Potential conflict of interest: Nothing to report.
High-grade dysplastic nodules (HGDNs) are known to be premalignant lesions of hepatocellular carcinoma (HCC). We devised a model to estimate the long-term survival benefit of treating HGDNs by radiofrequency ablation (RFA) (Group I), as compared with regular follow-up and timely treatment by resection (Group II). A hypothetical 60-year-old compensated patient with cirrhosis was assumed. The system being modeled was assumed to be a simple Markov process, and state transition probabilities were given as parameters. Data used for simulation were obtained by a systematic review of the literature. The reported overall malignant transformation rates of HGDNs ranged from 12.5% to 80.8%, and were assumed to be 20%, 50%, and 80% using the best, moderate, and worst scenarios for Group II, respectively. The 5-year overall survival benefit of Group I compared with Group II was calculated by summing the detrimental effect of overtreatment and the beneficial effects of avoiding operative mortality. When the overall malignant transformation rate was set at 20%, 50%, or 80% the expected additional 5-year overall survival benefits of Group I compared with Group II were −0.05%, 0.20%, and 0.47%, respectively, and the corresponding additional 10-year overall survival benefits were 0.03%, 0.33%, and 0.55%, respectively. One-way sensitivity analysis showed that Group I was preferable to Group II in terms of 5-year overall survival when the 5-year overall malignant transformation rate was greater than 25.9%. Conclusion: No definite evidence indicates that the treatment of HGDNs by RFA provides additional long-term overall survival benefit as compared with regular follow-up and timely treatment. The findings of the present study concur with the present American Association for the Study of Liver Diseases guidelines. (HEPATOLOGY 2011 )
Hepatocellular carcinoma (HCC) is the sixth most common malignant tumor worldwide, but is the third leading cause of cancer-related mortality due to its poor prognosis.1 By virtue of improvements in imaging techniques and surveillance programs, the early detection rate of HCC has increased in recent decades.2, 3 In general, HCC is known to be developed stepwise from benign precursors such as regenerative nodules or dysplastic nodules.4-9 In particular, high-grade dysplastic nodules (HGDNs), formerly known as atypical adenomatous hyperplasia, are known to have high malignant potential, and are regarded as premalignant lesions of HCC.4, 5, 9-12 The premalignant nature of HGDNs can raise a question as to whether the tumors should be treated immediately or should be followed closely until a definitive diagnosis of HCC is obtained.13, 14
Although some investigators advocated the advantage of treatment of HGDN by radiofrequency ablation (RFA),14 the mortality and morbidity associated with RFA of liver tumors may not be negligible, and can raise ethical concerns regarding a possibility of overtreatment, despite the minimally invasive nature of RFA.15, 16 Other ablative therapies such as microwave ablation or percutaneous ethanol injection therapy (PEIT) also can be considered, but are inferior to RFA in terms of local tumor control.17-19 According to the latest guidelines from the American Association for the Study of Liver Diseases (AASLD), HGDNs should not be treated or managed as cancers.20
To secure a reliable contribution on this issue, a prospective randomized trial will be required. However, such a randomized trial may require enrollment of a huge number of patients, not to mention the ethical concerns. In the present study, by using a simulation model, we tried to estimate the long-term survival benefit of immediate treatment of HGDNs by RFA as compared with regular follow-up and timely treatment.
AASLD, American Association for the Study of Liver Diseases; BCLC, the Barcelona Clinic Liver Cancer; CT, computed tomography; FNB, fine needle biopsy; HCC, hepatocellular carcinoma; HGDN, high-grade dysplastic nodule; MELD, model for end-stage liver disease; MR, magnetic resonance; MVI, microvascular invasion; PEIT, percutaneous ethanol injection therapy; RFA, radiofrequency ablation; TVDT, tumor volume doubling time; WHO, World Health Organization.
Materials and Methods
We attempted to evaluate the possible advantage of RFA for compensated patient with cirrhosis with HGDN by estimating the additional overall survival benefits of immediate treatment by RFA as compared with regular follow-up and timely treatment by partial hepatectomy, which is known as the standard treatment. The primary endpoint was overall survival.
Selection of Trials.
We selected all articles published as full articles in English from 1985 to February 2011 in peer-reviewed journals that assessed the risk of malignant transformation of the premalignant lesions into HCC, the risk of development of tumor recurrence or microscopic regional tumor spread from HCCs, or the associated impacts of each prognostic factor on the survival of patients. Case reports were not included in the analysis. Studies were identified by searching MEDLINE on PubMed, the Cochrane Library database, and CANCERLIT (National Cancer Institute) using “dysplastic nodule,” “adenomatous hyperplasia,” “hepatocellular carcinoma,” “liver cancer,” or “primary liver carcinoma” as common text words combined with “hepatocarcinogenesis,” “recurrence,” “microsatellite,” or “microvascular invasion.”
We created a simulation model to evaluate the expected overall survival rates of hypothetical 60-year-old patients with cirrhosis and with Child-Pugh class A and with HGDN ≥1 cm treated immediately by percutaneous RFA (Group I), compared with those of patients regularly followed until a definite diagnosis of HCC is obtained and then treated by resection (Group II) (Fig. 1). In Group I all HGDNs were assumed to be completely eradicated by percutaneous RFA due to the relatively small tumor size and the absence of metastatic potential.12, 21-24 The baseline Model for Endstage Liver Disease (MELD) score was assumed to be ≤10.25
The system being modeled was assumed to be a simple Markov process, whereas state transition probabilities were given as parameters. Each cycle was set to be 1 year. The cycles were repeated for 10 times. All of the estimates of the variables used in the simulation model were extracted by a systematic review of published articles. However, the annual malignant transformation rates of HGDNs during the first 5-year follow-up were extrapolated for the rest of the follow-up period, because no long-term follow-up data are known for the tumors over 5 years as yet. The overall malignant transformation rates of HGDNs were estimated using the best, moderate, or worst scenario for Group II, respectively.
The overall survival benefit of Group I compared with Group II was calculated by summing the detrimental effect caused by overtreating premalignant lesions and the beneficial effects caused by the avoidance of operative mortality. One-way sensitivity analysis was performed to evaluate the effects of changing each single variable value, while the values of other parameters remained constant. Excel software (Microsoft 2003) was used for all the calculations and more details are described in the Supporting Information.26, 27
Diagnostic Criteria of HGDN.
Percutaneous ultrasonography-guided fine-needle biopsy (FNB) was assumed to be technically feasible for all HGDNs. The histologic diagnosis of HGDN was assumed to be made by expert pathologists.20, 27, 28 According to the revised World Health Organization (WHO) guideline, the histologic features of HGDN include cytological or structural atypia, which are insufficient for a confident diagnosis of HCC.28 According to the same guideline, HCCs are classified as early HCCs, which are small (<2 cm) well-differentiated HCCs without encapsulation, and progressed HCCs, which are more advanced malignant tumors.28
When the distinction between HGDN and early HCC was difficult,28 specific immunostaining techniques were assumed to be used.20, 28-33 In particular, distinction with high specificity was reported by the application of multiple tumor markers on liver biopsy specimens.29-31, 34-36 Enhanced dynamic computed tomography (CT) or magnetic resonance imaging (MRI) was obtained in all patients.37-39 Dysplastic nodules usually do not show early arterial enhancement on dynamic CT or MR images.6, 37-40 If typical HCC findings of enhancement in the arterial phase and wash-out in the delayed phase, or nodule-in-nodule appearance suggesting malignant foci were noted on the CT or MR images, such hepatic nodules were excluded from the analysis.38-41
Follow-up Protocol for the HGDNs.
In Group II, periodic sonography was assumed to be performed per 3 months.4, 5 If definite interval growth in tumor size or changes in the echogenicity of a nodule were noted, enhanced dynamic CT or MRI was performed.4, 5, 42 When a tumor showed any typical feature indicating a definitive diagnosis of HCC or dysplastic nodule with malignant foci, malignant transformation of HGDN into HCC was diagnosed radiologically.37, 40 If definite tumor growth was accompanied by short tumor volume doubling time (TVDT) ≤6 months, similar to that of overt HCC,43 malignant transformation of HGDN was clinically diagnosed. Histologic confirmation was considered only when a tumor did not show the clinical or radiological features of HCC described above.
Estimations of Prognostic Factors for the Simulation Model.
The procedure-related mortality rates of RFA for HCC have been reported to range from 0.1% to 0.5% in large-scale surveys (Table 1).15, 16, 44, 45 After pooling data, the procedure-related mortality rate was estimated to be 0.21% when the overlapping of cases was considered. The perioperative mortality of partial hepatectomy for HCC was estimated to be 0.8%, as determined by recent large-scale studies.46-48 The overall malignant transformation rates of HGDN reported in retrospective longitudinal studies vary widely from 12.5% to 80.8% (Table 2).4, 5, 9, 42, 49 The majority have reported rates greater than 70%, but the reported mean diameters of HGDNs range from 13 to 30 mm. In long-term follow-up studies, approximately one-third of HCCs were reported to be moderately differentiated and poorly differentiated HCCs were nearly absent.4, 5, 9
Table 1. Procedure-Related Mortality of Radiofrequency Ablation for Patients with Hepatocellular Carcinoma
Table 2. Estimations and Assumptions for the Prognostic Factors of Survival in Patients with High-Grade Dysplastic Nodule (HGDN)
Under the best, moderate, and worst scenario for regular follow-up and timely treatment of HGDNs by partial hepatectomy, as compared with immediate treatment by radiofrequency ablation, the 5-year overall malignant transformation rates of HGDNs were estimated to be 20%, 50%, and 80%, respectively.
Microvascular invasion (MVI) is known to be a potent adverse prognostic factor for survival of patients resected for HCC.23, 50-52 Although MVI usually does not occur in association with early HCC,12, 21 the rates of occurrence of MVI associated with progressed HCCs are known to range from 5% to 40% depending on tumor gross morphological type.22-24, 53, 54 In particular, the rates of microsatellite formation from single nodular types of HCC are generally much lower than those from other more invasive types of HCCs.22, 23, 53, 54 However, data are still lacking for the incidence and clinical impact of MVIs developed from malignantly transformed HGDNs, and we did not incorporate MVI into our simulation model as a prognostic factor for survival of patients with HGDN.55
Estimations of liver-related mortality or mortality in the general population were the same as those used in our previous simulation study on the Barcelona Clinic Liver Cancer (BCLC) stage 0 HCC, which is an asymptomatic solitary small HCC ≤2 cm (Table 2; see the Supporting Information).25-27, 56-61 In patients with HGDN, the incidence of HCC in locations other than the index HGDN has been estimated to be 50% during 5 years of follow-up.4, 14, 62, 63 The risk of biopsy-related tumor seeding is known to be nearly absent for the Tru-cut needle, and we assumed that this type of needle was used for all FNBs in the simulation study.20, 64, 65 Mortalities associated with FNB were estimated to be absent because such cases were not reported in large-scale studies that included over 1,000 patients.66, 67
When the overall malignant transformation rate of HGDN was assumed to be 20%, 50%, or 80%, the expected 5-year and 10-year overall survival benefits of Group I compared to Group II were calculated to be −0.05%, 0.20%, and 0.47%, respectively, and the corresponding 10-year overall survival benefits of Group I were calculated to be 0.03%, 0.33%, and 0.55%, respectively (Table 3). In particular, the benefit was less than 1% irrespective of the overall malignant transformation rate. One-way sensitivity analysis showed that Group I was preferable to Group II if the malignant transformation rate of HGDN exceeded 25.9% or 17.3% in terms of 5-year and 10-year overall survival, respectively (Table 4).
Table 3. Expected Additional 5-Year and 10-Year Overall Survival Benefits of Immediate Treatment of High-Grade Dysplastic Nodule (HGDN) by Radiofrequency Ablation (RFA) (Group I), as Compared with Regular Follow-up and Timely Treatment by Resection (Group II)
Overall 5-Year Malignant Transformation Rate
calculations were made under the variable assumptions for the 5-year overall malignant transformation rates of HGDNs. RFA, radiofrequency ablation.
Risks and benefits in terms of 5-year overall survival
Increased mortality in Group I due to overtreatment by RFA
Increased procedure-related mortality in Group II
Additional 5-year overall survival benefit of Group I compared with Group II
Risks and benefits in terms of 10-year overall survival
Increased mortality in Group I due to overtreatment by RFA
Increased procedure-related mortality in Group II
Additional 10-year overall survival benefit of Group I compared with Group II
Table 4. One-Way Sensitivity Analysis
List of variables and respective threshold values influencing the treatment strategy between immediate treatment by radiofrequency ablation (Group I) and regular follow-up and timely treatment of high-grade dysplastic nodule (HGDN) by resection (Group II). The threshold values other than the overall malignant transformation rate were obtained while the 5-year overall malignant transformation rate of HGDN was set to be 50%.
Five-year overall survival rates were calculated to be 72.6% in Group I and 72.1% to 72.6% in Group II, according to the estimated malignant transformation rates of HGDNs. For internal validation purposes for Group I, the 5-year overall survival rate has been reported to be 73.0%, which is nearly identical with the rate in the present study.16 External validation for Group I and internal and external validations for Group II were not possible due to data deficiencies.62
HGDN is a premalignant lesion of HCC, which raises the question of whether this lesion should be treated or followed up.13, 14 However, the malignant potential of HGDN is still uncertain, and no prospective study has been conducted on this issue. Although most retrospective studies have reported malignant transformation rates of 70% or more, the mean diameters of the HGDNs in those studies were much larger than the generally known values (≤1.2 cm), suggesting a possibility of selection bias.42, 62, 63, 68, 69 In contrast, the overall malignant transformation rate of HGDN was only 12.5% in an Italian study, where the mean diameter was 13 mm.49 This suggests that the wide variation in the malignant transformation rates of HGDNs may be, at least in part, attributed to the differences in the tumor size.
It may be straightforward that early treatment of premalignant lesions can lead to improved survival by reducing local tumor spread.14, 70 However, considering that some dysplastic nodules do not show any interval change over several years or even disappear during follow-up,6, 49 and that the mortality and morbidity associated with RFA of liver tumors may not be negligible,15, 16 early treatment of HGDN intrinsically can raise a concern about overtreatment.70 The current simulation study shows that the additional 5-year overall survival benefit in Group I compared with Group II was only 0.20% when the overall malignant transformation rate was set at 50%.
Although the histologic differentiation of HGDN and early HCC is difficult,41, 53 recent genetic and immunologic advances show that this differentiation can be performed with high specificity.27-33, 36, 71 Furthermore, false-negative diagnosis of HCC due to sampling error may be reduced by obtaining a hepatobiliary phase MRI, although further investigations are needed.6, 37, 40 The differentiation of HGDN and LGDN is another diagnostic challenge.34 However, misdiagnosis of LGDN as HGDN would reduce the apparent malignant transformation rate of HGDN, and further favor Group II as compared with Group I.
In this simulation study a well-compensated patients with cirrhosis and with Child-Pugh class A was assumed. For patients with liver fibrosis of greater severity, mortalities unrelated to the malignant transformation of HGDN would increase, further favoring Group II.25, 72 Furthermore, for patients with ≥4 HGDN nodules the ablation of all premalignant nodules is not feasible clinically, as it can compromise liver function.
In addition to the inherent limitations of a simulation study, the present study has the following potential limitations. First, this study was limited to HGDNs well visualized by ultrasonography, and the detection of small HGDNs (<1 cm) can be difficult by ultrasonography.73 Second, although MVI is a significant adverse prognostic factor for survival of patients with HCC,50, 52 the clinical impact of MVIs developed from malignantly transformed HGDNs during regular follow-up may be controversial.22, 53, 54 In particular, microscopic invasion of small vessels without a muscle layer located within 10 mm from the index HCC does not seem to have significant clinical impact on the survival of HCC patients following resection,55 although further prospective investigations are necessary on this issue. When we consider the multistep hepatocarcinogenesis,6 most MVIs developed from malignantly transformed HGDNs during regular follow-up are likely to be such minor MVIs, at least in our opinion. Third, we think that long-term follow-up for HGDNs over 5 years will be necessary to obtain a more potent conclusion on the issue of early treatment of HGDNs. Fourth, biopsy-related bleeding is an additional risk in Group II, but may be substantially reduced in the future due to the improved early diagnosis of HCC by liver MRI.37-40 Finally, the psychological stress due to the presence of a premalignant lesion can be remedied by treating HGDNs in Group I.
In conclusion, no definite evidence indicates that the treatment of HGDNs by RFA could provide a definite long-term survival benefit as compared with regular follow-up and timely treatment by resection, which is the current standard option. This study shows that the present AASLD guideline remains valid, but a careful individual analysis of cases with HGDN by expert teams is recommended in order to provide a tailored treatment strategy for each patient, based on size of the nodule and patients' psychological attitude.