] and the review by Jang et al.[26] , both promoting the role of CEUS at the same time as the depiction of a nodule during ultrasound surveillance. Jo et al.[27] In a recent review suggest that CEUS could take place also in the characterization of indeterminate masses found on CT and MRI in cirrhotic patients. Indeed, the low sensitivity of CEUS[28] means that CT or MRI should be performed in more than half of patients to further characterize nodules that do not meet typical features on CEUS.

Another way to increase specificity for the diagnosis of HCC (especially in nodules between 10 and 20 mm) is to combine two imaging modalities. Our results confirm previous data reported in the literature, with specificity between 99.4% and 100% for any combination. Yet, as already reported, the sensitivity of combined imaging drops to unacceptable values. This explains the sequential strategy used in the EASL/EORTC and AASLD guidelines allowing HCC diagnosis in cirrhosis based on a single technique (CT or MRI).

To choose the best sequential approach, we evaluated the performance of the different imaging techniques in second line when a first one did not show the typical features of HCC. To the best of our knowledge only Khalili and al. have performed such analysis.[17] They showed that this sequential strategy improves sensitivity with a mild drop in specificity. But, as they underlined, this strategy appears interesting as long as both imaging scans have high specificity. In our study, because of a lower specificity of individual imaging techniques, the specificity reached by the sequential strategy was lower for the 10 ‐ 20 mm nodules than in the study of Khalili and al. Again, in this situation, it emphasizes the high specificity of CEUS to diagnose HCC especially when performed after a negative MRI, for 10 ‐ 20 mm nodules.

Based on the results of this study, we might propose the following sequential approach: MRI should be proposed as first line examination whatever the size stratification (10 ‐ 20 mm or 20 ‐ 30 mm). If MRI is inconclusive for the diagnosis of HCC, we propose CEUS, because its specificity is excellent and the cost of slight drop of sensitivity limited to 10 ‐ 20 mm nodules.

Our study suffers from some limitations. Firstly, like other studies we used a composite reference method for the diagnosis of HCC including imaging. Yet, in more than half of the cases, the diagnosis was obtained by histology. Considering only an absolute reference (as surgical resection) would have introduced more bias. Gold standard classically used is histology and follow ‐ up, but this gold standard also has limitations, in particular, with negative biopsy that can be because of sampling error or mistargeting. Furthermore, follow ‐ up is not always solving problem because histology of the nodule at the time of the imaging could never been affirmed. Instead, the composite reference method we built was a step ‐ by ‐ step process taking into account only an absolute specificity of diagnostic tests. The similar diagnostic performance of imaging using surgical resection as reference compared to the whole population confirmed the appropriateness of our composite reference method. Secondly, the lack of centralized reading might have led to misreading of washout. But on the other hand our results of this large multicentre study are close to those of daily clinical practice at expert centres.

Finally, this study included proportion of patients under surveillance using CT and MRI, while the AASLD and EASL criteria deal specifically with nodules found using US surveillance. But in daily practice, ultrasound may be limited in some patients and CT or MRI may find new nodules especially when they are performed in the follow ‐ up of treated HCC. Therefore, this study is not an exact assessment of the AASLD and EASL guidelines, but the extension of these criteria to the daily practice.

This large multicentre study confirms in daily practice the high specificity of any of the three imaging techniques for HCC between 20 and 30 mm, and therefore we validate the EASL ‐ AASLD recommendations for these lesions. We found a drop in specificity using CT or MRI in 10 ‐ 20 mm HCC but we do not recommend systematic combined imaging at first as sensitivity would be very low. The best sequential approach combined MRI and CEUS.

We sincerely acknowledge all of the technicians of clinical research and especially Marie Bost and the project manager of the CHIC study Pascaline Aucouturier.

No conflict of interest.

Jean Pierre Tasu: Département de radiologie, CHU Poitiers, 86000 Poitiers, France

Christine Sylvain: Service d'Hépato ‐ Gastroentérologie, CHU Poitiers, 86000 Poitiers, France

Alain Luciani: Imagerie Medicale, AP ‐ HP, Groupe Henri Mondor Albert Chenevier, Creteil, F ‐ 94010, France

Christophe Duvoux: Service d'Hépatogie, Hôpital Henri Mondor APHP, Université Paris Est Créteil ‐ France

Frank Pilleul: Centre LCC ‐ Léon Bérard, 28 rue Laennec, 69008 Lyon, France

Jérôme Dumortier, Fédération des Spécialités Digestives, Hôpital Edouard Herriot, Université Claude Bernard Lyon 1, Lyon, France

Guillaume Baudin: Service de radiologie et imagerie médicale, Hôpital Archet, CHU Nice, 06200 Nice Nice, France

Rodolphe Anty: Service d'Hépato ‐ gastroentérologie, Hôpital Archet, CHU Nice, 06200 Nice Nice, France

Patrick Chevallier: Service de radiologie et imagerie médicale, Hôpital Archet, CHU Nice, 06200 Nice Nice, France

Vincent Le Pennec: Service de Radiologie diagnostique et thérapeutique, CHU Côte de Nacre, 14033 Caen, France

Isabelle Ollivier Hourmand: Service d'Hépato ‐ gastro ‐ entérologie, CHU Côte de Nacre, 14033 Caen, France

Louis Estivalet: Radiologie et Imagerie médicale diagnostique et thérapeutique, CHU Dijon, 21079 Dijon, France

Anne Minello: Département d'hépatologie et gastroentérologie, CHU Dijon, Université Burgundy, INSERM U866, 21079 Dijon, France

Eric Frampas: Service Central de Radiologie et Imagerie Médicale, Hôtel Dieu, CHU Nantes, 44093 Nantes, France

Isabelle Archambeaud: Service d'hépato gastroentérologie, Hôtel Dieu, CHU Nantes, 44093 Nantes, France

Anne Guillygomarc'h: service des maladies du foie, CHU Pontchaillou, Rennes, France

Vanessa Brun: Imagerie médicale, CHU Pontchaillou, 35033 Rennes, France

Yves Gandon: Imagerie Médicale, CHU Pontchaillou, 35033 RENNES, France

Muriel Cuilleron, Service de radiologie, Hopital Nord, CHU, 42055 Saint ‐ Etienne, France

Brigitte Patouillard: Service d'hépato gastroentérologie, Hopital Nord, CHU, 42055 Saint ‐ Etienne, France

Boris Guiu: Département de Radiologie, INSERM U896, CHU Saint ‐ Eloi, Université de Montpellier, 34295 Montpellier, France

Benoit Gallix: 1) Département d'Imagerie Medicale, CHU Montpellier, CHU Saint ‐ Eloi, Université de Montpellier, 34295 Montpellier, France. 2) McGill University Health Center, Montreal, QC, Canada

Jérôme Lebigot: Département de radiologie, CHU d'Angers, 49933 Angers, France

.

Graph: Flow chart and gold standard (following the STARD statement)

Graph: Step by step algorithm of the reference methods

Graph