Comparison of ultrasound-detected and undetected hepatocellular carcinomas: a post-hoc, subgroup analysis from the MAGNUS-HCC surveillance trial
Article information
Abstract
Purpose
This study aimed to compare the clinical and imaging features of hepatocellular carcinomas (HCCs) detected versus undetected on surveillance ultrasound (US), based on a prospective cohort.
Methods
This post-hoc subgroup study analyzed data from MAGNUS-HCC trial, a prospective multicenter study evaluating both biannual US and annual non-contrast abbreviated magnetic resonance imaging (NC-AMRI) for HCC surveillance in high-risk patients. Among 34 HCCs, 16 tumors were detected and 18 tumors were undetected on US. We compared clinical features such as sex, age, body mass index (BMI), liver disease etiology, α-fetoprotein (AFP) level, and Child-Pugh class between participants who were diagnosed on US and those who were not. Imaging features including size, hemiliver distribution, anatomical section, subcapsular location, blind spot location, the presence of hepatic steatosis, and the presence of poor sonic window were also compared. NC-AMRI features were also assessed.
Results
No significant differences were observed between US-detected and US-undetected groups in terms of sex, age, liver disease etiology, AFP level, or Child-Pugh class. Patients with US-undetected HCCs had higher BMI (25.7 vs. 23.8 kg/m2, P=0.049). Lesions in blind spots were significantly more common in the US-undetected group (55.6% vs. 18.8%, P=0.039), as were lesions in the left hemiliver (38.9% vs. 6.3%, P=0.043). No significant differences were found in tumor size, anatomical section, subcapsular location, presence of poor sonic window, presence of hepatic steatosis, or magnetic resonance imaging characteristics. Of the 16 US-detected tumors, 62.5% were hypoechoic and 37.5% were hyperechoic.
Conclusion
US is less effective in detecting HCCs in patients with a higher BMI and tumors located in the blind spots or left hemiliver. These limitations should be considered in planning and interpreting surveillance strategies.
Introduction
Surveillance for hepatocellular carcinoma (HCC) in high-risk groups is essential to increase the chance of early detection and improve overall survival [1,2]. International guidelines from the American Association for the Study of Liver Diseases (AASLD), the European Association for the Study of the Liver (EASL), and the Korean Liver Cancer Association (KLCA), recommend biannual ultrasonography (US) with or without serum α-fetoprotein (AFP) [3-5]. US is the standard surveillance tool due to its accessibility, noninvasiveness, cost-effectiveness, and lack of radiation hazards.
However, since US has limited sensitivity in detecting early-stage HCC—reported at only 47% in a meta-analysis [6]—there is ongoing debate about the potential for complementary or alternative imaging methods. To address the limitations of US, the Liver Imaging Reporting and Data System (LI-RADS) proposes the use of contrast-enhanced US and a visualization score system. A visualization score can quantify examination adequacy and recommend alternative imaging including computed tomography (CT) or magnetic resonance imaging (MRI) when visualization is limited [7]. In addition, abbreviated MRI with various protocols and low radiation and/or low contrast media dose CT have been proposed as a surveillance method [8-11].
Understanding the risk factors for failure to detect HCC on US will play an important role in developing and establishing future guidelines for HCC surveillance. Meta-analyses have demonstrated that high body mass index (BMI), nonalcoholic steatohepatitis, high AFP level, and cirrhosis were significant factors for inadequate US or surveillance failure by US [12,13]. In addition to patient-related factors, ultrasound itself has intrinsic limitations due to blind spots [14]—areas that cannot be adequately evaluated because of their anatomical location. Despite retrospective analyses being carried out in several of the preceding studies, to date, no prospective study has comprehensively evaluated both clinical and imaging factors contributing to the failure of US to detect HCC during surveillance.
The MAGNUS-HCC trial [8] is a multicenter prospective trial conducted in Korea that directly compared biannual US and annual non-contrast abbreviated MRI (NC-AMRI) in high-risk populations undergoing HCC surveillance. Only a subset of the HCCs identified in this study were detected by US, with the others being missed. Therefore, we planned a subgroup, post-hoc analysis to determine if there were distinguishing clinical and imaging characteristics between HCCs detected by US and those that were not.
Materials and Methods
Compliance with Ethical Standards
The study was approved by the institutional review boards of all participating centers (Seoul St. Mary’s Hospital, XC15ONMI0023K; Uijeongbu St. Mary's Hospital, XC15ONMI0023U; St. Vincent's Hospital, XC15ONMI0023V; Korea University Guro Hospital, MD15022-001; Seoul National University Hospital, H-1603-082-749; Samsung Medical Center, SMC2015-07-064-003), and written informed consent was obtained from every participant.
Data Sources and Research Overview
Here, we performed a post-hoc, subgroup analysis utilizing data from a previous multicenter prospective clinical trial (MAGNUS-HCC; Clinicaltrial.gov number, NCT02551250 and cris.nih.go.kr number KCT0001739) that compared the performance of annual NC-AMRI and biannual (twice a year) US for HCC surveillance in high-risk populations [8]. The trial protocol has been previously reported [15]. During the follow-up period, participants underwent six biannual US screenings (Fig. 1). NC-AMRI was additionally performed at the 1st, 3rd, and 5th rounds, on the same day as US. If HCC was not identified by the 6th round, a dynamic contrast-enhanced liver CT scan was subsequently performed to verify whether HCC was present or absent.
Flow diagram of the MAGNUS-HCC trial.
Participants underwent six biannual (twice a year) ultrasound (US) surveillance during the follow-up period. In addition, non-contrast abbreviated magnetic resonance imaging was performed during the first, third, and fifth biannual rounds, concurrently with the US. If hepatocellular carcinoma (HCC) was not detected by the sixth round, participants then received a dynamic contrast-enhanced liver computed tomography scan to confirm whether HCC was present or absent. A total of 34 HCCs were detected in 31 patients. Of these, 16 were detected by US and the remaining 18 were not. CT, comupted tomography; MRI, magnetic resonance imaging; NC-AMRI, non-contrast abbreviated magnetic resonance imaging.
In that trial, annual NC-AMRI demonstrated a marginally higher sensitivity for HCC detection (71.0%, 22/31) compared to biannual US (45.2%, 14/31), although this difference did not reach statistical significance (P=0.077). Annual NC-AMRI yielded a significantly higher diagnostic yield (4.26% vs. 1.43%, P<0.001) and a significantly higher positive predictive value (59.5% vs. 31.8%, P=0.006) than biannual US, without a significant increase in the false referral rate (2.91% vs. 3.06%, P=0.885). A simulated strategy of alternating US and NC-AMRI at 6-month intervals significantly improved overall sensitivity to 83.9% (26/31), significantly surpassing biannual US alone (P=0.006) [8].
US Examinations
The liver US examinations were performed by abdominal radiologists who are both experienced over 6 years and board-certified in the MAGNUS-HCC trial. According to the National Cancer Center of Korea, the standard liver US protocol includes the following images: (1) transverse and longitudinal views of the left hemiliver; (2) subcostal and intercostal views of the right hemiliver; (3) transverse image of the right and left portal veins; (4) hepatic veins at the right hepatic dome; (5) longitudinal image of the gallbladder; and (6) longitudinal image of the extrahepatic duct [16]. The radiologists performing US scans and interpretations had access to patients' prior imaging results. However, they were unaware of the results from any other imaging study conducted during the same evaluation round.
Suspicious findings for HCC on US are characterized by two main criteria: (a) a newly identified hepatic nodule measuring 1 cm or more in size is considered suspicious; and (b) this nodule must not display typical features of a simple cyst or hemangioma. A hepatic nodule that meets both of these ultrasound criteria will be classified as suspicious for HCC. However, hepatic nodules ≥1 cm that have been previously diagnosed as benign or have shown only minimal interval change are excluded from this suspicious classification.
Patients
In the MAGNUS-HCC trial, 208 participants were finally enrolled from October 2015 to April 2017 from six tertiary hospitals in Korea. Inclusion criteria are (a) age 40 years or older; (b) clinical or histological diagnosis of liver cirrhosis, regardless of the presence of portal hypertension symptom; (c) risk score of 2.33 or higher according to the model by Velazquez et al. [17], indicating an estimated annual risk of HCC ≥5%. Participants were excluded if they had a previous history of HCC or other malignancies, or if they were pregnant or breastfeeding.
Out of 208 participants, 34 HCCs were diagnosed in 31 patients during a median follow-up period of 30 months, with 64.5% classified as very early-stage (Barcelona Clinic Liver Cancer stage 0) and 35.5% as early-stage HCC (Barcelona Clinic Liver Cancer stage A). Among 31 patients with 34 newly diagnosed HCCs, 16 tumors were detected by US, and 18 were undetected by US and these tumors were the objective of this subgroup analysis.
Data Analysis
Clinical features of 31 participants who were diagnosed as having HCCs, obtained from the pre-existing electronic case report forms (eCRF), included sex, age, BMI, liver disease etiology, AFP level, and Child-Pugh class. Imaging characteristics were collected for 34 tumors and comprised tumor size, anatomical location, tumor echogenicity, and the presence of a poor sonic window also from the eCRF. The presence of a poor sonic window was assessed by the radiologist performing the US during the trial, using a two-point scale (either "present" or "absent"), and without utilizing the US LI-RADS visualization score, as this study was conducted prior to its introduction. In the MAGNUS-HCC trial, a poor sonic window was broadly defined as a condition in which more than 30% of the liver could not be visualized by ultrasound due to bowel gas or a lung shadow. However, this assessment was subjectively evaluated by the US-performing radiologist.
Hepatic steatosis is also a significant factor influencing lesion conspicuity in surveillance US. Although the presence of hepatic steatosis was not included in the eCRF, it was retrospectively evaluated by additionally reviewing the T1 in- and opposed-phase images included in NC-AMRI. Hepatic steatosis was considered present when a diffuse signal drop of the liver was observed on opposed-phase imaging.
In addition to the anatomical classifications of tumor location by liver section (right anterior, right posterior, left medial, left lateral, and caudate) and hemiliver (right vs. left including caudate lobe) which were included in the eCRF, we performed an additional imaging review to ascertain whether the tumor was located in subcapsular area or in ultrasonographically blind spots. Subcapsular location was defined as the tumor directly abutting the liver capsule. Ultrasonographic blind spots, where lesions are difficult to visualize on US, were defined as (1) hepatic dome (right and left hemiliver), (2) the lateral end of the left hemiliver, (3) the hepatic angle (inferior tip of segment 6 of the right hemiliver), (4) <1 cm beneath ribs, (5) caudate lobe or around the inferior vena cava (Fig. 2) [14,18]. The additional imaging review was independently performed by two experienced abdominal radiologists (JIC and SYY), and any discrepancies were resolved by consensus.
Schematic illustration of ultrasonographic "blind spots".
Schematic illustrations show ultrasonographic blind spots of the liver from the anterior view (A) and posterior view (B). Ultrasonographic blind spots are defined as (1) hepatic dome (orange-colored area), (2) the lateral end of the left hemiliver (red-colored area), (3) the hepatic angle, i.e., inferior tip of segment 6 of the right hemiliver (blue-colored area), (4) <1 cm beneath ribs (green-colored areas), and (5) caudate lobe or around the inferior vena cava (teal-colored area).
Also, NC-AMRI features including T2 mild hyperintensity, diffusion restriction, and the presence of fat content on T1 in- and opposed-phase images were obtained from participants from the eCRF when tumors were visible on NC-AMRI.
For comparing variables between tumors which were detected by US and those which did not, either the independent t-test or the Mann-Whitney U test were performed for continuous variables, while categorical variables were assessed using the chi-square test or Fisher’s exact. All statistical analyses were performed with MedCalc Statistical Software version 23.2.7 (MedCalc Software Ltd, Ostend, Belgium). Statistical significance was defined as a P-value <0.05.
Results
A comparison of clinical features between the US-detected and US-undetected groups revealed no significant differences in sex, age, underlying liver disease etiology, or liver function status (Table 1). Serum AFP levels were also comparable between groups, with median values of 8.0 ng/mL (interquartile range, 3.7 to 12.3) in the US-detected group and 12.2 ng/mL (interquartile range, 3.2 to 21.2) in the US-undetected group (P=0.399). In contrast, BMI showed a statistically significant difference between the two groups. Patients with US-undetected HCCs had a higher average BMI of 25.7±2.5 kg/m2 compared to 23.8±2.8 kg/m2 in the US-detected group (P=0.049), suggesting that increased adiposity may interfere with optimal sonographic visualization of liver lesions.
Comparison of clinical features
In a comparison of imaging features, lesions located in the left hemiliver were significantly more prevalent in the US-undetected group compared to the US-detected group (38.9% vs. 6.3%, P=0.043) (Table 2). Also, lesions situated within anatomical blind spots were markedly more frequent in the US-undetected group than the US-detected group (55.6% vs. 18.8%, P=0.039) (Figs. 3-6). There were no significant differences in subcapsular location, anatomical section, or presence of poor sonic window. Hepatic steatosis was more prominently seen in the US-detected group than the US-undetected group (25.0% vs. 44.4%), although this difference did not reach statistical significance (P=0.243). Tumors undetected on US tended to be smaller in size (16.9±9.7 mm vs. 20.0±11.4 mm) than tumors detected on US, even though there was no statistical difference (P=0.137).
Comparison of imaging features
A 73-year-old man with alcoholic cirrhosis.
On ultrasonography, there are two nodular lesions (2 cm and 1.9 cm, arrows) in segment 6 (A, B). These lesions (arrows) show T2 intermediate hyperintensity (C) and diffusion restriction (D) on non-contrast abbreviated magnetic resonance imaging. Dynamic liver computed tomography performed for confirmation demonstrates two tumors (arrows) showing arterial phase hyperenhancement (E).
A 66-year-old man with hepatitis B-associated cirrhosis.
No suspicious lesion suggestive of hepatocellular carcinoma (HCC) was identified in the hepatic dome of the left hemiliver on ultrasonography (A), T2-weighted image (B), or diffusion-weighted image (C). However, due to elevated α-fetoprotein level, contrast-enhanced computed tomography was performed, which revealed a 1.2 cm HCC (arrow) in segment 2 of the hepatic dome, showing arterial phase hyperenhancement (D) and delayed phase washout (E).
A 64-year-old man with hepatitis B-associated cirrhosis.
An approximately 1.2 cm nodule (arrow) in liver segment 8 hepatic dome area shows T2 intermediate hyperintensity (A) and diffusion restriction (B) on non-contrast abbreviated magnetic resonance imaging. However, on ultrasonography (US) (C), the corresponding area was not covered. On arterial phase of dynamic liver magnetic resonance imaging (MRI), there is an arterial phase hyperenhancement (arrow) in corresponding area (D). This participant was scheduled to undergo treatment with radiofrequency ablation. During ultrasound examination for radiofrequency ablation planning, artificial ascites was introduced to improve the sonic window of the hepatic dome. Hepatocellular carcinoma was visualized as a low-echoic lesion (arrow), and the corresponding mass (arrow) was also identified on MRI–US fusion imaging (E).
A 61-year-old man with hepatitis B-associated cirrhosis.
There is no suspicious lesion suggestive of hepatocellular carcinoma in hepatic angle on ultrasonography (A). However, a 1.5-cm-sized lesion (arrow) with T2 mild hyperintensity (B) and diffusion restriction (C) is noted on non-contrast abbreviated magnetic resonance imaging.
On analyses of NC-AMRI, performed in 13 lesions from each group, revealed no statistically significant intergroup differences in T2 mild hyperintensity, diffusion restriction, or presence of intratumoral fat. Among the 16 HCCs identified by US, 10 tumors (62.5%) appeared hypoechoic and six tumors (37.5%) hyperechoic; notably, all two fat-containing tumors on MRI showed hyperechogenicity.
Discussion
This multicenter study is the first to prospectively compare the clinical and imaging factors between patients with HCC detected on surveillance US and those with failed detection. The US-undetected group had a significantly higher BMI than those in the US-detected group (25.7 vs. 23.8 kg/m2, P=0.049). The detectability of HCC on US can be influenced by its anatomical location. HCCs that were missed on US were more frequently located in the left hemiliver (38.9% vs. 6.3%, P=0.043) and blind spots (55.6% vs. 18.8%, P=0.039). However, subcapsular location, poor sonic window, hepatic steatosis and MR imaging factors were not associated with US detection failure.
It's widely believed that certain areas of the liver, called blind spots, are prone to missing lesions during a US examination [18-20]. Therefore, the US operators must have a precise understanding of the blind spot areas and pay close attention to carefully evaluating these areas during the examination. However, only a limited number of studies have thoroughly investigated this issue in the context of HCC surveillance. In this study, we defined blind spot areas as the hepatic dome, caudate lobe, the area beneath the ribs, the lateral end of the hemiliver, and the hepatic angle (Fig. 2) [14,18]. While previous research by Lee et al. [14] revealed that HCC in their US-missed group were more frequently located in these blind spots, defining the US-missed group as cases with no HCC detected on US and an AFP level of ≥20 ng/mL, our study offers a more precise comparison of HCC detectability. We achieved this by prospectively performing liver US and NC-AMRI on the same day.
High BMI is a well-known risk factor for surveillance failure in HCC screening [13,21,22]. A study by Del Poggio et al. [22] found that surveillance failure increased significantly when BMI exceeded 25 kg/m². Furthermore, a meta-analysis revealed that US detectability of HCC decreases with increasing BMI, particularly at BMI ≥35 kg/m² or within the 25-30 kg/m2, depending on tumor size [13]. In our study, patients with US-undetected HCCs had a higher average BMI of 25.7 kg/m2, which falls within the range identified in previous reports. However, BMI cutoff values may vary by age, sex, and ethnicity, requiring a population-specific assessment. Higher BMI is associated with thicker subcutaneous fat, which attenuates the ultrasound beam and may reduce image quality [23]. Nevertheless, BMI may not fully reflect subcutaneous fat distribution or its actual impact on US image quality. Therefore, future studies using waist circumference or subcutaneous fat quantification may provide more precise understanding of the relationship between body composition and US detectability of HCC.
The US LI-RADS introduced the concept of a visualization score to assess the quality of ultrasound examinations [3]. A visualization score of C has been associated with higher false-negative rates for detecting HCC compared to scores of A or B [24,25]. However, since the visualization scoring system proposed by LI-RADS was introduced after the initiation of the MAGNUS-HCC trial, it could not be applied retrospectively. Instead, we defined a poor sonic window as a condition in which more than 30% of the liver could not be visualized by ultrasound due to bowel gas or a lung shadow, without consideration of parenchymal echogenicity. There was no significant difference in the proportion of poor sonic windows between the US-detected and US-undetected groups, suggesting that poor sonic windows did not significantly affect detectability. In addition, the pooled proportion of patients with the LI-RADS visualization score of C in previous studies is reported to be approximately 6.7% [21], whereas the proportion of poor sonic window in our cohort was considerably higher at 26.5% (9 of 34 cases). The definition for a poor sonic window used in our study did not follow the LI-RADS visualization score C, which is based on a 50% cutoff and takes heterogeneous echogenicity into consideration. Also, the cohort of our study differs from the general surveillance group. These differences may explain the lack of statistical significance observed. The LI-RADS visualization score of C has been reported to occur more frequently in patients with cirrhosis, non-alcoholic fatty liver disease, and a high BMI (>25 kg/m2).
In this study, HCCs located in the left liver are more frequently missed. During US examination, the right hemiliver is typically evaluated multiple times through both subcostal and intercostal scanning approaches, whereas the left hemiliver is assessed using only transverse and longitudinal views from a subxiphoid location, which may increase the likelihood of missed lesions. In contrast, subcapsular location did not significantly affect detectability. Although subcapsular locations were expected to be difficult to visualize due to acoustic interference from the abdominal wall, their impact was minimal, possibly because the proximity of the lesions to the probe enhanced their visibility. To our knowledge, no previous study has demonstrated that HCCs located in the left liver are more frequently missed. However, as this result was derived from a relatively small study population, it should be interpreted with caution, and warrants validation in studies with larger populations.
Our study has several limitations. First, although the prospective design is a strength, the relatively small sample size limits statistical significance. The results related to the left hemiliver, hepatic steatosis, and poor sonic window cannot exclude the possibility of having arisen from the small sample size. Second, as the study cohort consisted of patients with an annual risk of HCC exceeding 5%, the generalizability of our findings to a standard HCC surveillance population may be limited. Nevertheless, this inclusion criterion was necessary to ensure an adequate number of newly detected HCC cases during surveillance. Third, although prior studies have suggested that HCC detection may be more challenging in patients with macronodular cirrhosis, we could not analyze this factor because all patients in our study had cirrhosis. Fourth, as most of our study population had hepatitis B, the findings may not be generalizable to regions where hepatitis B is not endemic. Lastly, the LI-RADS visualization score had not been published at the time of this study and therefore could not be applied to our analysis.
In conclusion, surveillance US for detecting HCC is less effective in patients with a high BMI, when the lesion is located in the left hemiliver, or when it lies within a blind spot. Recognizing the limitations of ultrasound is important for both clinicians and radiologists. These limitations should be considered in planning and interpreting surveillance strategies.
Notes
Author Contributions
Conceptualization: Youn SY, Choi JI, Kang TW. Data acquisition: Choi JI, Choi MH, Yoon JH, Lee CH, Kim HA, Ku YM, Lee JM, Kim SH, Kim KA, Lee SL. Data analysis or interpretation: Youn SY, Choi JI. Drafting of the manuscript: Youn SY, Choi JI. Critical revision of the manuscript: Youn SY, Choi MH, Yoon JH, Lee CH, Kim HA, Ku YM, Lee JM, Kim SH, Kim KA, Lee SL, Kang TW. Approval of the final version of the manuscript: all authors.
Conflict of Interest
No potential conflict of interest relevant to this article was reported.
Acknowledgments
This study was supported by a grant from the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No.2022R1F1A1074852).
References
Article information Continued
Notes
Key point
Higher body mass index was significantly associated with missed detection of hepatocellular carcinoma on surveillance ultrasound. Tumors located in blind spots or the left hemiliver were more likely to be undetected on ultrasound.