Liver International ISSN 1478-3223
LIVER CANCER
Prognostic role of magnetic resonance imaging vs. computed tomography for hepatocellular carcinoma undergoing chemoembolization Beom Kyung Kim1,2, Kyung Ah Kim3, Chansik An4, Eun Jin Yoo1, Jun Yong Park1,2, Do Young Kim1,2, Sang Hoon Ahn1,2, Kwang-Hyub Han1,2,5, Seung Up Kim1,2,* and Myeong-Jin Kim4,* 1 2 3 4 5
Department of Internal Medicine, Yonsei University College of Medicine, Seoul, Korea Yonsei Institute of Gastroenterology, Yonsei University College of Medicine, Seoul, Korea Department of Radiology, St. Vincent’s Hospital, The Catholic University of Korea, Gyeonggi-do, Korea Department of Radiology, Yonsei University College of Medicine, Seoul, Korea Brain Korea 21 Project for Medical Science, Seoul, Korea
Keywords chemoembolization – computed tomography – hepatocellular carcinoma – magnetic resonance imaging – prognosis Abbreviations AFP, a-fetoprotein; CI, confidence intervals; CR, complete response; CT, Computed tomography; GBI, gross bile duct invasion; GVI, gross vascular invasion; HBV hepatitis B virus; HCC, hepatocellular carcinoma; HR, hazard ratio; IQR, interquartile range; ITM, irregular tumour margin; MRI, magnetic resonance imaging; NA, not applicable; OR odds ratio; OS, overall survival; PD, progressive disease; PRE, peripheral ragged enhancement; PR, partial response; SD, stable disease; TACE, transarterial chemoembolization. Correspondence Myeong-Jin Kim, MD, PhD, Department of Radiology, Yonsei University College of Medicine, 50-1 Yonsei-ro, Seodaemun–gu, Seoul 120–752, Korea Tel: +82 2 2228 7400 Fax: +82 2 393 3035 e-mail: [email protected]
Abstract Background & Aims: Computed tomography (CT) and magnetic resonance imaging (MRI) play important roles in diagnosis and staging of hepatocellular carcinoma (HCC). However, prognostic roles of radiological characteristics are not yet determined. Methods: Eighty-eight patients treated with chemoembolization were analysed. Radiological parameters at baseline were assessed in all patients using both dynamic CT and MRI. Treatment responses were assessed using modified RECIST 4 weeks after the first chemoembolization. Results: Gross vascular invasion (GVI), bile duct invasion, irregular tumour margin (ITM), peripheral ragged enhancement (PRE) and satellite nodules on CT or MRI were associated with non-response (stable disease or progression) after chemoembolization respectively (all P ≤ 0.05). GVI, ITM and PRE on CT or MRI were also independently associated with poor overall survival (OS) respectively (all P ≤ 0.05). Using these results, a prognostic scoring system for CT and MRI were developed; 0, absence of all three features (GVI, ITM and PRE); 1, presence of one feature; 2, presence of two features; and 3, presence of three features. After adjusting tumour size, tumour number and alpha-foetoprotein level, both CT and MRI scores were independently associated with OS (both P < 0.001). Patients with CT or MRI score ≥2 had a worse OS than those with score 0.05). Conclusions: Radiological parameters by CT and MRI may be useful in biological characterization of tumours and prognostification for HCC treated with chemoembolization.
Received 3 June 2014 Accepted 18 November 2014 DOI:10.1111/liv.12751 Liver Int. 2015; 35: 1722–1730
Although progress has been made in the treatment of hepatocellular carcinoma (HCC), prognosis remains very poor as only a small group of patients with early*Seung Up Kim and Myeong-Jin Kim contributed equally to this manuscript.
1722
stage HCC has access to curative options such as orthotopic liver transplantation, surgical resection and local ablative therapy. Most patients fail to qualify for these surgical interventions primarily because of advanceddisease stage and/or compromised liver function at diagnosis (1). Liver International (2015) © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Kim et al.
Key Points
We firstly defined prognostic imaging biomarkers for hepatocellular carcinoma apart from tumour size, number and vascular invasion, using computed tomography (CT) or magnetic resonance imaging (MRI). Among radiological parameters by CT and MRI, gross vascular invasion (GVI), gross bile duct invasion, complete capsule formation, irregular tumour margin (ITM), peripheral ragged enhancement (PRE), satellite nodules and washout appearance, we selected GVI, ITM and PRE as key variables. They are significantly associated with survival outcomes and treatment responses. Using these three key radiological parameters, we have proposed a scoring system for stratification of post-TACE outcome. Radiological examinations, such as computed tomography (CT) and magnetic resonance imaging (MRI) play important roles for the diagnosis and staging of HCC. For diagnosis of HCC, presence of increased arterial enhancement on early phase and decreased enhancement (‘washout appearance’) on late phase images of dynamic CT and MRI may spare an invasive histological examination. Of the radiological features assessed in the diagnosis of HCC using CT or MRI, apart from obvious macroscopically visible vascular invasion, only the size and number criterion have been utilized in prognostication by tumour staging (2). In addition to the roles for diagnosis and staging, recent studies have suggested that CT and MRI may help identify valuable imaging features that are useful for pre-treatment estimation of prognosis of patients with HCC (3–5). These imaging features include presence of capsule formation (6–8), enhancement pattern (3, 5, 9), fatty change (10) and signal intensity on diffusion-weighted imaging (11). However, application of these radiological biomarkers as useful prognostic predictors have been limited mostly to the analysis of surgically resected cases and furthermore, their roles have not been validated yet in larger studies with reference to the most robust prognostic factors such as staging (tumour size, number, vascular invasion) and clinical factors (tumour markers, patients’ status, etc.). On the other hand, for most HCC patients who are not candidate for surgical treatments, transarterial chemoembolization (TACE) is the mainstay of treatment (2). A meta-analysis showed that TACE improved survival in patients with unresectable HCC, thus ending a controversy for its survival benefits (12) and thus, currently, TACE is being widely used and has become the standard treatment modality for patients with unresectable, specifically those with intermediate-stage HCC. Since a histological evaluation of the whole tumour specimen for elucidation of the biological characteristics is not eligible in this setting, it is valuable to define addiLiver International (2015) © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Radiological parameters in hepatocellular carcinoma
tional reliable imaging biomarkers, without the need for further tests or treatment challenges for not only appropriate prediction of prognosis but also establishment of treatment strategies. Therefore, in this study, we evaluated whether the imaging features of CT and MRI can be useful for the prediction of prognosis in patients undergoing TACE and which modality is the better in term of prediction of prognosis. Materials and methods Patients’ eligibility
Patients were selected from the database of Severance Hospital, Yonsei University College of Medicine. Treatment-na€ıve patients with intrahepatic HCC who underwent both liver dynamic CT and MRI prior to first-line TACE therapy between 2008 and 2010 were considered eligible for this study. Liver dynamic MRI was performed to assess the necessity for modification of treatment strategies. The exclusion criteria were an unmeasurable lesion (i.e. a diffuse type HCC), the presence of an additional primary malignancy in another organ, the presence of extrahepatic lesions or main portal vein invasion, Child–Pugh class of B or C and the presence of uncontrolled functional or metabolic disease. This study protocol conformed to the ethical guidelines outlined in the 1975 Declaration of Helsinki and was approved by the Institutional Review Board of Severance Hospital, Yonsei University College of Medicine, Seoul, Republic of Korea. Diagnosis of HCC
HCC was diagnosed by histological or radiological evaluation with reference to American Association for the Study of Liver Diseases or European Association for the Study of the Liver guidelines (2, 13). A positive finding on dynamic CT or MRI was defined as an increased arterial enhancement, followed by washout appearance (defined as decreased enhancement relative to liver in the portal or delayed phase) (14). Assessment of radiological parameters at baseline using CT and MRI
Radiological parameters were evaluated in all patients using both CT and MRI before TACE was performed. For both imaging modalities, the presence of gross vascular invasion (GVI), gross bile duct invasion (GBI), complete capsule formation, irregular tumour margin (ITM), peripheral ragged enhancement (PRE), satellite nodules and washout appearance were evaluated. Furthermore, MRI-specific parameters including signal hyperintensity on T1- and T2-weighted images, an evidence of intralesional fat (signal drop on
1723
Kim et al.
Radiological parameters in hepatocellular carcinoma
opposed-phase than on in-phase images), and a hyperintensity on DW images (DWI) were also evaluated. GVI was diagnosed when soft tissue or tumour thrombi within vessel were directly visualized and a case with only extrinsic compression of the vessels by tumour was not regarded as GVI. Figures S1 and S2 shows representative images using CT and MRI respectively. Treatment modality
TACE was performed in two steps: infusion of a mixture of 5 ml iodized oil contrast medium (Lipiodol) and 30– 50 mg adriamycin, followed by embolization of the feeding arteries using gelatin sponge particles. Sequential TACE was scheduled at 6- to 8-week intervals to achieve the better response on an ‘on-demand’ basis if eligible, provided that patients’ clinical and laboratory findings permitted and there was no evidence of extrahepatic spread. Assessment of treatment responses according to modified Response Evaluation Criteria in Solid Tumours
Treatment responses were assessed 4 weeks after the initial TACE treatment according to modified Response Evaluation Criteria in Solid Tumours guidelines (15), which define viable tumours according to the uptake of contrast material in the arterial phase of either dynamic CT or MRI. Tumours retaining iodized oil, as well as necrotic lesions without intratumoural arterial enhancement, were regarded as necrotized tumour foci. Up to two target lesions were assessed, based on previous studies (16, 17). Objective response refers to sum of complete response (CR) and partial response (PR) cases, whereas non-response refers to sum of stable disease (SD) and progressive disease (PD) cases. Statistical analysis
Intermethod concordances between CT and MRI were estimated for each radiological parameter using the kappa (j) coefficient. The strength of agreement based on j values was interpreted as follows: j < 0.21, poor; j = 0.21–0.40, fair; j = 0.41–0.60, moderate; j = 0.61– 0.80, good; and j > 0.80, excellent (18). The chi-square test or Fisher’s exact test was used as appropriate to determine the association of each radiological parameter with the radiological response (objective response vs. non-response). Binary logistic regression analysis was used to calculate odds ratios and 95% confidence intervals (CI). Overall survival (OS) was calculated as the interval between the date of TACE initiation and the date of death or last follow-up evaluation. Survival time was estimated using the Kaplan–Meier method, and the survival difference between groups was assessed using the
1724
log-rank test. In order to predict OS, an adjusted hazard ratio was calculated for each radiological parameter using the multivariate Cox proportional hazard model, adjusting clinical variables. Statistical analyses were performed using SAS statistical software (ver. 9.1.3; SAS Institute, Inc., Cary, NC, USA). A two-sided P-value of ≤0.05 was considered significant. Results Patients’ characteristics
A total of 88 patients (70 men) were included in the final analysis. The median interval between CT and MRI was 13 [interquartile range (IQR) 8–19] days. The baseline demographical and clinical characteristics of these patients are shown in Table 1. The median age of the patients was 58.5 (IQR 51.3–66.8) years. Chronic hepatitis B virus (HBV) infection was the major aetiology for chronic liver disease. All patients had well-preserved liver function (Child–Pugh class A). The median diameter of the largest measurable lesion was 2.9 cm (IQR 2.1–4.5 cm). The number of measurable lesions at baseline was 1 in 52 patients (59.1%), 2 in 19 patients (21.6%), 3 in 5 patients (5.7%), 4 in 5 patients (5.7%) and 5 or more in 7 patients (7.9%). The median alphafoetoprotein (AFP) level was 48.1 ng/ml (IQR 12.0– 310.5 ng/ml). Treatment outcomes
The median interval between the baseline MRI and TACE was 6 (IQR 2–13) days. The median follow-up period after TACE was 28.5 (IQR 21.2–35.3) months. After the first TACE session, CR was achieved in 39 patients, PR in 29 and SD in 16 respectively. PD was observed in four patients. Table 1 also shows the patients’ characteristics between objective responders and non-responders. Among 45 patients who achieved PR or SD after the first TACE session, additional TACE was performed for residual viable tumour to achieve the better response, if eligible, as follows; one (n = 28), two (n = 7), four (n = 1) and six (n = 1) times. However, eight patients did not undergo additional TACE owing to patients’ refusal or deteriorated liver function. As a result, among 29 patients who achieved PR after the first TACE session, CR was achieved additionally in 16 patients during repeated TACE sessions, and among 16 patients who achieved SD after the first TACE session, CR and PR were achieved additionally in 3 and 5 patients during repeated TACE sessions respectively. Overall, all three survival curves were statistically different, with the median OS of ‘not reached’, 25.6 and 8.2 months in patients who achieved CR as the best response, those who achieved PR as the best response, and those who did not achieve the objective response Liver International (2015) © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Kim et al.
Radiological parameters in hepatocellular carcinoma
Table 1. Patients’ characteristics Variables
Entire population (n = 88)
Objective responders (n = 68)
Non-responders (n = 20)
P-value
Age (years) Male gender Aetiology HBV HCV Alcohol Others Tumour number 1 2 3 4 ≥5 Tumour size (cm) MELD score Alpha-foetoprotein (ng/ml)
58.5 (51.3–66.8) 70 (79.5)
58.4 ± 8.2 54 (79.4)
61.4 ± 12.9 16 (80)
0.347 1.000
10 (50.0) 3 (15.0) 4 (20.0) 3 (15.0)
0.458
55 (62.5) 12 (13.6) 13 (14.8) 8 (9.1) 52 (59.1) 19 (21.6) 5 (5.7) 5 (5.7) 7 (7.9) 2.9 (2.1–4.5) 5.16 (2.74–7.02) 48.1 (12.0–310.5)
45 (66.2) 9 (13.2) 9 (13.2) 5 (7.4) 41 (60.3) 16 (23.5) 5 (7.4) 5 (7.4) 1 (1.5) 2.8 (2.0–4.30) 4.93 (2.71–7.56) 46.5 (12.9–272.2)
11 (55.0) 3 (15.0) 0 (0.0) 0 (0.0) 6 (30.0) 3.5 (2.2–5.0) 5.90 (3.54–6.63) 48.5 (9.1–16601.7)
0.015*
0.344 0.530 0.605
Values are expressed as median (interquartile range), mean (±standard deviation), or no. (%). *P-value was calculated to compare the proportion of tumor number ≥4 between two groups. HBV, hepatitis B virus; HCV, hepatitis C virus; MELD, model for end-stage liver disease.
eventually during treatment courses, respectively (logrank test, P < 0.001), suggesting the impact of the best response during repeated TACE sessions (if eligible) on the survival outcome. Assessment and concordance of radiological parameters using dynamic CT and MRI
An assessment of radiological parameters using dynamic CT and MRI is presented in Table S1. These parameters include presence of GVI, GBI, complete capsule formation, ITM, PRE, satellite nodules and washout appearance on dynamic CT and MRI and MRI-specific parameters such as hyperintensity on T1-and T2-weighted images, intralesional fat and a hyperintensity on DWI. Overall, both imaging modalities showed moderate to excellent concordance as reflected by calculated j-values (0.652–1.000). In the context of common radiological parameters assessed using both imaging modalities, equivalent or marginally better detection rates were achieved using MRI when compared to CT, with the exception of the detection of ITM. Association between radiological parameters and treatment responses
The association between radiological parameters and treatment responses is outlined in Table 2. When assessed by CT, there was a significant association between the poor treatment response (non-response) and the presence of GVI [odds ratio (OR): 5.333, 95% confidence interval (CI): 1.276–22.285; P = 0.026], GBI [OR: not applicable (NA); P = 0.050], ITM (OR: 2.825, 95% CI: 1.018–8.621; P = 0.048), PRE (OR: Liver International (2015) © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
4.457, 95% CI: 1.549–12.821; P = 0.006) and satellite nodules (OR: 6.889, 95% CI: 2.022–23.469; P = 0.001). Similar results were obtained for these five radiological parameters when assessed by MRI: there was a significant association between the poor treatment response (non-response) and the presence of GVI (OR: 3.214, 95% CI: 0.961–10.756; P = 0.050), GBI (OR: NA; P = 0.050), ITM (OR: 3.584, 95% CI: 1.170– 10.989; P = 0.021), PRE (OR: 3.194, 95% CI: 1.126– 9.062; P = 0.029) and satellite nodules (OR: 3.867, 95% CI: 1.264–11.831; P = 0.024). However, there was no significant association between MRI-specific radiological parameters (hypersignal intensity on T1- and T2-weighted images, a fat signal and a hyperintense signal on DWI) and therapeutic response (P > 0.05 for all parameters), indicating that these are poor predictive factors. Association between radiological parameters and overall survival
Table 3 shows the prognostic significance of radiological parameters in terms of OS. Using univariate analysis, patients with GVI, GBI, ITM, PRE or satellite nodules on CT scans were likely to have a poorer OS than those without these features (P ≤ 0.05 for all parameters). These five radiological parameters were individually entered into multivariate analysis with adjustment for tumour size, tumour number and an AFP level of ≥400 ng/ml, since all three clinical variables were wellknown prognostic factors. So, GVI [adjusted hazards ratio (HR): 5.608, 95% CI: 2.109–14.912; P < 0.001], ITM (adjusted HR 1.992, 95% CI: 0.978–4.046;
1725
Kim et al.
0.024 0.339 0.221 0.405 1.000 0.457 CT, computed tomography; MRI, magnetic resonance imaging; OR, odds ratio; CI, confidence interval; NA, not applicable; DWI, diffusion-weighted image.
3.867 (1.264–11.831) NA 0.273 (0.036–2.073) 3.526 (0.211–59.056) 0.000 (0.000–0.000) 0.669 (0.230–1.940) 10 (14.7%) 65 (95.6%) 66 (97.1%) 1 (1.5%) 1 (1.5%) 50 (73.5%) 0.001 1.000 NA NA NA NA 8 (40.0%) 18 (90.0%) NA NA NA NA 6 (8.8%) 60 (92.3%) NA NA NA NA
6.889 (2.022–23.469) 1.200 (0.234–6.165) NA NA NA NA
8 (40.0%) 20 (100.0%) 18 (90.0%) 1 (5.0%) 0 (0.0%) 13 (65.0%)
0.050 0.050 0.304 0.021 0.029 3.214 (0.961–10.756) NA 0.512 (0.176–1.490) 3.584 (1.170–10.989) 3.194 (1.126–9.062) 6 (30.0%) 2 (10.0%) 6 (30.0%) 15 (75.0%) 13 (65%) 8 (11.8%) 0 (0.0%) 31 (45.6%) 31 (45.6%) 25 (36.7%) 0.026 0.050 0.381 0.048 0.006 5.333 (1.276–22.285) NA 0.611 (0.198–1.997) 2.825 (1.018–8.621) 4.457 (1.549–12.821) 5 (25.0%) 2 (10.0%) 5 (25.0%) 15 (75.0%) 13 (65.0%)
Gross vascular invasion Gross bile duct invasion Complete capsule formation Irregular tumour margin Peripheral ragged enhancement Satellite nodule Washout appearance T2 hypersignal intensity T1 hypersignal intensity Fat signal Hyperintense signal on DWI
4 (5.9%) 0 (0.0%) 24 (35.3%) 35 (51.5%) 20 (29.4%)
Pvalue
MRI
OR (95% CI) Non-responders (n = 20) Radiological parameters
Objective responders (n = 68)
CT
Table 2. Association between radiological parameters and treatment responses 1726
Objective responders (n = 68)
Non-responders (n = 20)
OR (95% CI)
Pvalue
Radiological parameters in hepatocellular carcinoma
P = 0.050) and PRE (adjusted HR: 3.899, 95% CI: 1.867–8.142; P < 0.001) remained significant unfavourable predictors. Similarly, using univariate analysis, patients with GVI, GBI, ITM, PRE and satellite nodule on MRI scans were likely to have a poorer OS than those without these features (P ≤ 0.05 for all parameters). Likewise, when these five radiological parameters were individually entered into multivariate analysis with adjustment for the three previously mentioned clinical variables, GVI (adjusted HR: 4.052, 95% CI: 1.693–9.695; P = 0.002), ITM (adjusted HR: 2.525, 95% CI: 1.245–5.128; P = 0.010) and PRE (adjusted HR: 2.294, 95% CI: 1.095–4.807; P = 0.028) remained significant unfavourable predictors. However, none of the MRI-specific radiological parameters such as hyperintensity on T1and T2-weighted images, a fat signal and a hyperintensity on DWI proved significant predictors for OS using univariate analysis (P > 0.05 for all parameters). In summary, three key parameters GVI, ITM and PRE on CT were significantly associated with both unfavourable treatment response and poorer OS and when radiological parameters were assessed using MRI, the same results were maintained. Assessment of a prognostic scoring system based on radiological parameters
Further to above results, we developed prognostic formulas using radiological parameters that significantly predicted treatment response and OS. CT or MRI prognostic scores were derived independently as follows; 0, absence of all 3 significant radiological features (GVI, ITM and PRE); 1, presence of one significant radiological feature; 2, presence of two radiological features; and 3, presence of all three radiological features. After adjusting for tumour size, tumour number and the AFP level, both the CT and MRI scores were independently associated with OS (P < 0.001 for both parameters) (Table 4). Thus, patients with a CT or MRI score of ≥2 had a poorer OS than those with a score of
LIVER CANCER
Prognostic role of magnetic resonance imaging vs. computed tomography for hepatocellular carcinoma undergoing chemoembolization Beom Kyung Kim1,2, Kyung Ah Kim3, Chansik An4, Eun Jin Yoo1, Jun Yong Park1,2, Do Young Kim1,2, Sang Hoon Ahn1,2, Kwang-Hyub Han1,2,5, Seung Up Kim1,2,* and Myeong-Jin Kim4,* 1 2 3 4 5
Department of Internal Medicine, Yonsei University College of Medicine, Seoul, Korea Yonsei Institute of Gastroenterology, Yonsei University College of Medicine, Seoul, Korea Department of Radiology, St. Vincent’s Hospital, The Catholic University of Korea, Gyeonggi-do, Korea Department of Radiology, Yonsei University College of Medicine, Seoul, Korea Brain Korea 21 Project for Medical Science, Seoul, Korea
Keywords chemoembolization – computed tomography – hepatocellular carcinoma – magnetic resonance imaging – prognosis Abbreviations AFP, a-fetoprotein; CI, confidence intervals; CR, complete response; CT, Computed tomography; GBI, gross bile duct invasion; GVI, gross vascular invasion; HBV hepatitis B virus; HCC, hepatocellular carcinoma; HR, hazard ratio; IQR, interquartile range; ITM, irregular tumour margin; MRI, magnetic resonance imaging; NA, not applicable; OR odds ratio; OS, overall survival; PD, progressive disease; PRE, peripheral ragged enhancement; PR, partial response; SD, stable disease; TACE, transarterial chemoembolization. Correspondence Myeong-Jin Kim, MD, PhD, Department of Radiology, Yonsei University College of Medicine, 50-1 Yonsei-ro, Seodaemun–gu, Seoul 120–752, Korea Tel: +82 2 2228 7400 Fax: +82 2 393 3035 e-mail: [email protected]
Abstract Background & Aims: Computed tomography (CT) and magnetic resonance imaging (MRI) play important roles in diagnosis and staging of hepatocellular carcinoma (HCC). However, prognostic roles of radiological characteristics are not yet determined. Methods: Eighty-eight patients treated with chemoembolization were analysed. Radiological parameters at baseline were assessed in all patients using both dynamic CT and MRI. Treatment responses were assessed using modified RECIST 4 weeks after the first chemoembolization. Results: Gross vascular invasion (GVI), bile duct invasion, irregular tumour margin (ITM), peripheral ragged enhancement (PRE) and satellite nodules on CT or MRI were associated with non-response (stable disease or progression) after chemoembolization respectively (all P ≤ 0.05). GVI, ITM and PRE on CT or MRI were also independently associated with poor overall survival (OS) respectively (all P ≤ 0.05). Using these results, a prognostic scoring system for CT and MRI were developed; 0, absence of all three features (GVI, ITM and PRE); 1, presence of one feature; 2, presence of two features; and 3, presence of three features. After adjusting tumour size, tumour number and alpha-foetoprotein level, both CT and MRI scores were independently associated with OS (both P < 0.001). Patients with CT or MRI score ≥2 had a worse OS than those with score 0.05). Conclusions: Radiological parameters by CT and MRI may be useful in biological characterization of tumours and prognostification for HCC treated with chemoembolization.
Received 3 June 2014 Accepted 18 November 2014 DOI:10.1111/liv.12751 Liver Int. 2015; 35: 1722–1730
Although progress has been made in the treatment of hepatocellular carcinoma (HCC), prognosis remains very poor as only a small group of patients with early*Seung Up Kim and Myeong-Jin Kim contributed equally to this manuscript.
1722
stage HCC has access to curative options such as orthotopic liver transplantation, surgical resection and local ablative therapy. Most patients fail to qualify for these surgical interventions primarily because of advanceddisease stage and/or compromised liver function at diagnosis (1). Liver International (2015) © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Kim et al.
Key Points
We firstly defined prognostic imaging biomarkers for hepatocellular carcinoma apart from tumour size, number and vascular invasion, using computed tomography (CT) or magnetic resonance imaging (MRI). Among radiological parameters by CT and MRI, gross vascular invasion (GVI), gross bile duct invasion, complete capsule formation, irregular tumour margin (ITM), peripheral ragged enhancement (PRE), satellite nodules and washout appearance, we selected GVI, ITM and PRE as key variables. They are significantly associated with survival outcomes and treatment responses. Using these three key radiological parameters, we have proposed a scoring system for stratification of post-TACE outcome. Radiological examinations, such as computed tomography (CT) and magnetic resonance imaging (MRI) play important roles for the diagnosis and staging of HCC. For diagnosis of HCC, presence of increased arterial enhancement on early phase and decreased enhancement (‘washout appearance’) on late phase images of dynamic CT and MRI may spare an invasive histological examination. Of the radiological features assessed in the diagnosis of HCC using CT or MRI, apart from obvious macroscopically visible vascular invasion, only the size and number criterion have been utilized in prognostication by tumour staging (2). In addition to the roles for diagnosis and staging, recent studies have suggested that CT and MRI may help identify valuable imaging features that are useful for pre-treatment estimation of prognosis of patients with HCC (3–5). These imaging features include presence of capsule formation (6–8), enhancement pattern (3, 5, 9), fatty change (10) and signal intensity on diffusion-weighted imaging (11). However, application of these radiological biomarkers as useful prognostic predictors have been limited mostly to the analysis of surgically resected cases and furthermore, their roles have not been validated yet in larger studies with reference to the most robust prognostic factors such as staging (tumour size, number, vascular invasion) and clinical factors (tumour markers, patients’ status, etc.). On the other hand, for most HCC patients who are not candidate for surgical treatments, transarterial chemoembolization (TACE) is the mainstay of treatment (2). A meta-analysis showed that TACE improved survival in patients with unresectable HCC, thus ending a controversy for its survival benefits (12) and thus, currently, TACE is being widely used and has become the standard treatment modality for patients with unresectable, specifically those with intermediate-stage HCC. Since a histological evaluation of the whole tumour specimen for elucidation of the biological characteristics is not eligible in this setting, it is valuable to define addiLiver International (2015) © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Radiological parameters in hepatocellular carcinoma
tional reliable imaging biomarkers, without the need for further tests or treatment challenges for not only appropriate prediction of prognosis but also establishment of treatment strategies. Therefore, in this study, we evaluated whether the imaging features of CT and MRI can be useful for the prediction of prognosis in patients undergoing TACE and which modality is the better in term of prediction of prognosis. Materials and methods Patients’ eligibility
Patients were selected from the database of Severance Hospital, Yonsei University College of Medicine. Treatment-na€ıve patients with intrahepatic HCC who underwent both liver dynamic CT and MRI prior to first-line TACE therapy between 2008 and 2010 were considered eligible for this study. Liver dynamic MRI was performed to assess the necessity for modification of treatment strategies. The exclusion criteria were an unmeasurable lesion (i.e. a diffuse type HCC), the presence of an additional primary malignancy in another organ, the presence of extrahepatic lesions or main portal vein invasion, Child–Pugh class of B or C and the presence of uncontrolled functional or metabolic disease. This study protocol conformed to the ethical guidelines outlined in the 1975 Declaration of Helsinki and was approved by the Institutional Review Board of Severance Hospital, Yonsei University College of Medicine, Seoul, Republic of Korea. Diagnosis of HCC
HCC was diagnosed by histological or radiological evaluation with reference to American Association for the Study of Liver Diseases or European Association for the Study of the Liver guidelines (2, 13). A positive finding on dynamic CT or MRI was defined as an increased arterial enhancement, followed by washout appearance (defined as decreased enhancement relative to liver in the portal or delayed phase) (14). Assessment of radiological parameters at baseline using CT and MRI
Radiological parameters were evaluated in all patients using both CT and MRI before TACE was performed. For both imaging modalities, the presence of gross vascular invasion (GVI), gross bile duct invasion (GBI), complete capsule formation, irregular tumour margin (ITM), peripheral ragged enhancement (PRE), satellite nodules and washout appearance were evaluated. Furthermore, MRI-specific parameters including signal hyperintensity on T1- and T2-weighted images, an evidence of intralesional fat (signal drop on
1723
Kim et al.
Radiological parameters in hepatocellular carcinoma
opposed-phase than on in-phase images), and a hyperintensity on DW images (DWI) were also evaluated. GVI was diagnosed when soft tissue or tumour thrombi within vessel were directly visualized and a case with only extrinsic compression of the vessels by tumour was not regarded as GVI. Figures S1 and S2 shows representative images using CT and MRI respectively. Treatment modality
TACE was performed in two steps: infusion of a mixture of 5 ml iodized oil contrast medium (Lipiodol) and 30– 50 mg adriamycin, followed by embolization of the feeding arteries using gelatin sponge particles. Sequential TACE was scheduled at 6- to 8-week intervals to achieve the better response on an ‘on-demand’ basis if eligible, provided that patients’ clinical and laboratory findings permitted and there was no evidence of extrahepatic spread. Assessment of treatment responses according to modified Response Evaluation Criteria in Solid Tumours
Treatment responses were assessed 4 weeks after the initial TACE treatment according to modified Response Evaluation Criteria in Solid Tumours guidelines (15), which define viable tumours according to the uptake of contrast material in the arterial phase of either dynamic CT or MRI. Tumours retaining iodized oil, as well as necrotic lesions without intratumoural arterial enhancement, were regarded as necrotized tumour foci. Up to two target lesions were assessed, based on previous studies (16, 17). Objective response refers to sum of complete response (CR) and partial response (PR) cases, whereas non-response refers to sum of stable disease (SD) and progressive disease (PD) cases. Statistical analysis
Intermethod concordances between CT and MRI were estimated for each radiological parameter using the kappa (j) coefficient. The strength of agreement based on j values was interpreted as follows: j < 0.21, poor; j = 0.21–0.40, fair; j = 0.41–0.60, moderate; j = 0.61– 0.80, good; and j > 0.80, excellent (18). The chi-square test or Fisher’s exact test was used as appropriate to determine the association of each radiological parameter with the radiological response (objective response vs. non-response). Binary logistic regression analysis was used to calculate odds ratios and 95% confidence intervals (CI). Overall survival (OS) was calculated as the interval between the date of TACE initiation and the date of death or last follow-up evaluation. Survival time was estimated using the Kaplan–Meier method, and the survival difference between groups was assessed using the
1724
log-rank test. In order to predict OS, an adjusted hazard ratio was calculated for each radiological parameter using the multivariate Cox proportional hazard model, adjusting clinical variables. Statistical analyses were performed using SAS statistical software (ver. 9.1.3; SAS Institute, Inc., Cary, NC, USA). A two-sided P-value of ≤0.05 was considered significant. Results Patients’ characteristics
A total of 88 patients (70 men) were included in the final analysis. The median interval between CT and MRI was 13 [interquartile range (IQR) 8–19] days. The baseline demographical and clinical characteristics of these patients are shown in Table 1. The median age of the patients was 58.5 (IQR 51.3–66.8) years. Chronic hepatitis B virus (HBV) infection was the major aetiology for chronic liver disease. All patients had well-preserved liver function (Child–Pugh class A). The median diameter of the largest measurable lesion was 2.9 cm (IQR 2.1–4.5 cm). The number of measurable lesions at baseline was 1 in 52 patients (59.1%), 2 in 19 patients (21.6%), 3 in 5 patients (5.7%), 4 in 5 patients (5.7%) and 5 or more in 7 patients (7.9%). The median alphafoetoprotein (AFP) level was 48.1 ng/ml (IQR 12.0– 310.5 ng/ml). Treatment outcomes
The median interval between the baseline MRI and TACE was 6 (IQR 2–13) days. The median follow-up period after TACE was 28.5 (IQR 21.2–35.3) months. After the first TACE session, CR was achieved in 39 patients, PR in 29 and SD in 16 respectively. PD was observed in four patients. Table 1 also shows the patients’ characteristics between objective responders and non-responders. Among 45 patients who achieved PR or SD after the first TACE session, additional TACE was performed for residual viable tumour to achieve the better response, if eligible, as follows; one (n = 28), two (n = 7), four (n = 1) and six (n = 1) times. However, eight patients did not undergo additional TACE owing to patients’ refusal or deteriorated liver function. As a result, among 29 patients who achieved PR after the first TACE session, CR was achieved additionally in 16 patients during repeated TACE sessions, and among 16 patients who achieved SD after the first TACE session, CR and PR were achieved additionally in 3 and 5 patients during repeated TACE sessions respectively. Overall, all three survival curves were statistically different, with the median OS of ‘not reached’, 25.6 and 8.2 months in patients who achieved CR as the best response, those who achieved PR as the best response, and those who did not achieve the objective response Liver International (2015) © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Kim et al.
Radiological parameters in hepatocellular carcinoma
Table 1. Patients’ characteristics Variables
Entire population (n = 88)
Objective responders (n = 68)
Non-responders (n = 20)
P-value
Age (years) Male gender Aetiology HBV HCV Alcohol Others Tumour number 1 2 3 4 ≥5 Tumour size (cm) MELD score Alpha-foetoprotein (ng/ml)
58.5 (51.3–66.8) 70 (79.5)
58.4 ± 8.2 54 (79.4)
61.4 ± 12.9 16 (80)
0.347 1.000
10 (50.0) 3 (15.0) 4 (20.0) 3 (15.0)
0.458
55 (62.5) 12 (13.6) 13 (14.8) 8 (9.1) 52 (59.1) 19 (21.6) 5 (5.7) 5 (5.7) 7 (7.9) 2.9 (2.1–4.5) 5.16 (2.74–7.02) 48.1 (12.0–310.5)
45 (66.2) 9 (13.2) 9 (13.2) 5 (7.4) 41 (60.3) 16 (23.5) 5 (7.4) 5 (7.4) 1 (1.5) 2.8 (2.0–4.30) 4.93 (2.71–7.56) 46.5 (12.9–272.2)
11 (55.0) 3 (15.0) 0 (0.0) 0 (0.0) 6 (30.0) 3.5 (2.2–5.0) 5.90 (3.54–6.63) 48.5 (9.1–16601.7)
0.015*
0.344 0.530 0.605
Values are expressed as median (interquartile range), mean (±standard deviation), or no. (%). *P-value was calculated to compare the proportion of tumor number ≥4 between two groups. HBV, hepatitis B virus; HCV, hepatitis C virus; MELD, model for end-stage liver disease.
eventually during treatment courses, respectively (logrank test, P < 0.001), suggesting the impact of the best response during repeated TACE sessions (if eligible) on the survival outcome. Assessment and concordance of radiological parameters using dynamic CT and MRI
An assessment of radiological parameters using dynamic CT and MRI is presented in Table S1. These parameters include presence of GVI, GBI, complete capsule formation, ITM, PRE, satellite nodules and washout appearance on dynamic CT and MRI and MRI-specific parameters such as hyperintensity on T1-and T2-weighted images, intralesional fat and a hyperintensity on DWI. Overall, both imaging modalities showed moderate to excellent concordance as reflected by calculated j-values (0.652–1.000). In the context of common radiological parameters assessed using both imaging modalities, equivalent or marginally better detection rates were achieved using MRI when compared to CT, with the exception of the detection of ITM. Association between radiological parameters and treatment responses
The association between radiological parameters and treatment responses is outlined in Table 2. When assessed by CT, there was a significant association between the poor treatment response (non-response) and the presence of GVI [odds ratio (OR): 5.333, 95% confidence interval (CI): 1.276–22.285; P = 0.026], GBI [OR: not applicable (NA); P = 0.050], ITM (OR: 2.825, 95% CI: 1.018–8.621; P = 0.048), PRE (OR: Liver International (2015) © 2014 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
4.457, 95% CI: 1.549–12.821; P = 0.006) and satellite nodules (OR: 6.889, 95% CI: 2.022–23.469; P = 0.001). Similar results were obtained for these five radiological parameters when assessed by MRI: there was a significant association between the poor treatment response (non-response) and the presence of GVI (OR: 3.214, 95% CI: 0.961–10.756; P = 0.050), GBI (OR: NA; P = 0.050), ITM (OR: 3.584, 95% CI: 1.170– 10.989; P = 0.021), PRE (OR: 3.194, 95% CI: 1.126– 9.062; P = 0.029) and satellite nodules (OR: 3.867, 95% CI: 1.264–11.831; P = 0.024). However, there was no significant association between MRI-specific radiological parameters (hypersignal intensity on T1- and T2-weighted images, a fat signal and a hyperintense signal on DWI) and therapeutic response (P > 0.05 for all parameters), indicating that these are poor predictive factors. Association between radiological parameters and overall survival
Table 3 shows the prognostic significance of radiological parameters in terms of OS. Using univariate analysis, patients with GVI, GBI, ITM, PRE or satellite nodules on CT scans were likely to have a poorer OS than those without these features (P ≤ 0.05 for all parameters). These five radiological parameters were individually entered into multivariate analysis with adjustment for tumour size, tumour number and an AFP level of ≥400 ng/ml, since all three clinical variables were wellknown prognostic factors. So, GVI [adjusted hazards ratio (HR): 5.608, 95% CI: 2.109–14.912; P < 0.001], ITM (adjusted HR 1.992, 95% CI: 0.978–4.046;
1725
Kim et al.
0.024 0.339 0.221 0.405 1.000 0.457 CT, computed tomography; MRI, magnetic resonance imaging; OR, odds ratio; CI, confidence interval; NA, not applicable; DWI, diffusion-weighted image.
3.867 (1.264–11.831) NA 0.273 (0.036–2.073) 3.526 (0.211–59.056) 0.000 (0.000–0.000) 0.669 (0.230–1.940) 10 (14.7%) 65 (95.6%) 66 (97.1%) 1 (1.5%) 1 (1.5%) 50 (73.5%) 0.001 1.000 NA NA NA NA 8 (40.0%) 18 (90.0%) NA NA NA NA 6 (8.8%) 60 (92.3%) NA NA NA NA
6.889 (2.022–23.469) 1.200 (0.234–6.165) NA NA NA NA
8 (40.0%) 20 (100.0%) 18 (90.0%) 1 (5.0%) 0 (0.0%) 13 (65.0%)
0.050 0.050 0.304 0.021 0.029 3.214 (0.961–10.756) NA 0.512 (0.176–1.490) 3.584 (1.170–10.989) 3.194 (1.126–9.062) 6 (30.0%) 2 (10.0%) 6 (30.0%) 15 (75.0%) 13 (65%) 8 (11.8%) 0 (0.0%) 31 (45.6%) 31 (45.6%) 25 (36.7%) 0.026 0.050 0.381 0.048 0.006 5.333 (1.276–22.285) NA 0.611 (0.198–1.997) 2.825 (1.018–8.621) 4.457 (1.549–12.821) 5 (25.0%) 2 (10.0%) 5 (25.0%) 15 (75.0%) 13 (65.0%)
Gross vascular invasion Gross bile duct invasion Complete capsule formation Irregular tumour margin Peripheral ragged enhancement Satellite nodule Washout appearance T2 hypersignal intensity T1 hypersignal intensity Fat signal Hyperintense signal on DWI
4 (5.9%) 0 (0.0%) 24 (35.3%) 35 (51.5%) 20 (29.4%)
Pvalue
MRI
OR (95% CI) Non-responders (n = 20) Radiological parameters
Objective responders (n = 68)
CT
Table 2. Association between radiological parameters and treatment responses 1726
Objective responders (n = 68)
Non-responders (n = 20)
OR (95% CI)
Pvalue
Radiological parameters in hepatocellular carcinoma
P = 0.050) and PRE (adjusted HR: 3.899, 95% CI: 1.867–8.142; P < 0.001) remained significant unfavourable predictors. Similarly, using univariate analysis, patients with GVI, GBI, ITM, PRE and satellite nodule on MRI scans were likely to have a poorer OS than those without these features (P ≤ 0.05 for all parameters). Likewise, when these five radiological parameters were individually entered into multivariate analysis with adjustment for the three previously mentioned clinical variables, GVI (adjusted HR: 4.052, 95% CI: 1.693–9.695; P = 0.002), ITM (adjusted HR: 2.525, 95% CI: 1.245–5.128; P = 0.010) and PRE (adjusted HR: 2.294, 95% CI: 1.095–4.807; P = 0.028) remained significant unfavourable predictors. However, none of the MRI-specific radiological parameters such as hyperintensity on T1and T2-weighted images, a fat signal and a hyperintensity on DWI proved significant predictors for OS using univariate analysis (P > 0.05 for all parameters). In summary, three key parameters GVI, ITM and PRE on CT were significantly associated with both unfavourable treatment response and poorer OS and when radiological parameters were assessed using MRI, the same results were maintained. Assessment of a prognostic scoring system based on radiological parameters
Further to above results, we developed prognostic formulas using radiological parameters that significantly predicted treatment response and OS. CT or MRI prognostic scores were derived independently as follows; 0, absence of all 3 significant radiological features (GVI, ITM and PRE); 1, presence of one significant radiological feature; 2, presence of two radiological features; and 3, presence of all three radiological features. After adjusting for tumour size, tumour number and the AFP level, both the CT and MRI scores were independently associated with OS (P < 0.001 for both parameters) (Table 4). Thus, patients with a CT or MRI score of ≥2 had a poorer OS than those with a score of
