Indeterminate 1-2-cm nodules found on hepatocellular carcinoma surveillance: Biopsy for all, some, or none?

Authors


  • Potential conflict of interest: Nothing to report.

  • This study has not been funded by grants or other financial support.

Abstract

In the latest hepatocellular carcinoma (HCC) management guidelines by the American Association for the Study of Liver Diseases, biopsy is advocated for all nodules deemed indeterminate after imaging work-up by contrast-enhanced scans. However, the latest guidelines' imaging work-up algorithm has been shown to improve sensitivity of characterization of HCC for 1-2-cm nodules, decreasing the proportion of HCCs that remain indeterminate after imaging work-up. We undertook a study of 1-2-cm indeterminate nodules to determine what proportions are malignant and which variables can be used to limit biopsy to a subset of nodules at higher risk of malignancy. Eighty consecutive patients with 93 indeterminate nodules were included. Final diagnosis was established in 85 nodules, with 13 malignant (9 by biopsy, 4 by growth) and 72 benign (stability of ≥18 months). Cause of liver disease, ethnicity, size, arterial hypervascularity, venous hypoenhancement, and presence of synchronous typical HCC were analyzed by univariate logistic analysis to determine significant predictors of malignancy. Rate of malignancy among indeterminate 1-2-cm nodules was found to be 14%-23%. Only arterial hypervascularity [odds ratio (OR), 3.7) and presence of synchronous HCC (OR, 7.1) were significant predictors of malignancy. A strategy of limiting biopsy to nodules that had either feature would result in 23 biopsies and potentially detect 8 of 13 malignant nodules, yielding a sensitivity of 62% and specificity of 79%. Conclusion: The prevalence of malignancy among 1-2-cm indeterminate nodules is low (14%-23%), and biopsy of all such nodules results in many negative results. Limiting biopsy to nodules with arterial hypervascularity or in the presence of a synchronous typical HCC would detect the majority of HCCs while substantially reducing the number of biopsies. (HEPATOLOGY 2011)

The American Association for the Study of Liver Diseases (AASLD) hepatocellular carcinoma (HCC) practice guidelines recommend a biopsy when imaging work-up of nodules is indeterminate.1 The biopsy of nodules in the background of cirrhosis has several implications. The nodule has to be visible on ultrasound (US) to be practically biopsied; additional nodules found on computed tomography/magnetic resonance imaging (CT/MRI) work-up of that found on surveillance may not be visible on US. The biopsy of the nodule has to be technically feasible; vaguely seen nodules or those close to large blood vessels in the central liver may be very difficult to biopsy. In patients with several indeterminate nodules, multiple biopsies increase the risks of the procedure and may be impractical. The false-negative rate of biopsy in such a setting is substantial, with up to 30%, as reported by Forner et al., among nodules under 2 cm, and nodules deemed benign on biopsy still require close imaging follow-up for 18-24 months.2 Finally, the use of biopsy after costly imaging work-up requires high enough prevalence of malignancy to justify the procedure, especially considering the false-negative rate and potential complications of biopsy, such as bleeding. It is this final point that was the subject of our study.

The adoption of sequential imaging work-up for nodules detected on HCC surveillance results in improved sensitivity of imaging to detect malignancy, and consequently, nodules that are indeterminate would be a smaller proportion of malignant nodules.1, 3 Given that the probability of malignancy is lower in smaller nodules, the prevalence of HCC among 1-2-cm indeterminate nodules may be low enough so as not to warrant biopsy for all. The aim of this study was twofold. First, it was to determine the prevalence of HCC among indeterminate 1-2-cm nodules. Second, it was to identify variables with significant association with HCC that would allow selective application of biopsy to a subset of indeterminate 1-2-cm nodules.

Abbreviations

AASLD, American Association for the Study of Liver Diseases; AFP, alpha-fetoprotein; CEUS, contrast-enhanced ultrasound; CT, computed tomography; HCC, hepatocellular carcinoma; MRI, magnetic resonance imaging; OR, odds ratio; 3D, three-dimensional; US, ultrasound.

Patients and Methods

This was a retrospective study of data acquired from a “standardized” clinical program, based on the AASLD HCC management guidelines, implemented for all patients undergoing US surveillance at the institution.4 Between January 2006 and December 2007, consecutive asymptomatic cirrhotic patients at a single center were referred for standardized work-up of nodules found during routine US surveillance. Standardized imaging work-up of 1-2-cm nodules included CT, MRI, and contrast-enhanced US (CEUS). Patients with indeterminate nodules measuring 1-2 cm and without a previous history of HCC, Child-Pugh score of C, or on the transplant waitlist were included in this study. Indeterminate nodules were defined as not demonstrating enhancement greater than the liver in the arterial phase (i.e., arterial hypervascularity) and less than the liver in the venous or delayed phases (i.e., washout) on contrast-enhanced imaging on at least two contrast-enhanced scans, based on AASLD HCC management guidelines.1 Nodules demonstrating the typical enhancement pattern of hemangioma were excluded. If the patients had undergone three contrast-enhanced scans, nodules would have had to have demonstrated an indeterminate enhancement pattern in at least two scans for inclusion. Patients with more than one nodule, including those with a typical HCC, elsewhere were included. Patient characteristics are summarized in Table 1, and the breakdown of the study population is shown in Fig. 1.

Figure 1.

Breakdown of study population by final diagnosis (pts = patients).

Table 1. Patient Characteristics
 All Patients
VariableN%
  1. Abbreviations: CT, computed tomography; MRI, magnetic resonance imaging; CEUS, contrast-enhanced ultrasound.

PatientsN = 72
Age (years)  
 Median (range)59 (22-75)
Gender  
 Male4258
 Female3042
Ethnicity  
 Asian3244
 Non-Asian4056
Risk factors  
 Hepatitis B3751
 Hepatitis C2129
 Hepatitis B and C11
 Nonalcoholic steatohepatitis57
 Autoimmune hepatitis46
 Alcoholic cirrhosis23
 Other causes of liver cirrhosis23
Child-Pugh score
 A7097
 B23
No. of nodules
 16286
 2811
 311
 411
No. of scans
 CT68 
 MRI69 
 CEUS70 

Reference Standard.

A nodule was considered malignant if it demonstrated growth on follow-up imaging or had features of HCC on biopsy. As part of the standardized work-up protocol, all indeterminate nodules were recommended for biopsy as per first AASLD HCC management guidelines.4 Though the reason for the recommendation of biopsy was detailed in a standardized report summarizing all imaging findings, the decision to biopsy was left to individual hepatologists responsible for the patient. Before 2006, the usual practice for indeterminate 1-2-cm nodules was close imaging follow-up, and subsequent to the implementation of the standardized program, there was slow acceptance of the new recommendation. Nodules not visible on grayscale US, and those in patients with other larger nodules, were unlikely to have been biopsied.

For the aims of this study, irrespective of biopsy findings, a nodule was considered benign only if it remained stable on imaging for a minimum of 18 months. Given the small size of the nodules and increase in variability in measurement, growth was defined as 30% change in lesional diameter. Follow-up imaging was performed by the detecting modality or CT scan every 3 months for 18 months and every 6 months thereafter.

Potential Predictors of Malignant Behavior.

The following variables were analyzed to determine whether malignant behavior in indeterminate 1-2-cm nodules could be predicted: cause of liver disease (i.e., hepatitis B, C, or other), ethnicity, nodule size, arterial hypervascularity, hypoenhancement on the venous/delayed phase relative to the liver, and presence of synchronous typical HCC.

Surveillance Protocol.

Surveillance was performed by US at a hepatobiliary referral center where approximately 3,000 patients undergo routine surveillance every 6 months. Scans were performed by US technologists with an on-site abdominal radiologist checking all images. Direct physician scanning was performed if a new abnormality was noted by the sonographer. Any well-defined, reproducible nodule ≥1 cm detected on US was included in this study. The test of reproducibility was detection of lesion on grayscale at the time of CEUS, which was performed personally by a radiologist with expertise in sonography of cirrhotic patients. The nodule was remeasured and confirmed as a true nodule. Hepatic lobulations and pseudonodules caused by a coarse liver were excluded.

Imaging Techniques and Analysis.

CT scans were performed using 64 detector scanners (Toshiba Aquilion 64; Toshiba Medical Systems, Inc., Tustin, CA). A four-phase CT scan was performed, with precontrast, arterial (20 seconds after trigger using bolus tracking in aorta), portal venous (70 seconds), and delayed phases (180 seconds). MRI scans were performed on a 1.5-T system (Excite HD and Excite HD; GE Healthcare, Milwaukee, WI), with a four- or eight-channel phased-array torso coil. The standard protocol included dynamic three-dimensional (3D) fluoro-triggered, fat-suppressed, volumetric, fast-spoiled, gradient-echo images (3D LAVA) with unenhanced, arterial, portal venous, late portal venous, and delayed (300-second delay) phases. CEUS was performed using Acuson Sequoia (Siemens, Mountain View, CA), with a 4C1 multifrequency curved array probe. Between one and three bolus injections of 0.2-0.3 mL of perflutren lipid microsphere (Definity; Lantheus Medical Imaging, N. Billerica, MA) was injected intravenously and a low-mechanical-index (0.10-0.20) technique was used to dynamically image any nodule of interest to 5 minutes postinjection.

Imaging analysis was performed prospectively and in consensus by three fellowship trained hepatobiliary imagers with 7-11 years of experience. For each nodule, hypervascularity in the arterial phase and hypoenhancement in venous and/or delayed phases relative to the liver were determined. Hypoenhancement on the venous/delayed phase meant the lesion appeared lower in signal than adjacent liver irrespective of its appearance in the arterial phase.

Malignancy was considered when a nodule demonstrated arterial hypervascularity and hypoenhancement in the venous or delayed phase relative to the liver. The lack of definition of a nodule was the result of the absence of the diagnostic profile. Isodensity in each evaluated phase was considered a negative finding.

Biopsy Technique.

Biopsy was performed using US guidance with an 18-gauge needle (Temno Biopsy System; Allegiance Systems, McGaw Park, IL) as part of routine clinical practice. Between one and three core biopsies were obtained.

Statistical Analysis.

Univariate logistic regression analysis was used to determine the findings predictive of HCC. Generalized estimating equations were used to adjust for the correlation between multiple nodules within a patient. Odds ratios (ORs) with confidence intervals were calculated for each variable; significance was defined as P ≤ 0.05. Sensitivity, specificity, and positive and negative predictive values were calculated for each of the variables with significant association with malignant behavior.

Results

Rate of Malignancy.

Ninety-three indeterminate 1-2-cm nodules were found in 80 patients (Fig. 1). These include 10 nodules in 9 patients, which were detected on contrast-enhanced work-up imaging, but not seen on the original surveillance US where another nodule had been detected. In 8 patients (with 8 nodules), final diagnosis could not be established: 6 patients underwent radiofrequency ablation without tissue biopsy or recurrence, and 2 died of non-HCC-related liver failure before 18 months of imaging follow-up. The final diagnosis was established in 85 nodules (72 patients). Benignity was determined in 72 of 85 (85%) nodules by long-term stability on follow-up imaging (mean follow-up, 29.9 months; range, 19-44). Malignancy was determined in 13 of 85 (15%) nodules, 9 by biopsy and 4 by growth, on follow-up imaging (at 13, 14, 24, and 25 months after detection). The 4 malignant nodules detected by growth also exhibited typical HCC enhancement characteristics (i.e., arterial hypervascularity and washout) on follow-up imaging. None of the 13 malignancies, including those detected by growth, demonstrated vascular invasion and all were treated by radiofrequency ablation. Because a diagnosis could not be established in 8 of 93 nodules, the rate of malignancy was considered within a range, with a minimum of 13 of 93 (14%) and maximum of 13 (+ 8) of 93 (23%).

Biopsy was performed in 30 of 85 nodules, yielding 9 (29%) malignant results. Twenty-one biopsies were not malignant, with results being nonlesional liver parenchyma (n = 12), regenerative nodule (n = 5), dysplastic nodule (n = 3), and angiomyolipoma (n = 1). Two of the biopsied nodules deemed nonlesional grew on follow-up imaging (at 14 and 25 months) and were designated as malignant.

Discrepancy Between Contrast-Enhanced Imaging Scans.

Discrepancy between contrast-enhanced scans as to the diagnosis of malignancy were noted in 5 nodules (5 patients). On CEUS, 3 nodules showed the enhancement pattern of malignancy, with CT and MRI being negative. The final diagnoses in these were benign for 2 and malignant for 1. On CT scan, 1 nodule showed the enhancement pattern of malignancy (CEUS and MRI negative), with the final diagnosis being malignant. Finally, one nodule showed the enhancement pattern of malignancy on MRI (CT and CEUS negative), with the final diagnosis being benign.

Synchronous Nodules.

In 8 of 72 (11%) patients with indeterminate nodules, a synchronous HCC was seen at the time of initial imaging work-up, with typical imaging appearance on at least 2 of 3 modalities. In 1 of these patients, an indeterminate nodule was observed with 3 typical HCCs, whereas in the remaining 7, a single synchronous HCC was noted. The mean size of these synchronous HCCs was 2.0 cm (range, 1.1-2.7). As well, in 7 of 72 (10%) patients, 7 additional nodules were seen, which were benign on final diagnosis, and 2 of these measured >2 cm and the other 5 were <1 cm.

Development of New Nodules.

Over the follow-up period, 27 of 72 (38%) patients developed 31 new liver nodules. Of these, 5 nodules in 5 patients were deemed new HCC, with 3 being in the 13 patients with indeterminate nodules whose final diagnosis was malignant.

Potential Predictors of Malignant Behavior.

Univariate logistic regression was performed in the 85 nodules with a known diagnosis to determine which variables had a significant association with malignancy (Table 2). Arterial hypervascularity (OR, 3.7; P = 0.04) and presence of a synchronous typical HCC (OR, 7.1; P = 0.01) were factors with significant association with the final diagnosis of malignancy. Cause of background liver disease, Asian ethnicity, and size of the nodule showed no significant association. Hypoenhancement relative to the liver in the venous or delayed phases demonstrated increased likelihood of malignancy (OR, 3.1; P = 0.14), but this was not significant. Alpha-fetoprotein (AFP) levels, measured within 90 days of detection, were available in 41 patients only, because it was not part of the standard surveillance protocol. Among patients with indeterminate nodules that were malignant on final diagnosis, the AFP level was elevated (≥20 ng/mL) in 2 of 9 (22%) patients; 1 of 2 patients had a synchronous HCC. Among patients with indeterminate nodules that were benign on final diagnosis, the AFP level was elevated in 9 of 32 (28%) patients; 2 of these 9 patients had a synchronous HCC.

Table 2. Association Between Patient and Nodule Characteristics and the Diagnosis of Malignancy
 All Nodules (N = 85)Benign (N = 72)Malignant (N = 13)Univariate*
VariableN%N%N%OR (95% CI)P Value
  • Abbreviations: OR, odds ratio; CI, confidence interval; HCC, hepatocellular carcinoma.

  • *

    Univariate logistic regression using generalized estimating equations; adjusted for multiple nodules per patient.

  • One patient excluded from analysis of background liver disease who had both hepatitis B and C.

  • OR for lesion size represents every 1-mm increase in size.

Background liver disease        
 Hepatitis B435137516501.000.89
 Other414935496501.1 (0.3-3.8) 
 Hepatitis C283325353251.000.60
 Other566747659751.5 (0.4-6.2) 
Lesion size (cm)        
 Median (range)1.2 (1.0-2.0)1.2 (1.0-1.9)1.3. (1.0-2.0)1.1 (0.9-1.4)0.29
Ethnicity        
 Asian374430427541.000.44
 Non-Asian485642586460.6 (0.2-2.1) 
Other HCC        
 No779168949691.000.01
 Yes89464317.1 (1.5-33.6) 
Arterial enhancement        
 No667859827541.000.04
 Yes192213186463.7 (1.1-12.8) 
Relative venous hypoenhancement        
 No283326362151.000.14
 Yes5767466411853.0 (0.7-12.8) 

Limiting Biopsy to Nodules With Significant Features.

If AASLD guidelines were to be followed for all indeterminate 1-2-cm nodules, 85 biopsies would be performed, yielding 13 malignancies (assuming flawless biopsy performance). Nineteen indeterminate nodules demonstrated arterial phase enhancement on at least one of the enhanced scans, 6 of which were malignant (Table 3). Limiting biopsies to these nodules, as opposed to all, would have resulted in a sensitivity of 46% and specificity of 82% in the detection of malignancy. Eight indeterminate nodules were in the presence of a synchronous typical HCC elsewhere in the liver, 4 of which were malignant. Limiting biopsy to these nodules, as opposed to all nodules, would have resulted in a sensitivity of 31% and specificity of 94% in the detection of malignancy. Two malignant nodules, and 2 benign nodules, had both arterial hypervascularity and a synchronous HCC. If biopsy were to have been limited to nodules that had arterial hypervascularity or were in the presence of a synchronous typical HCC, 23 of 85 nodules would have been biopsied, 8 of which would have been malignant. Such a strategy would have resulted in a sensitivity of 62% and specificity of 79% (Table 3).

Table 3. Performance Characteristics of Limiting Biopsy to 1-2-cm Nodules Having Significant Association With Malignancy
VariableTPTNFPFNSensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)
  1. Ninety-five percent CIs adjusted for clustering, where necessary.

  2. Abbreviations: TP, true positive; TN, true negative; FP, false positive; FN, false negative; CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value; HCC, hepatocellular carcinoma.

Synchronous HCC4684931945088
     (12.04-59.07)(85.61-97.86)(20.01-79.99)(77.73-93.52)
Arterial enhancement65913746823289
     (22.36-71.83)(69.62-88.97)(15.39-56.43)(78.23-94.66)
Presence of either85715562793592
     (34.36-83.02)(66.20-86.57)(19.00-57.17)(81.19-96.51)

Discussion

We have demonstrated that the prevalence of malignancy is relatively low among 1-2-cm nodules deemed indeterminate by two contrast-enhanced imaging scans (between 14% and 23%). The application of the new AASLD criteria to recently published reports of small nodules found on HCC surveillance also shows a low prevalence of malignancy among indeterminate 1-2-cm nodules. In a study of 89 nodules measuring up to 2 cm found in an HCC surveillance population, Forner et al. found that only 3 of 60 HCCs did not fulfill AASLD enhancement criteria for malignancy on at least one of the contrast-enhanced scans. Therefore, the proportion of malignancy among indeterminate nodules requiring biopsy, according to the new AASLD criteria, was 3 of 32 (9%).2 Similarly, subanalysis of the 1-2-cm nodules in the Leoni et al. and Sangiovanni et al. studies revealed malignancy rates among indeterminate nodules of 24% (5 of 21) and 33% (12 of 32), respectively.5, 6

Given the low likelihood of malignancy, especially in the setting of a substantial false-negative rate, we believe that biopsy may not be practical for all indeterminate 1-2-cm nodules. The latest AASLD HCC guidelines have altered the diagnostic algorithm for 1-2-cm nodules, shifting from a requirement of two coincidental contrast-enhanced scans (i.e., both being positive) to sequential scans (performing a second only when the first is negative).1 This change was based on recent publications demonstrating that sequential imaging improved sensitivity while maintaining high specificity.3, 6 For example, we had shown that using CT and MRI, the sensitivity of HCC detection improved from 41% using a coincidental algorithm to 74% using a sequential algorithm.3 The improved sensitivity of the new diagnostic algorithm changes the definition of an indeterminate nodule (to lack of typical features on only one imaging modality), with the result that there will be fewer HCCs among these nodules than among those defined by the older standard. Therefore, biopsying all indeterminate 1-2-cm nodules appears impractical.

Our study demonstrates that selective application of biopsy to nodules with specific features can substantially reduce the number of biopsies and increase the proportion of malignant nodules found. Our results show that had biopsy been reserved for indeterminate 1-2-cm nodules demonstrating arterial hypervascularity in at least one of the scans or in the presence of a typical synchronous HCC, 8 of 13 malignancies would have been detected, and the number of biopsies would have decreased 3-fold (from 85 to 23). Such a strategy would yield a sensitivity of 62% and specificity of 79% (Table 3) for detection of malignancy by biopsy among indeterminate 1-2-cm nodules. Though this sensitivity is not optimal, close imaging follow-up will very likely detect malignant transformation in the remainder of the nodules within the curable stage. All such nodules in our study were thus treated by radiofrequency ablation. Furthermore, arterial hypervascularity has been linked to more sinister tumor differentiation by a number of publications, and limiting biopsy to these nodules may result in the identification of tumors with a worse prognosis.7-10

Conversely, it is possible that the application of biopsy to all indeterminate 1-2-cm nodules may result in an overdiagnosis of HCC (i.e., the diagnosis and treatment of histologically malignant nodules), which may not cause significant disease in patients.11 Although the concept of “very early HCCs,” which do not exhibit the imaging findings of typical HCCs, has now been adopted internationally by liver pathologists, there is no study of the clinical significance of these nodules.12 What proportion of very early HCCs will progress to the more advanced HCCs, and over what time frame? The rationale behind the current framework of HCC treatment algorithms, including transplantation criteria, have been based primarily on more advanced HCCs, diagnosed using imaging studies in the 1990s.13 In the setting of a competing potentially fatal disease (i.e., cirrhosis), both the identification and treatment of “very early HCCs” has yet to be justified.

One strength of this study is that long-term stability was used as the reference standard for benignity. Nodules not clearly malignant, even those diagnosed as benign by biopsy, were followed for a mean of 29.9 months (range, 19-44). Because of a substantial false-negative rate of biopsy, the AASLD guidelines recommend follow-up of all indeterminate nodules by imaging for 18-24 months.1 Furthermore, the differentiation of very early HCC and high-grade dysplasia may be especially difficult on a small biopsy sample and may depend on the judgment and experience of the pathologist. Histological findings of malignancy on biopsy are biomarkers of malignant behavior. Rather than relying on predictive biomarkers, long-term follow-up allows the determination of “actual” behavior. Though this is not a convenient method, we believe that it represents a stronger reference standard than biopsy.

A limitation of this study was that the reference standard of follow-up imaging was not applied to all nodules. In 9 of 93 indeterminate nodules where the biopsy showed malignancy, treatment was applied before imaging documentation of growth. Given the adoption of AASLD HCC management guidelines in our practice as the standard of care, this limitation was unavoidable. Furthermore, 8 of 93 indeterminate nodules were not followed for a minimum of 18 months, 2 of which were caused by non-HCC-related deaths. Therefore, a range had to be provided for prevalence of malignancy, representing either assumption that none or all of the 8 nodules may have been malignant. The multiplicity of indeterminate nodules in an individual patient may result in clustering bias; we, therefore, applied a correction in our statistical methods for such clustering effects. Although this is the largest study of indeterminate nodules, the overall number of nodules and, especially, the low prevalence of malignancy provided for suboptimal confidence intervals. Finally, this is a retrospective analysis of prospectively acquired data with inherent bias; our findings require confirmation in a larger, prospective trial.

In conclusion, our study demonstrates a low prevalence of malignancy (14%-23%) among 1-2-cm nodules deemed indeterminate by two contrast-enhanced imaging scans. Limiting biopsy to those indeterminate 1-2-cm nodules with arterial hyperenhancement on at least one scan or in the presence of a synchronous typical HCC would detect the majority of malignant nodules while substantially reducing the number of biopsies. We advocate a strategy of close imaging follow-up for most indeterminate 1-2-cm nodules, with selective application of biopsy to the above-listed indications.

Ancillary