| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Full Paper |
1 Liver Transplantation and Hepatobiliary Medicine, The Royal Free Hospital, Pond Street, London NW3 2QG and 2 Department of Radiology, The Royal Free Hospital, Pond Street, London NW3 2QG, UK
Correspondence: Prof. A K Burroughs
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
FP) is controversial as definitive evidence for survival benefit in cirrhotic patients has only been shown in some retrospective or uncontrolled prospective studies [2–4]. There are no randomized controlled clinical trials concerning surveillance of HCC in patients with chronic liver disease, which show a reduction of mortality. However, a higher chance of receiving treatment is shown amongst patients with HCC diagnosed by surveillance in an Asian retrospective study [5]. Irrespective of the lack of evidence, surveillance for HCC is widely practised as it is recognised that all treatment modalities yield better results in smaller and unifocal tumours [6]. An expert panel on HCC from the European Association for the Study of the Liver (EASL) recommends that cirrhotic patients should undergo surveillance if potentially effective treatment (i.e. surgical resection, liver transplantation and percutaneous ablation) for HCC can be offered to them for tumours detected early [6]. Therefore, it is not uncommon to find a focal hepatic lesion on US or following an elevated FP value during surveillance in cirrhotic patients.
The further investigation of an abnormal US during surveillance programs or an incidental finding of a suspicious nodule on US is based on a wide range of radiological investigations. There is no consensus regarding the best next investigation to confirm, characterize, and detect any further HCC lesions. This point is also discussed in a very recent paper studying HCC imaging [7]. According to the non-invasive diagnostic criteria for HCC in cirrhotic patients defined in an EASL monothematic conference [6], a focal lesion greater than 2 cm diameter with arterial hypervascularization as shown by two coincident imaging techniques or a focal lesion greater than 2 cm with arterial hypervascularization as shown by one imaging technique with associated FP level greater than 400 ng ml–1 is diagnostic of HCC in cirrhosis. Further imaging techniques include spiral CT, MRI, contrast ultrasound and angiography. For the diagnosis of HCC not exceeding 2 cm diameter, biopsy of the nodule is recommended because the imaging techniques do not have sufficient accuracy to distinguish HCC from other benign or malignant conditions and the FP level is usually within normal values or is only slightly elevated [6].
Apart from the imaging techniques mentioned above, there are less invasive ones such as US, CT and MRI. More invasive ones include iodized oil-CT, CT during hepatic arteriography (CTHA), CT arterial portography (CTAP) and conventional hepatic angiography.
It is important to confirm the diagnosis of HCC and to assess the number of tumour nodules accurately as this can affect the choice of subsequent treatment modalities. For orthotopic liver transplantation, the survival rates are better for patients with a single tumour 5 cm or less in diameter or three tumour nodules, each 3 cm or less in diameter, based on pathological examination of the explanted livers [8]. Similar findings are also found in other studies [9, 10]. Amongst those patients planned for liver resection, it is also important to assess the number of nodules to plan the feasibility and extent of resection.
For histological diagnosis of HCC, percutaneous needle biopsy is an efficient procedure. However, it has complications. In one retrospective study [11], needle tract seeding occurred in 9 (2.1%) and internal bleeding occurred in 5 of 420 (1.2%) patients. There are also case reports of needle tract recurrence after needle aspiration rather than cutting biopsy [12, 13]. The sensitivity and specificity of US guided fine-needle aspiration biopsy for HCC are only 69% and 93%, respectively [14]. Taking into account the potential risk of biopsy and its accuracy, a diagnostic biopsy should only be needed if the result has a clinical impact or if there is a considerable risk in the treatments planned, against the background of a possible false positive diagnosis based solely on imaging techniques [6].
The aim of this article is to review systematically the evidence currently available for determining the best radiological investigation for confirming and characterizing HCC when a suspicious nodule is detected by ultrasound in a cirrhotic patient.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Types of studies included
Only prospective cohorts of patients with either prospective or retrospective review of films by radiologists, and retrospective cohorts of patients were included. Case reports and series of less than 10 or unknown number of patients, and studies limited to assessing patients with previous therapy (radiofrequency ablation, transcatheter arterial embolisation or percutaneous ethanol injection) were excluded. Studies involving imaging for HCC patients were also excluded if the aims of the studies were not explicitly defined in terms of lesion characterization. Furthermore, some imaging studies on HCC included patients with and without cirrhosis. These studies were excluded if no separate data on the imaging results in the subgroup of cirrhotic patients with HCC could be extracted. Studies on imaging performed solely on explanted livers were excluded.
Methods of review
Studies considered for inclusion were examined to determine their relevance to the objectives of the review. Data were recorded in data extraction sheets. Variables recorded included study design, radiological imaging modality, number of patients, proportion of cirrhotic patients, reference standard for diagnosis of HCC and outcome measures of the radiological imaging modality concerned. Reference standards for diagnosis of HCC included histological findings from the explanted liver, liver resection or needle biopsy; confirmatory radiological findings from US, CT, MRI or angiography as defined by the authors; elevated FP level as defined by the authors; and follow up for tumour growth. Time lag between imaging and liver transplantation was also recorded if available.
Types of outcome measures
Data on diagnostic impact, diagnostic performance, therapeutic impact, patient outcome and health economics of each imaging technique were extracted. For diagnostic impact and diagnostic performance, the sensitivity, specificity, positive predictive value and negative predictive value of each imaging modality to detect and characterize HCC were recorded, if they had been reported. Diagnostic performance referred to the performance characteristics of a single radiological imaging modality. Diagnostic impact referred to the investigation of the best radiological imaging modality in terms of its diagnostic performance, when several were compared directly in a study. Sensitivity of an imaging technique was either expressed as the percentage of patients with HCC detected (patient-by-patient sensitivity) or as the percentage of HCC nodules detected (lesion-by-lesion sensitivity). Specificity of an imaging technique was either expressed as the percentage of patients correctly identified to have no HCC (patient-by-patient specificity) or as the percentage of non-HCC nodules correctly identified (nodule-by-nodule specificity). For data on health economics, patient outcome and therapeutic impact, a descriptive summary for each study was recorded.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Studies addressing diagnostic impact and diagnostic performance of imaging techniques
These articles were classified according to whether cirrhotic patients alone constituted the whole study population, or whether HCC per se or various types of focal hepatic lesion were the focus of interest. The reference standard used for the diagnosis of HCC was either an elevated FP, or follow-up for tumour growth, or confirmatory imaging as defined by the authors, or needle biopsy, surgical biopsy or explanted liver histology.
Explanted liver histology as reference standard: studies limited to hepatocellular carcinoma in cirrhotic patients
10 studies with 159 patients fulfilled these criteria. In six studies [15–20], only a single imaging modality was assessed so that solely diagnostic performance was addressed. The other four studies had a direct comparison of two or three imaging modalities [21–24] and therefore addressed both the issues of diagnostic impact and diagnostic performance for each imaging modality. The results are summarized in Tables 1 and 2
.
|
|
Bizollo and colleagues [17], in a subgroup of 6 patients known to have HCC, the lesion-by-lesion sensitivity of iodized oil-CT was 33% probably because no daughter nodules, found in the explanted liver, could be detected by iodized oil-CT. For the group of 66 patients without known history of HCC, lesion-by-lesion sensitivity was 40%.
A potential factor affecting the diagnostic performance of an imaging technique is the interval from imaging to transplantation. In the study by Bizollo and colleagues [17], the sensitivity of iodized oil-CT among patients transplanted within 3 months of imaging was 42% but that among patients transplanted after 3 months of imaging was 37%. No statistical test was performed to assess the significance of the difference. Although data from three studies [15, 17, 23] investigating iodized oil-CT reported the mean duration from imaging to transplantation, the protocol of performing iodized-oil CT differed from one to another in terms of the amount of iodized oil injected, CT scanner model and scan thickness. It was difficult to draw firm conclusions from these three studies regarding the effect of the interval from iodized oil CT imaging to transplantation on the sensitivity of iodized oil CT. As regards to other imaging techniques, the effect of the interval from imaging to transplantation was not investigated in these 10 studies [15–24].
Four explant studies [15, 17, 21, 23], all of which were prospective, reported the diagnostic performance of iodized oil-CT for HCC. In the study by Taourel and colleagues [15], nodules less than 8 mm in size were not examined microscopically according to protocol. Detected nodules were larger than missed nodules (mean diameter: 27.3 mm versus 16.6 mm), though the difference was not assessed by any significance test. Hence, the sensitivity of iodized oil-CT could be overestimated in this study. This point was also illustrated in a study [21] in which the sensitivity of iodized oil-CT was just 7%, but the HCC nodules were much smaller than the HCC reported in other three studies. In addition, Spreafico and colleagues [23] found that the detection rate was increased significantly only in HCC nodules greater than 2 cm in diameter or with a hypervascular pattern at angiography (p<0.0001).
Two studies [16, 22], both prospective, assessed the performance of spiral CT. In one study [16], sensitivity for tumours more than 2 cm (82%) was greater than that for tumours less than 2 cm in size (60%) though no significance testing was performed to assess the difference.
Four studies [19, 21, 22, 24] assessed the performance of grey-scale US in pre-transplant patients. The highest sensitivity value was obtained in the retrospective cohort [24]. Again, the lowest sensitivity rate was found in the study [21] in which the size of HCC nodules was the smallest. According to Kim and colleagues [19], sensitivities for HCC nodules greater or less than 2 cm were 38% and 30%, respectively.
Two prospective studies [20, 22] assessed the performance of MRI in pre-transplant patients with cirrhosis but no known HCC history. Unfortunately, specificity could not be compared directly as one was presented on patient-by-patient basis [20] and another one was presented on nodule-by-nodule basis [22].
For the diagnostic performance of angiography for HCC amongst cirrhotic patients, three prospective studies [18, 21, 23] were identified. In one study [21], it is not mentioned clearly whether the digital subtraction technique had been used or not. In one study [18], seven patients were found to have more than 10 HCC nodules at pathology and only the 10 largest lesions were counted. For lesions greater than 2 cm, it was 100%; and for lesions 1–2 cm, it was 39%; and for lesions less than or equal to 1 cm, it was 12%. Unfortunately, no statistical analysis on these differences was performed. The relationship between the size of HCC nodules and sensitivity of DSA was not reported in the remaining two studies [21, 23].
Non-liver explant studies limited to hepatocellular carcinoma in cirrhotic patients
There were 16 studies [25–40] with 607 patients fulfilling these criteria. In six studies [25–30], there were direct comparisons between different modalities of imaging, so both diagnostic impact and diagnostic performance of imaging modalities could be addressed. In 10 studies [31–40], only one modality of imaging was studied in each of them, so only data for diagnostic performance could be obtained. In these 16 studies, true patient-by-patient specificity cannot be determined as they did not use the explanted liver as the reference standard for diagnosis and the whole liver could not be examined by other histological methods. Surgical resection or needle biopsy only provides histological findings from the part of the liver from which they are taken. HCC detection by radiological imaging depends to a great extent on the vascularity of the tumour. In this group of studies [25–40], the reference standard for diagnosis included surgical specimen, needle biopsy, confirmatory imaging findings from iodized oil-CT, spiral CT, CTAP, CTHA or angiography, elevated FP level and follow up for tumour growth. Salient results of this group of studies are summarized in Tables 2 and 3
.
|
Two studies [32, 33] evaluated spiral CT alone. In the first [32], sensitivity of HCC detection was up to 100% from results of imaging with the four phase CT in tumours greater than 2 cm in size, 93% in tumours 1–2 cm in size, and 60% in tumours less than 1 cm in size. However, no statistical test was performed to assess the difference. The major limitation of this study was that not all the nodules were confirmed histologically.
Gadopentetate dimeglumine-enhanced dynamic MRI was investigated in two studies [29, 30]. Again, the limitation of the first study [29] was that not all lesions had histological confirmation. Some lesions had the diagnosis of HCC confirmed by radiological follow up for tumour growth.
Three studies [35–37] focused on superparamagnetic iron oxide (SPIO)-enhanced MRI only. In the first study [35], SPIO was infused over 60 min. Among two [36, 37] of the three studies, a new SPIO known as SHU-555-A was used. It has the advantage of not causing an acute hypotension reaction when it is administered by rapid intravenous injection, enabling dynamic MRI to be performed.
There were two studies [39, 40] using colour Doppler US with or without Levovist as the contrast agent for the detection of HCC among cirrhotic patients. However, they did not report the sensitivity of HCC nodule detection.
Non-explant studies limited to hepatocellular carcinoma in patients with and without cirrhosis
No articles from this group fulfilled the selection criteria, so no data could be extracted.
Studies on focal hepatic lesions amongst patients with and without cirrhosis, using explanted liver histology or other investigations for confirming diagnosis
Two studies [41, 42] reported result of imaging for a subgroup of HCC amongst cirrhotic patients. The results are shown in Table 4. No conclusion could be drawn on the radiological imaging of choice.
|
FP level in the presence of a space occupying lesion in a cirrhotic liver. This would inevitably cause verification bias. In addition, the model and parameter setting of the initial CT machine was not clearly stated although this may have affected the performance of imaging.
Studies addressing patient outcome and health economics of imaging techniques
No articles which addressed health economics or patient outcome of imaging for HCC in cirrhotic patients were identified.
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Amongst the 29 studies fulfilling our selection criteria, it was difficult to compare the results due to the heterogeneity in the study population, imaging modalities, imaging protocols, method of film review and study design. It would be important to establish uniform protocols in the future in comparing the study results. Furthermore, different studies used different reference standards for the diagnosis of HCC. All these factors may induce verification bias, if attempts are made to compare the result between different studies. This is well illustrated by the fact that the sensitivity of an imaging study from non-explant studies is higher than that from the explant studies. For example, the sensitivities of spiral CT for HCC nodules from two non-explant studies were 92% to 95% [27, 32] but the sensitivities from two explant studies were just 53.8% to 71% [16, 22]. This shows that many HCC nodules may be missed by not using explanted liver as the reference standard for HCC diagnosis.
In addition, one potential factor affecting imaging sensitivity was whether the patients were known to have a HCC before transplantation [17].
There are very few studies addressing the issue of diagnostic impact of the choice of radiological imaging. Data from one explant study [23] showed that CTAP was significantly more sensitive than iodized oil-CT and DSA. All the other three explant studies did not find any significant difference of the sensitivities between the various imaging modalities. One possible reason is that the study populations are too small in the three negative studies [21, 22, 24]. There are just 6 to 19 patients being diagnosed as having HCC, from the explanted liver in each study. US, non-spiral CT, spiral CT, iodized oil-CT and angiography had nodule-by-nodule specificities greater than 92% in these studies [17, 21, 22, 24]. The nodule-by-nodule specificity of dynamic gadolinium enhanced MRI was significantly lower than those of US and spiral CT [22] but just six patients were diagnosed with HCC in that study. Unfortunately, the specificity of CTAP was not reported in the only explant study identified in this review.
Among the non-explant studies limited to HCC in cirrhotic patients, all except one [27] did not have any reference standard for the diagnosis of HCC in studying the diagnostic impact of HCC imaging result. For the study [27] with single reference standard, the diagnosis of HCC was made by US-guided fine needle biopsy. This may favour Levovist enhanced colour Doppler US rather than spiral CT when assessing sensitivity because it may not be possible to biopsy lesions detected by CT under US guidance. Again, this may cause verification bias, as HCC nodules detected by one imaging modality are more likely to be detected again by the same imaging technique. A significant drawback of non-explant studies is that the true specificity of an imaging technique cannot be determined because the whole liver is not examined histologically, the most robust method of identifying true false negatives. Using other radiological imaging techniques as the reference standard for diagnosis solely provides a specificity value relative to that technique, but the absolute or true specificity value cannot be determined. In addition, some non-explant studies did not use sensitivity, specificity or accuracy as outcome measures. Instead of using these more clinically relevant endpoints, they used tumour-liver contrast enhancement on CT [33], tumour signal intensity on MRI [36] and flow signals in harmonic [28], colour or power Doppler US [39–42] as outcome measures, preventing firm conclusions on the choice of the best radiological imaging for HCC from this group of studies.
Again assessing diagnostic impact is difficult; one study showing that CTAP has a significantly higher sensitivity among invasive imaging techniques and another showing that dynamic gadolinium enhanced MRI has significantly lower specificity among non-invasive imaging techniques. Unfortunately, it is not possible to derive a conclusion on the choice of imaging for HCC among cirrhotic patients and the results are based on single studies without comparison of invasive and non-invasive techniques. Studies with small sample sizes have shown no difference in sensitivity between enhanced MRI and spiral CT, but the wide confidence intervals do not allow a conclusion as to the best technique.
For diagnostic performance assessment, the results of imaging techniques varied widely from one explant study to another. A contributing factor is the small sample size of each explant study. Nine of the 10 studies had less than 30 patients diagnosed with HCC in the explanted liver [15–22, 24].
Only one study addressed the therapeutic impact effect of the choice of radiological imaging, and, unfortunately, the detail of how treatment was altered was not clearly explained in the article, for example, from transplantation to local resection or from local ablative therapy to general supportive care [43]. More studies on therapeutic impact need to be performed, as ultimately this may be the best test of the use of imaging techniques and should include patient outcome and a health economic assessment.
In conclusion, this systematic review has shown that there is insufficient evidence to decide on the best imaging technique for the detection of HCC in cirrhotic patients after a routine US.
We believe that future evaluation of radiological tests for the detection of HCC in cirrhotic patients should focus on the detection and characterization of tumours less than 2 cm in size. The optimal imaging parameters and contrast injection protocols for lesion detection should be defined. Currently, it is difficult to compare the results of the same imaging modality from different studies because different imaging protocols are used. For assessing the diagnostic performance of an imaging modality, clinically relevant endpoints for lesion detection such as sensitivity and specificity should be used, not just lesion conspicuity. In order to avoid wasting valuable research data, it would be important to report the result from various subgroups of patients such as those with or without cirrhosis, and from various subgroups of focal hepatic lesions such as HCC, metastatic tumours and other benign lesions. Ideally, the explanted liver should be used as the reference standard for the diagnosis of HCC because the whole liver can be examined and verification bias caused by radiological imaging can be avoided. However, it may be difficult to correlate in vivo imaging with explanted livers, particularly of small lesions in a liver of relatively uniform nodular parenchyma and to recognise the potential for missing HCC nodules identified on imaging at histological examination.
Received for publication January 10, 2003. Revision received January 19, 2004. Accepted for publication March 24, 2004.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  F. Eckel, K. Herrmann, S. Schmidt, C. Hillerer, H. A. Wieder, B.-J. Krause, T. Schuster, R. Langer, H.-J. Wester, R. M. Schmid, et al. Imaging of Proliferation in Hepatocellular Carcinoma with the In Vivo Marker 18F-Fluorothymidine J. Nucl. Med., September 1, 2009; 50(9): 1441 - 1447. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Loubeyre Moving Field of Oncologic Diagnosis: Another Reason for Going Beyond the Picture Am. J. Roentgenol., November 1, 2005; 185(5): 1368 - 1369. [Full Text] [PDF]
|
|
|
|
 |
|
 P J A Robinson EBM and the diagnosis of hepatocellular carcinoma Br. J. Radiol., April 1, 2005; 78(928): 370 - 371. [Full Text] [PDF]
|
|
|
|
|
|
A K Burroughs, K T T Fung, F T W Li, M L Raimondo, D Maudgil, A Mancuso, J M Tibballs, A A Watkinson, and D Patch Authors' reply Br. J. Radiol., April 1, 2005; 78(928): 371 - 372. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
