Summary of findings
Description of the condition
Epidemiology and risk factors
Hepatocellular carcinoma (liver cancer) is the fifth most common cancer worldwide (El Serag 2007). The highest incidence of hepatocellular carcinoma is found in regions with endemic hepatitis B virus (HBV) such as Eastern Asia and Central Africa. A rising incidence of hepatocellular carcinoma in high-income countries has been reported, along with a decrease in high-incidence areas (El Serag 1999; El Serag 2007; Jepsen 2007). This might be caused by differences in hepatitis C virus (HCV) prevalence (e.g. increase in the USA) (Jepsen 2007), and increased HBV vaccination rates (e.g. decrease in China) (El Serag 2007). Overall, 75% to 80% of patients with primary liver cancer are attributable to persistent viral infections with either HBV (50% to 55%) or HCV (25% to 30%). High consumption of alcohol; cumulative amount of aflatoxins in the liver over time; and metabolic disorders such as non-alcoholic fatty liver disease, haemochromatosis, and alpha-1-antitrypsin deficiency are further risk factors for the development of cirrhosis and subsequent hepatocellular carcinoma (Sørensen 2003; Bosch 2004; EASL-EORTC 2012). Associations of primary liver cancer with diabetes (El Serag 2004), obesity (Calle 2003), and syndromes related to insulin resistance are subject of ongoing research, but their impact is currently unclear.
Liver cancer carries high mortality. In Europe and USA, the five-year survival is below 10% (Schoppmeyer 2009). In most patients, hepatocellular carcinoma is diagnosed at late stages of the disease and is mostly accompanied by liver cirrhosis. In high-income countries, only about 30% of patients present with the more favourable early hepatocellular carcinoma (Bruix 2001). Percutaneous ablation techniques, surgical resection, and liver transplantation are currently considered potentially curative treatments for early hepatocellular carcinoma.
Due to the neovascularity of hepatocellular carcinoma and its characteristic arterial hypervascularity and venous/late phase washout (so-called hepatocellular carcinoma radiological hallmark) in rapid-sequence cross-sectional imaging, an accurate diagnosis of hepatocellular carcinoma lesions larger than 2 cm in diameter can be made non-invasively in patients with cirrhosis. According to the current guidelines of the European Association for the Study of the Liver (EASL) and the European Organisation for Research and Treatment of Cancer (EORTC) (EASL-EORTC 2012), non-invasive diagnostic criteria for patients with cirrhosis and hepatocellular carcinoma are:
- for lesions larger than 2 cm: one positive imaging technique (four-phase computed tomography or dynamic contrast enhanced magnetic resonance imaging (MRI))
- for lesions from 1 to 2 cm: two coincidental techniques (computed tomography, MRI, or contrast-enhanced ultrasound.
Image-guided biopsies of the tumour are common practice in cases of inconclusive radiological findings, that is, a lesion between 1 and 2 cm in diameter identified by only one imaging modality and such biopsies are mandatory for tumours in non-cirrhotic livers (Takamori 2000; EASL-EORTC 2012). Elevated serum levels of alpha-fetoprotein (AFP) that were proposed in addition to one typical radiological finding at the beginning of the century (Bruix 2001) have been discarded from the diagnostic scheme by the American Association for the Study of Liver Disease (AASLD) (Bruix 2005). Conversely, it is still regarded to be useful by EASL and EORTC (EASL-EORTC 2012).
Description of the intervention
Ablation techniques rely on the possibility to induce cell death by thermal or chemical means. The aim of the ablation is to completely destroy all tumour cells and thus provide cure of the cancer. The underlying liver disease (e.g. cirrhosis or hepatitis) is not affected by the treatment.
Radiofrequency ablation or radiofrequency thermal ablation (RFA) uses frictional heat to induce cell death from coagulation necrosis. After local anaesthesia of the skin, an RFA-electrode is placed into the liver lesion under the guidance of magnetic resonance tomography, computed tomography, or ultrasound (Goldberg 2000). A radiofrequency-current-generator is connected to the needle and a zero-electrode is fixed onto the patient's back. Under short-term analgesia-sedation, energy is applied for a few minutes leading to spheric lesions of about 3 to 5 cm in diameter. When temperatures pace 60 °C, a coagulation necrosis is induced (Minami 2010; EASL-EORTC 2012). Several investigations have shown the efficacy and the safety of the procedure (Rossi 1996; Rossi 1998; Allgaier 1999; Curley 1999; Francia 1999; Goldberg 2000; Grasso 2000; Livraghi 2000; Nicoli 2000). If necessary, the procedure can be repeated to treat larger lesions or patients with more than one liver tumour.
Surgical resection demands a radical resection of the tumours. This is often not feasible due to the impairment of liver function caused by the underlying cirrhosis. Local recurrence and de novo hepatocellular carcinoma at other locations within the liver are about 50% at three years and 70% at five years (Arii 2000; Bismuth 2000; Llovet 2000).
Orthotopic liver transplantation is an alternative approach for selected patients with small hepatocellular carcinoma. The beneficial removal of liver cirrhosis as a predisposing factor for hepatocellular carcinoma is, however, often counterbalanced by progression of the tumour while the patient is waiting for a new organ. Moreover, organ shortage is a major factor limiting the availability of this procedure. The Milan criteria (one tumour of 5 cm or less in diameter, or three tumours with a diameter of 3 cm or less each) are the broadly accepted standard to identify patients suitable for liver transplantation (Mazzaferro 1996). Percutaneous techniques are used with increasing frequency to bridge the waiting time until transplantation (Dubay 2011).
Medical treatment options: Sorafenib, a multi-kinase inhibitor improved median survival by 2.8 months in one trial (Llovet 2008), and was approved for treatment of advanced hepatocellular carcinoma in 2007 (Wörns 2009). Sorafenib has not yet been compared to other interventions for hepatocellular carcinoma in randomised trials.
Percutaneous ethanol injection (PEI) causes dehydration and necrosis of tumour cells, accompanied by small vessel thrombosis, leading to tumour ischaemia and destruction. PEI is usually carried out under ultrasound guidance with repeated injections of ethanol on separate days. Best results for PEI are achieved in single hepatocellular carcinoma lesions of less than 3 cm in diameter, for which complete remission rates of 70% can be expected. Lower occurrences of complete remission have been observed in larger and multinodular tumours (Livraghi 1995; Lencioni 1997). In one Cochrane review with meta-analysis, we concluded that there was insufficient evidence to determine whether PEI or segmental liver resection was more effective (Schoppmeyer 2009). In a recent up-date of the Schoppmeyer 2009 review, no additional information could be added and there is little evidence to expect such evidence within the next years (Weis 2013).
Percutaneous acetic acid injection (PAI) has been used as an alternative for ethanol in percutaneous treatment (Ohnishi 1998). Contraindications for PEI and PAI are cirrhosis with poor liver function (Child C cirrhosis), complete portal vein thrombosis, and massive ascites. We chose to analyse data from trials using PEI and PAI together because of the similarities between these two interventions in contrast to other treatment modalities. However, when results from individual trials are reported, we will report if PEI or PAI was used. In subgroup analysis, we will assess if there seems to be a difference in the overall survival in trials comparing RFA versus PEI or PAI.
Other interventions: Percutaneous cryoablation or interstitial laser photocoagulation are not yet routine clinical practice.
Why it is important to do this review
Treatment options of hepatocellular carcinoma have substantially increased since the introduction of ablation therapies in the 1990ies. While PEI was still the preferred treatment option at the beginning of the century (Bruix 2001), contradicting results have been reported in recently performed trials. Some were in favour of RFA (Lin 2004 (RFA vs. PEI); Lin 2005 (RFA vs. PEI), while others showed no differences in the outcome between RFA and PEI (Brunello 2008; Giorgio 2011). In addition, the effects of RFA compared with hepatic resection for hepatocellular carcinoma are unclear. The current recommendations by the EASL and the EORTC are to use hepatic resection as a primary treatment and to use RFA when surgery is not possible (EASL-EORTC 2012). As new trials have been published, we thought that this might change the evidence.
To assess the beneficial and harmful effects of RFA versus placebo, no intervention, or any other therapeutic approach in patients with hepatocellular carcinoma.
Criteria for considering studies for this review
Types of studies
We included randomised clinical trials, irrespective of publication status or blinding. We applied no language limitations.
Types of participants
All patients with hepatocellular carcinoma without contraindications for RFA (e.g. too many or too large tumours, extrahepatic malignant manifestation).
Types of interventions
RFA compared with placebo, no intervention, or any other intervention.
Types of outcome measures
- Overall survival.
- Two-year survival.
- Event-free survival concerning local recurrences, distant recurrence, de novo (a new) carcinoma, death.
- Local progression/recurrences.
- Frequency of complications/serious adverse events. Serious adverse events: Number of participants with serious adverse events defined as "any untoward medical occurrence that at any dose results in death, is life-threatening, requires inpatient hospitalisation or prolongation of existing hospitalisation, results in persistent or significant disability/incapacity". All other adverse events were be considered as non-serious (ICH-GCP 1997).
- Quality of life.
- Duration of hospital stay/health economics.
Search methods for identification of studies
We performed electronic searches in the Cochrane Hepato-Biliary Group Controlled Trials Register (Gluud 2013), the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, and ISI Web of Science (Science Citation Index Expanded) (Royle 2003) to 3 September 2012. We did not update searches in CancerLit and Current Contents as well as searches of the journals (Zeitschrift für Gastroenterologie (suspended in 2000) and Endoskopie Heute (suspended in 1999). CancerLit is no longer updated and Current Contents has merged into ISI Web of Science. We used the search strategies with the time spans shown in Appendix 1, which ended September 2012.
Searching other resources
We also handsearched proceedings of conferences of five oncological and hepatological societies (ASCO, ESMO, AASLD, EASL, APASL), as well as references of articles. We contacted researchers in the field and companies producing RFA equipment for information on further (unpublished) trials.
Data collection and analysis
We performed this systematic review according to a peer-reviewed, published protocol (Galandi 2001), and followed the recommendations of The Cochrane Collaboration (Higgins 2011), and the Cochrane Hepato-Biliary Group Module (Gluud 2013). We performed the analyses using Review Manager 5 (RevMan 2012).
Selection of studies
One review author (SW) collected the publications of all trials potentially relevant for this review. Two review authors (SW, KS) independently selected trials for inclusion.
Data extraction and management
Two review authors (SW, AF) independently extracted the data from all trials. We included data from the first versions of the review by Galandi (Galandi 2002; Galandi 2004). We contacted the principal authors of the trials to retrieve missing data. Where there were missing data for an appropriate judgement of the risk of bias, we judged the risk of bias as 'high'. We accepted the original trials' definitions for all beneficial and harmful outcomes although we attempted to extract information on all outcomes specified for this review. We resolved disagreements by discussion. We extracted type and number of serious adverse events, length of hospitalisation, and quality of life if available.
Assessment of risk of bias in included studies
Two review authors (SW, KS) assessed the risk of bias of all trials included in this review, following the risk of bias domains, with definitions presented below (Schulz 1995; Moher 1998; Kjaergard 2001; Wood 2008; Lundh 2012; Savovic 2012; Savovic 2012a; Gluud 2013).
Allocation sequence generation
- Low risk of bias: sequence generation was achieved using computer random number generation or a random number table. Drawing lots, tossing a coin, shuffling cards, and throwing dice were adequate if performed by an independent research assistant not otherwise involved in the trial.
- Uncertain risk of bias: the method of sequence generation was not specified.
- High risk of bias: the sequence generation method was not random.
- Low risk of bias: the participant allocations could not have been foreseen in advance of, or during, enrolment. Allocation was controlled by a central and independent randomisation unit. The allocation sequence was unknown to the investigators (e.g. if the allocation sequence was hidden in sequentially numbered, opaque, and sealed envelopes).
- Uncertain risk of bias: the method used to conceal the allocation was not described so that intervention allocations may have been foreseen in advance of, or during, enrolment.
- High risk of bias: the allocation sequence was likely to be known to the investigators who assigned the participants.
Blinding of participants, personnel, and outcome assessors
- Low risk of bias: blinding was performed adequately, or the assessment of outcomes was not likely to be influenced by lack of blinding.
- Uncertain risk of bias: there was insufficient information to assess whether blinding was likely to induce bias on the results.
- High risk of bias: no blinding or incomplete blinding, and the assessment of outcomes were likely to be influenced by lack of blinding.
Incomplete outcome data
- Low risk of bias: missing data were unlikely to make treatment effects depart from plausible values. Sufficient methods, such as multiple imputation, had been employed to handle missing data.
- Uncertain risk of bias: there was insufficient information to assess whether missing data in combination with the method used to handle missing data were likely to induce bias on the results.
- High risk of bias: the results were likely to be biased due to missing data.
Selective outcome reporting
- Low risk of bias: all outcomes were pre-defined (e.g. in a published protocol) and reported, or all clinically relevant (e.g. overall survival, event-free survival, or adverse events) were reported.
- Uncertain risk of bias: it was unclear whether all pre-defined and clinically relevant outcomes were reported.
- High risk of bias: one or more clinically relevant outcomes were not reported, and data on these outcomes were likely to have been recorded.
Vested interest bias
- Low risk of bias: the trial appeared to be free of other components (should be listed, e.g. industry bias, academic bias, etc.) that could put it at risk of bias.
- Uncertain risk of bias: it was unclear whether the trial was at risk of vested interest bias.
- High risk of bias: there were other factors in the trial that could put it at risk of bias (e.g. for-profit involvement, study authors had conducted trials on the same topic, etc.).
- Low risk of bias: no other potential source of bias could be detected.
- Uncertain risk of bias: potential sources of bias could not be ruled out (e.g. missing outcome definition, unclear baseline differences, etc.).
- High risk of bias: other relevant sources of bias were present (e.g. small cohorts, significant baseline differences, etc.).
A trial was considered with a low risk of bias, if it was assessed with low risk of bias in all of the specified bias risk domains, and a trial was considered with high risk of bias, if one or more of the specific domains were assessed to be unclear or with high risk of bias.
We solved disagreements by discussion, or by contacting the Cochrane Hepato-Biliary Group in Copenhagen.
Units of analysis issues
We did not include cluster randomised trials. We did not expect or identify cross-over trials due to the nature of the disease and the interventions investigated. We included trials with three or more treatment groups if pair-wise comparison of a single intervention versus RFA was possible and if inclusion criteria of both intervention groups fulfilled the inclusion criteria. In order to exclude analysis bias by multiple counting of the shared intervention group, the shared intervention group was split into a corresponding number of subgroups with smaller sample size. We analysed and included each pair-wise comparison separately. We measured two survival outcomes. We analysed recurrence-free survival and overall survival separately.
Dealing with missing data
We considered missing data in the judgement of selective and incomplete reporting bias. In case of missing data that prevented trials from being included, we contacted trialists and requested missing data. In addition, we aimed to obtain hazard ratios (HRs) and standard deviations or Kaplan Meier survival plots for the survival outcome measures when these were not available from the original publication.
Assessment of heterogeneity
We performed analyses of heterogeneity between trials by calculating the Chi
Assessment of reporting biases
We planned to do funnel plots if more than 10 trials were identified to give an overview regarding potential reporting bias and other bias sources (Higgins 2011).
We performed the meta-analyses according to the recommendations of The Cochrane Collaboration (Higgins 2011), and the Cochrane Hepato-Biliary Group Module(Gluud 2013). We extracted HRs as relevant effect measures for overall survival, two-year survival, event-free survival, and local recurrences with 95% confidence intervals (CI) from publications. Alternatively, we estimated HRs from log rank Chi
We assigned trials to subcategories according to their type of control intervention. We presented pooled effects within each subcategory and as overall estimate of effects. We analysed trials comparing RFA versus two other interventions within a three-armed design in that both comparisons were included independently within the respective subcategory using the given - or estimated - effect differences. The sample size or events or both, observed in the RFA group, allowed only once for the overall analysis to maintain the total sample sizes or events.
We used random-effects models to analyse the effect sizes (DerSimonian 1986). In comparison to fixed-effect model meta-analyses, these are more conservative and result in wider CIs of the meta-analysed effect estimates, in order to avoid over-optimistic significances. In contrast, random-effects models put too much weight on small trials (which are often more biased). Therefore, we also conducted fixed-effect model meta-analysis and reported results from both models if the results differed. In the absence of relevant heterogeneity, both models will result in nearly identical effect estimations.
For all available outcomes, we presented data tables or forest plots. The risk of bias within the trials was considered a potential source of heterogeneity, and, therefore, it was assessed in detail according to the methodology described above. We intended to perform a subgroup analysis to compare trials with low risk of bias with the overall evidence received from all suitable trials.
We performed statistical analyses using Review Manager 5 Software (RevMan 2012).
Trial sequential analysis
We applied trial sequential analysis (CTU 2011; Thorlund 2011) because cumulative meta-analyses are at risk of producing random errors due to sparse data and repetitive testing of the accumulating data (Wetterslev 2008). To minimise random errors, we calculated the required information size (i.e. the number of participants needed in a meta-analysis to detect or reject a certain intervention effect) (Wetterslev 2008). The required information size calculation should also account for the heterogeneity or diversity present in the meta-analysis (Wetterslev 2008; Wetterslev 2009). In our meta-analysis, the required information size was based on the event proportion in the control group; assumption of a plausible RR reduction of 20%, or on the RR reduction observed in the included trials with low risk of bias; a risk of type I error of 5%; a risk of type II error of 20%; and the assumed diversity of the meta-analysis (Brok 2008; Wetterslev 2008; Brok 2009; Thorlund 2009, Wetterslev 2009; Thorlund 2010). The underlying assumption of trial sequential analysis was that testing for significance may be performed each time a new trial is added to the meta-analysis. We have added the trials according to the year of publication, and if more than one trial has been published in a year, we added trials alphabetically according to the last name of the first author. Based on the required information size, we constructed trial sequential monitoring boundaries (Brok 2008; Wetterslev 2008; Brok 2009; Thorlund 2009; Wetterslev 2009; Thorlund 2010). These boundaries determined the statistical inference one may draw regarding the cumulative meta-analysis that has not reached the required information size; if the trial sequential monitoring boundaries were crossed before the required information size was reached, firm evidence may perhaps be established and further trials may turn out to be superfluous. In contrast, if the boundaries were not surpassed, it was most probably necessary to continue doing trials in order to detect or reject a certain intervention effect if the trial sequential monitoring boundaries for futility were not crossed. If the latter was the case, futility could be declared (Brok 2008; Wetterslev 2008; Brok 2009; Thorlund 2009; Wetterslev 2009; Thorlund 2010).
Summary of findings tables
We summarised the evidence in the 'Summary of findings' tables using GRADEpro (ims.cochrane.org/revman/other-resources/gradepro). We used the GRADE classification in order to judge the quality of evidence.
Description of studies
Results of the search
Our literature search revealed 2008 references. We removed 718 duplicates. From the remaining 1290 references, we identified 11 randomised clinical trials that fulfilled our inclusion criteria (Characteristics of included studies). We identified no additional trials from other sources. Two trials still await classification with insufficient information that were reported in meeting abstracts available to date and with no additional information from the study authors (Di Costanzo 2011; Kuansheng 2011; Characteristics of studies awaiting classification). One trial was still ongoing (NCT00814242; Characteristics of ongoing studies). We excluded 23 studies due to insufficient design or lack of relevant information (Figure 1; Characteristics of excluded studies).
|Figure 1. Study flow diagram of the literature search as well as number and reasons of excluded studies.|
We included and analysed 11 trials in this review. The trials included 1819 randomised participants. We found no trials comparing RFA versus placebo, no intervention, chemotherapy, or liver transplantation.
Six trials with 1088 participants compared RFA versus PEI (Lencioni 2003; Lin 2004 (RFA vs. PEI); Lin 2005 (RFA vs. PEI); Shiina 2005; Brunello 2008; Giorgio 2011). Two of the trials with 344 participants randomised participants to three intervention groups (Lin 2004 (RFA vs. PEI); Lin 2004 (RFA vs. PEI-hd); Lin 2005 (RFA vs. PEI); Lin 2005 (RFA vs. PAI)). In one trial, RFA was compared with a standard dose of PEI (Lin 2004 (RFA vs. PEI)), and versus a high dose of PEI (Lin 2004 (RFA vs. PEI-hd)). In the other trial, RFA was compared with PEI (Lin 2005 (RFA vs. PEI)) and versus PAI (Lin 2005 (RFA vs. PAI)).
Regarding the comparison of RFA versus other interventions, 10 out of 11 comparisons (including also the two three-armed trials) reported on the primary outcome, overall survival ( Analysis 1.1). Two-year survival could be extracted from seven trials (with nine comparisons) ( Analysis 2.1). Data on event-free survival were available from eight trials and on local progression were available from six trials ( Analysis 3.1; Analysis 4.1). Rates of major complications or procedure-related deaths, or both, were reported in all trials.
Reasons for excluding identified studies were as follows: one trial compared RFA alone with RFA plus hepatic arterial occlusion (Kobayashi 2007), one with RFA plus PEI (Zhang 2007), and two with RFA plus transarterial chemo-embolisation (TACE) (Morimoto 2010; Morimoto 2011). Ten studies were not randomised clinical trials (Buscarini 1996; Goldberg 1998; Cuschieri 1999; Jiao 1999; Livraghi 1999; Kurokohchi 2005; Zhang 2005; Lü 2006; Khan 2007; Ohmoto 2009). Five studies were retrospective cohort studies (Amarnath 2006; Ueno 2009; Opocher 2010; Sreenivasan 2010; Gory 2012). One publication was retracted (Cheng 2008). One study did not include patient data but was a Markov model analysis (Cho 2010). Akamatsu et al switched patients from PEI to RFA during the study (Akamatsu 2004). It was unclear whether randomisation was performed in two trials (Gan 2004; Chen 2005). We contacted the study authors but they did not reply and, therefore, we did not include the trials (Gan 2004; Chen 2005). RFA was also used to treat portal vein thrombosis in one study and, therefore, the effects of RFA on tumour growth could not be extracted (Giorgio 2011a). Peng et al only included patients with recurrent hepatocellular carcinoma. As this cohort displays differences to patients with newly diagnosed hepatocellular carcinoma, we did not include this trial (Peng 2012). One trial used laparoscopic RFA and was excluded (Santambrogio 2012).
Risk of bias in included studies
Overall, risk of bias was low in six trials and high in five trials. We only included the low risk of bias trials in the subgroup analysis.
Generation of allocation sequence and allocation concealment
Allocation generation was done by a computer in seven trials, by random numbers in three trials (Shiina 2005; Giorgio 2011; Feng 2012), and the method of the allocation generation was not reported in one trial (Shibata 2002). The Shibata trial was described as a randomised trial. However, we were unable to obtain the missing information on allocation generation by the study authors. Concealed allocation was ensured by using a computer-based procedure in six trials, by sealed envelopes in three trials (Shibata 2002; Brunello 2008; Huang 2010), and by a coded list in one trial (Giorgio 2011). For the remaining trial, the study authors did not describe if allocation concealment was performed (Ferrari 2007).
Our most important outcomes were survival and tumour recurrence, which are unlikely to be significantly influenced by lack of blinding of both participants and outcome assessors. Moreover, as treatment modalities such as RFA and hepatic resection differ substantially with regard to the degree of invasiveness, procedure performance, etc., reasonable blinding to the treatment of an investigator or an informed patient is unlikely. Objective outcomes such as adverse events were blinded in one trial only and, therefore, only this trial was judged as a low risk of bias trial (Giorgio 2011); all other trials had a high risk of bias ( Analysis 5.1).
Incomplete outcome data
An intention-to-treat (ITT) analysis was explicitly stated in five trials (Shiina 2005; Chen 2006; Brunello 2008; Huang 2010; Feng 2012). Chen et al reported on 19 patients randomised to RFA who had later been subjected to resection (Chen 2006). These participants were excluded from the primary analysis. Results were adequately presented for both the ITT and the per-protocol analysis. The ITT results were considered for the meta-analysis of the data. Similarly, in Huang 2010, seven patients were switched from RFA to the resection group on their own decision but were analysed according to ITT (used in this review) and per-protocol. In two trials, it was unclear whether incomplete outcome data had been adequately addressed (Shibata 2002; Ferrari 2007).
Ten trials were free of selective outcome reporting bias as the most relevant clinical outcome (overall survival was not reported in Shibata 2002). The other 10 trials reported outcomes in a sufficiently detailed manner to allow estimations of HR based on Parmar's method (Parmer 1998).
Vested interest bias
Relevant vested interest bias such as industry funding or relevant academic bias was unclear in two trials (Ferrari 2007; Shibata 2002), and not present in any of the trials as stated in the manuscript or directly by the study authors. Lin et al. performed two randomised clinical trials during the same time period. Patients were assigned to one of the trials according to their hospital admission week (Lin 2004 (RFA vs. PEI); Lin 2005 (RFA vs. PEI)) However, we judged this to have low risk of bias.
Other source of bias
Other sources of bias, namely possible premature stopping (Brunello 2008), an unusual high number of patients switching groups (Chen 2006) or undefined outcomes (Ferrari 2007) were present in three trials and we judged them to have an unclear risk of bias.
Overall bias risk assessment
|Figure 2. Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies (survival outcomes).|
|Figure 3. Risk of bias summary: review authors' judgements about each risk of bias item for each included study (survival outcomes).|
Effects of interventions
See: Summary of findings for the main comparison Hepatic resection compared with radiofrequency ablation; Summary of findings 2 Percutaneous ethanol injection compared with radiofrequency ablation for early hepatocellular carcinoma; Summary of findings 3 Clinical characteristics of patients treated with radiofrequency ablation or percutaneous ethanol injection
Radiofrequency ablation versus hepatic resection
Comparison for efficacy outcomes
Three randomised clinical trials were identified that compared RFA versus hepatic resection (Chen 2006; Huang 2010; Feng 2012), of which two had low risk of bias (Huang 2010; Feng 2012), and one had high risk of bias (Chen 2006). Analysing the results from all three trials using a random-effects model showed no significant difference between the compared interventions regarding overall survival (HR 0.71; 95% CI 0.44 to 1.15) ( Analysis 1.1) and two-year survival (HR 0.51; 95% CI 0.24 to 1.08) ( Analysis 2.1). Meta-analysis using a fixed-effect model showed a result in favour of hepatic resection compared with RFA on overall survival (HR 0.76; 95% CI 0.58 to 1.00) ( Analysis 1.2). In addition to the calculation of the HR, dichotomous outcomes on overall survival could be extracted from the two trials with low risk of bias (Huang 2010; Feng 2012; Analysis 1.3). In concordance with the HR analysis, resection yielded statistically significant improved survival both using the fixed-effect and random-effects model meta-analyses (RR fixed 0.61; 95% CI 0.44 to 0.82 and RR random 0.60; 95% CI 0.44 to 0.82).
Trial sequential analysis
Heterogeneity was present (I
In a subgroup analysis that included low-risk of bias trials only, results indicated that hepatic resection yields better results than RFA regarding overall survival (HR 0.56; 95% CI 0.40 to 0.78) ( Analysis 1.3) and two-year survival (HR 0.38; 95% CI 0.17 to 0.84) ( Analysis 2.2). Regarding event-free survival, the results indicated that hepatic resection seemed better than RFA (HR 0.70; 95% CI 0.54 to 0.91) ( Analysis 3.1). Only one trial reported on local progression and found hepatic resection was superior to RFA (HR 0.48; 95% CI 0.28 to 0.82) ( Analysis 4.1).
Comparison for safety outcomes
Comparison for economic outcomes
Our results show that RFA was associated with shorter duration of hospital stay compared with surgery (standardised mean difference 2.18 days; 95% CI 1.97 to 2.39).
Radiofrequency ablation versus percutaneous ethanol injection/percutaneous acetic acid injection
Comparison for efficacy outcomes
Six trials compared RFA with PEI of which four had a low risk of bias (Lencioni 2003; Lin 2004 (RFA vs. PEI); Lin 2005 (RFA vs. PEI); Shiina 2005) and two had a high risk of bias (Brunello 2008; Giorgio 2011). The overall effects of RFA seemed superior to PEI regarding overall survival (HR 1.64; 95% CI 1.31 to 2.07) ( Analysis 1.1), two-year survival (HR 1.82; 95% CI 1.34 to 2.47) ( Analysis 2.1), event-free survival (HR 1.55; 95% CI 1.31 to 1.85) ( Analysis 3.1), and local progression (HR 2.44; 95% CI 1.71 to 3.49) ( Analysis 4.1). No significant difference was found if only the result from the four trials with low risk of bias were meta-analysed (overall survival: HR 1.19; 95% CI 0.79 to 1.77). One trial compared the two interventions for single and very small (2 cm or less) hepatocellular carcinoma and found no differences in overall survival (Giorgio 2011) ( Summary of findings 2).
In addition to the calculation of the HR, dichotomous outcomes (RR) on overall survival were calculated from all six trials. There was no significant difference between the RFA group versus the PEI/PAI group (RR 1.76; 95% CI 0.97 to 3.22) ( Analysis 6.1).
In addition, a subgroup analysis was performed in order to compare the results for overall survival of RFA versus PEI and RFA versus PAI. RFA versus PAI was only investigated in one three-armed trial (Lin 2005 (RFA vs. PAI)), as compared to the six trials that investigated RFA versus PEI (Lencioni 2003; Lin 2004 (RFA vs. PEI); Shiina 2005; Lin 2005 (RFA vs. PEI); Brunello 2008; Giorgio 2011). In a test of subgroup differences, we found no significant difference between the effect of RFA versus PEI or PAI (P value = 0.75) ( Analysis 1.5). The body of evidence was not changed if the comparison of RFA versus PAI was excluded ( Analysis 1.5; Analysis 5.2; Analysis 5.5).
Trial sequential analysis on overall mortality
Heterogeneity was not detected (I
A subgroup analysis including low-risk of bias trials only, was performed for all comparisons, but it did not change the body of evidence ( Analysis 2.3; Analysis 3.2; Analysis 4.1; Analysis 5.3; Analysis 5.4).
A separate post-hoc analysis of the three trials conducted in Italy compared to the three trials conducted in Asia that compared RFA with PEI was performed. In Italian patients, there was no difference between treatment groups (HR 1.24; 95%; CI 0.84 to 1.83), while in Asian patients, the RFA seemed superior to PEI (HR 1.95; 95% CI 1.38 to 2.75). We found no statistically significant difference between the two subgroups (P value = 0.09) ( Analysis 6.3).
Comparison for safety outcomes
All studies reported on minor complications. In six trials (Lencioni 2003; Lin 2004 (RFA vs. PEI); Lin 2004 (RFA vs. PEI-hd); Lin 2005 (RFA vs. PAI); Lin 2005 (RFA vs. PEI); Shiina 2005; Brunello 2008; Giorgio 2011), serious adverse events occurred. One trial reported a procedure-related death within the PEI group (Brunello 2008). The proportion of patients with a serious adverse event did not differ significantly between the compared intervention groups (PEI/PAI versus RFA; OR 0.70; 95% CI 0.33 to 1.48) ( Analysis 5.1).
Comparison for economic outcomes
Duration of hospital stay was documented in three trials including five comparisons (Lin 2004 (RFA vs. PEI); Lin 2004 (RFA vs. PEI-hd); Lin 2005 (RFA vs. PAI); Lin 2005 (RFA vs. PEI); Shiina 2005). However, the settings of the trials differed due to regional habits. While Shiina 2005 treated all patients as inpatients until treatment success was documented by computed tomography, PEI/PAI patients in the Lin trials were routinely treated as outpatients in the cases when there was no severe adverse effect during the first PEI/PAI session (Lin 2004 (RFA vs. PEI); Lin 2005 (RFA vs. PEI)). Shiina 2005 reported a shorter hospitalisation time within the RFA group. These patients showed a shorter mean duration of hospital stay for the few complications that occurred. Due to the heterogeneous settings - also demonstrated by high Chi
Radiofrequency ablation versus percutaneous microwave coagulation
Comparison for efficacy outcomes
Only one trial examined the effects of RFA versus percutaneous microwave coagulation. Shibata et al randomised 72 patients with 94 hepatic tumour nodules to RFA versus percutaneous microwave coagulation (Shibata 2002). Data on overall survival, two-year survival, and event-free survival could not be extracted as the data presentation was based on tumour nodules and were not done patient-wise. There was no significant difference between RFA and percutaneous microwave coagulation regarding local progression (HR 2.14; 95% CI 0.67 to 6.80) ( Analysis 4.1).
Comparison for safety outcomes
No death or other serious adverse events were observed during the trial in the RFA or percutaneous microwave coagulation group. In the RFA group, one patient had a segmental hepatic infarction. In the percutaneous microwave coagulation group, one patient each developed liver abscess, cholangitis, subcutaneous abscess, and skin burn. The OR of major complications was 4.38 (95% CI 0.46 to 41.22).
Comparison for economic outcomes
There were no data for economic outcomes. Data on the duration of hospital stay were not presented.
Radiofrequency ablation versus percutaneous laser ablation
One trial randomised 40 patients to RFA versus 41 patients to laser ablation (Ferrari 2007). There were no significant differences between RFA and percutaneous laser ablation with regards to overall survival (HR 1.62; 95% CI 0.62 to 4.22) ( Analysis 1.1), event-free survival (HR 1.20; 95% CI 0.50 to 2.89) ( Analysis 3.1), and local progression (HR 1.12; 95% CI 0.40 to 3.09) ( Analysis 4.1)..
Comparison for safety outcomes
There were no data reported on safety outcomes.
Comparison for economic outcomes
There were no data reported on economic outcomes.
Quality of life
None of the identified trials reported on quality of life.
Hepatocellular carcinoma represents a major health problem worldwide. Various therapeutic approaches such as percutaneous, transarterial, and surgical interventions are available. Prognosis and therapeutic options of patients with hepatocellular carcinoma depend on tumour extension and liver function, which is often reduced by underlying liver cirrhosis. Though sophisticated prognostic classifications such as the Barcelona Clinic Liver Cancer (BCLC), a more pragmatic approach based on number and size of tumour lesions is widely used to stratify patients (EASL-EORTC 2012). The Milan criteria define "early hepatocellular carcinomas" as a single tumour of 5 cm or less or up to three tumours of 3 cm or less (Mazzaferro 1996). The Milan criteria are also used to evaluate patients for orthotopic liver transplantation. Advanced stages of disease include extensive liver involvement and extrahepatic disease. Intermediate stages are ill-defined to date and belong to none of the former two groups. Furthermore, all treatments for hepatocellular carcinoma (excluding orthotopic liver transplantation) are reserved for patients with mild and moderate functional impairment of the liver. In our review, we aimed to determine the role of RFA as compared with any other intervention in patients with 'early hepatocellular carcinoma'. The primary outcome of our analyses was overall survival, which is the most relevant outcome for patients.
Summary of main results
In contrast to two former versions of this systematic review (Galandi 2002; Galandi 2004), new data on RFA for hepatocellular carcinoma has accumulated so that 11 randomised clinical trials have assessed RFA: three trials compared RFA versus resection (Chen 2006; Huang 2010; Feng 2012); six trials RFA versus PEI (Lencioni 2003; Lin 2004 (RFA vs. PEI); Lin 2005 (RFA vs. PEI); Shiina 2005; Brunello 2008; Giorgio 2011); one trial RFA versus PAI (Lin 2005 (RFA vs. PAI); one trial RFA versus microwave ablation (Shibata 2002); and one trial RFA versus laser ablation (Ferrari 2007). This represents an expansion of nine new trials with 1643 new participants.
Radiofrequency ablation versus hepatic resection
Our Cochrane systematic review included three randomised clinical trials comparing RFA versus surgical resection. Two of the identified trials had a low risk of bias (Huang 2010; Feng 2012), while the third trial had a high risk of bias (Chen 2006).
Analysing all three trials, there was no significant difference in the HR and the RR for overall survival ( Analysis 1.1) and the HR for two-year survival ( Analysis 2.1) between the treatment groups, but we found the outcome to be in favour of the surgically treated patients regarding event-free survival ( Analysis 3.1) and local progression ( Analysis 4.1). In the analysis of the two trials with a low risk of bias, hepatic resection resulted in an increased overall survival ( Analysis 2.2) and increased two-year survival ( Analysis 2.2) when compared with RFA using traditional and naive statistical boundaries. We conclude that there is moderate-quality evidence from randomised clinical trials that hepatic resection yields better results regarding benefits. However, the rate of complications is remarkably higher and the hospital stay is longer in the surgical resection group ( Analysis 5.1). In order to become certain about the balance between benefits and harms, it would be necessary to conduct larger well-designed randomised clinical trials (Implications for research).
Radiofrequency ablation versus local ablative interventions
In summary, our meta-analysis adduces moderate-quality evidence compiled from six randomised clinical trials that RFA seems superior to PEI/PAI regarding overall survival, two-year survival, event-free survival, and local recurrence, with a similar frequency of major complications (Lencioni 2003; Lin 2004 (RFA vs. PEI); Lin 2005 (RFA vs. PEI); Shiina 2005; Brunello 2008; Giorgio 2011) ( Summary of findings 2). In the comparison of RFA versus PEI, the low technical and instrumental requirements for PEI should also be taken into consideration (Schoppmeyer 2009). PEI may possibly also be a treatment option for very early hepatocellular carcinoma (single lesion 3 cm or less) (Giorgio 2011). There is evidence from single comparisons suggesting that RFA seems better than PAI and high-dose PEI (Lin 2004 (RFA vs. PEI-hd); Lin 2005 (RFA vs. PAI)). RFA was only compared with laser ablation and microwave coagulation in one randomised clinical trial (Shibata 2002; Ferrari 2007). The trial comparing RFA versus laser ablation found no difference between treatment modalities regarding the outcomes event-free survival and local progression (Ferrari 2007). No survival data were available for microwave coagulation. In conclusion, among the available local ablative therapies, RFA seemed to be the most efficient and safe technique of all the assessed local ablative therapies.
Italian versus East Asian trials
The trials comparing RFA versus PEI/PAI were performed either in Italy (Lencioni 2003; Brunello 2008; Giorgio 2011), or in East-Asia (Lin 2004 (RFA vs. PEI); Lin 2005 (RFA vs. PEI); Shiina 2005). Interestingly, there was no evidence for prolonged overall survival in European patients when treated with RFA compared with PEI (HR 1.24; 95% CI 0.84 to 1.83) ( Summary of findings 3), while East-Asian studies suggested that overall survival was superior after RFA (HR 1.95; 95% CI 1.38 to 2.75). In one post-hoc analysis, we compared both subgroups and found no significant difference between the groups (P value = 0.09). Therefore, undefined technical differences in tumour ablation, reporting bias, or chance might account for the difference between the European and Asian trials. However, this could be addressed in further trials.
In addition to the included trial publications, we found three additional ongoing trials. One randomised clinical trial seeks information on RFA versus resection for hepatocellular carcinoma adjacent to large vessels (NCT00814242). Randomised trials comparing (chemo-)embolisation versus supportive care suggest that (chemo-)embolisation may increase survival (Llovet 2003; Llovet 2003b); however, one Cochrane systematic review found no convincing evidence that transarterial (chemo)embolisation was superior compared with no intervention for unresectable hepatocellular carcinoma (Oliveri 2011). We plan to include the information from these trials in the next update of our meta-analysis.
Overall completeness and applicability of evidence
Efficacy outcomes were properly addressed in most trials. However, this was not the case for the pre-specified outcomes in this review, quality of life and health economics. None of the trials assessed effects of RFA on quality of life and only one trial reported on health economics. Giorgio et al. found RFA to be 100 times more expensive than PEI (Giorgio 2011). The study authors included the cost of the PEI needles and the cost of a radiofrequency generator. However, in contrast to the needles used for PEI, the generator can be used multiple times, and as such, this comparison has to be interpreted very cautiously. For the comparison between RFA and hepatic resection, only retrospective data with inherent high risk of bias are available. One Japanese working group compared medical costs in 213 patients who received either RFA or hepatic resection (Ikeda 2005). Resection yielded lower recurrence rates as compared with RFA while increasing hospital stay. As a consequence, RFA had to be repeated in a higher number of patients. The study authors stated that repeated RFA was cheaper than resection (1,086,000 Japanese yen (JPY) versus 1,745,100 JPY; conversion rate 100 JPY = EUR0.71, at December 2013).
The relevance of other alternative interventional therapies such as laser and microwave ablation remains undetermined. No randomised clinical trial compared RFA versus no treatment, with best supportive care, or with cryoablation. Given the large differences in survival between treatment and no treatment groups in cohort studies, randomised trials with an untreated arm are neither likely to be performed nor ethically justifiable. One unresolved problem of local percutaneous ablation techniques and hepatic resection is the appearance of disease at disseminated areas of the body, and the fact that new tumours often emerge in the remaining cirrhotic liver tissue. Orthotopic liver transplantation is the only treatment that also removes underlying cirrhosis. Randomised clinical trials comparing the effectiveness of RFA versus orthotopic liver transplantation or versus RFA as a bridging therapy to liver transplantation have not been performed.
The current guidelines by the EASL and EORTC state that RFA is the interventional treatment of choice in patients that are not suitable for hepatic resection. If RFA is not feasible, PEI is recommended (EASL-EORTC 2012). In our meta-analysis, we found moderate evidence to support this recommendation, but too few patients were included in order to draw firm conclusion on the pre-specified survival outcomes. To the best of our knowledge, this meta-analysis is the first that includes the trials by Giorgio 2011 and Feng 2012 and thus it provides the best evidence available to date.
Quality of the evidence
The Cochrane Collaboration stresses the importance of good-quality trials. It is recognised to date that methodological quality of trials influences estimates of intervention effects(Schulz 1995; Moher 1998; Kjaergard 2001; Wood 2008; Lundh 2012; Savovic 2012; Savovic 2012a). Considering possible sources of bias, we found five trials with low risk of bias. Six trials had methodological weaknesses that qualified them as trials with high risk of bias but were still eligible for inclusion (Shibata 2002; Chen 2006; Ferrari 2007; Brunello 2008; Giorgio 2011; Feng 2012). In summary, we found moderate-quality evidence derived from six trials with 1088 participants that RFA is superior to PEI regarding efficacy outcomes, while at the same time, the risk of adverse events does not seem to differ between the two treatment modalities ( Summary of findings 2). As for the comparison between RFA and hepatic resection, there is moderate-quality evidence from two low risk of bias trials that randomised 398 patients that hepatic resection is more effective than RFA regarding overall survival and two-year survival. However, if a third trial with an additional 168 patients and a high risk of bias is included, the evidence becomes weak. With regards to the outcomes event-free survival and local progression, hepatic resection was found to yield better results than RFA regardless of the inclusion of the third trial with high risk of bias. In order to strengthen the body of evidence, more trials with low risk of bias are needed. There is high-quality evidence that complication rates were consistently higher in patients treated with hepatic resection ( Summary of findings for the main comparison). Due to their high risk of bias, no firm conclusions regarding the pre-specified outcomes can be drawn for other interventional therapies.
Trial sequential analysis revealed that less than half of the required number of patients was recruited into the eligible trials in order to judge an RRR of 20% (Figure 4; Figure 5). In addition, the calculation of the RR for the comparison RFA versus PEI did not reach statistical significance ( Analysis 1.2). Therefore, we conclude that there is only moderate-quality evidence to support RFA versus PEI or hepatic resection ( Summary of findings for the main comparison; Summary of findings 2).
Potential biases in the review process
We excluded two trials from the analysis as it was unclear whether they were randomised clinical trials. We tried to contact the study authors but did not receive any replies (Gan 2004; Chen 2005). If these trials had been eligible for inclusion, we might have obtained information about the efficacy of RFA compared with chemotherapy and TACE. In addition, the two studies awaiting classification might have added important data to the meta-analysis (Di Costanzo 2011; Kuansheng 2011).
Agreements and disagreements with other studies or reviews
Radiofrequency ablation compared with hepatic resection
Some non-randomised studies (Abu-Hilal 2008; Ueno 2009; Cho 2010), one Markov model analysis (Molinari 2009), and one prior meta-analysis (Zhou 2010) reported superiority of hepatic resection as compared with RFA with regards to survival and recurrence. Although the post-interventional morbidity was less in RFA-treated patients, this did not significantly influence mortality. These results are in agreement with our review results.
Radiofrequency ablation compared with percutaneous ethanol injection or percutaneous acetic acid injection
In addition to the above mentioned meta-analysis (Zhou 2010), three meta-analyses on the effects of RFA on hepatocellular carcinoma have been published (Cho 2009; Bouza 2009; Lau 2009). Cho et al investigated the three-year survival in randomised clinical trials comparing RFA versus PEI for small hepatocellular carcinoma (Cho 2009). The four identified trials were also included in our review. The Lencioni trial included in our analyses was not considered for missing three-year survival rates. In accordance with our analyses, the study authors found evidence that RFA treatment was superior to PEI (OR 0.48; 95% CI 0.34 to 0.67, P value < 0.001). Another meta-analysis identified six studies that compared RFA versus PEI (Bouza 2009). An important difference to our analysis was that the previous authors included a quasi-randomised trial (Livraghi 1999). The RR for one-, two-, three-, and four-year survival, the local recurrence rates as well as the complications were assessed. There was strong (overall survival, one-year disease-free survival, disease free-survival, complete tumour response, major complications) or very strong (one-, two-, three-year survival, total complications) evidence in favour of RFA versus PEI. Lau and Lai referred to four categories in their trial publication: RFA versus other ablative techniques, RFA for unresectable hepatocellular carcinoma, RFA as a bridging therapy for liver transplantation, and RFA for recurrent hepatocellular carcinoma after hepatectomy (Lau 2009). We acknowledge this holistic and thorough approach. However, there are some major differences to our work. First, existing data were included regardless of the quality of trials. For instance, the trial of Lü 2006 was excluded from our analysis since effects of RFA were not extractable (see Characteristics of excluded studies table). Second, quantitative meta-analyses of the data were not performed. Thus, the conclusions presented there should be considered with caution.
Implications for practice
In this update of the previous Cochrane review from 2004 (Galandi 2004), we identified 11 trials comparing radiofrequency ablation (RFA) versus other interventions that fulfilled the inclusion criteria. There is moderate-quality evidence from two trials with low risk of bias that hepatic resection seems to yield better results than RFA in patients with early hepatocellular carcinoma. This is counterbalanced by an increased rate of adverse events and by a longer hospital stay in surgically treated patients requiring a cautious choice of the appropriate therapy for any individual patient. There is moderate-quality evidence that RFA seems to be to be more effective than percutaneous ethanol injection (PEI) or percutaneous acetic acid injection (PAI) in early hepatocellular carcinoma. However, trial sequential analyses showed that only 40% of the required number of patients were recruited in order to judge a relative risk reduction of 20%, therefore, more trials with low risk of bias are needed to confirm our results. A small number of studies have been conducted to compare RFA versus other interventions such as microwave and laser ablation. Due to a high risk of bias in these trials, these interventions cannot yet be recommended for clinical practice.
Implications for research
The is a need for well-designed, sufficiently powered, multicentre trials with low risk of bias comparing RFA versus hepatic resection, PEI, or PAI in patients with hepatocellular carcinoma. Although there is evidence from two low risk of bias trials that hepatic resection yields better survival rates, more data are needed in order to confirm these results. There is a lack of evidence whether RFA can be used as a bridging therapy for patients that are listed for liver transplantation.
In general, trial outcomes should comprise overall survival, event-free survival, adverse events, and quality of life. With regards to increasing financial restrictions in public healthcare systems, the economic costs should also be investigated. In addition, a sufficiently extended follow-up period should be ensured in every trial. To obtain a higher level of external validity, trials involving RFA should be conducted as multicentre trials because surgical as well as percutaneous interventions are strongly affected by the individual experience of the physician(s)/investigator(s).
We thank Dr. Galandi for handing over this meta-analysis to us. We also thank Dimitrinka Nikolova, Sarah Klingenberg, and Christian Gluud from the Cochrane Hepato-Biliary Group for their patience and ongoing support. We are grateful to Dr. Hannelore Tenckhoff from HepNet for database assistance and Ping La from the Children's Hospital of Philadelphia for the translation of the Chinese publications. We also want to express our thanks to Christian Gluud from the Cochrane Hepato-Biliary Group for the help with the trial sequence analysis.
Peer reviewers: Ronald L. Koretz, USA.
Contact editor: Christian Gluud, Denmark.
Data and analyses
- Top of page
- Summary of findings [Explanations]
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. Search strategies
Last assessed as up-to-date: 19 September 2012.
Contributions of authors
Konrad Schoppmeyer and Sebastian Weis searched the literature, extracted the data, and drafted the manuscript.
Annegret Franke extracted the data and performed the statistical evaluation. Annegret Franke also took part in the quality assessment of the trials.
Joachim Mössner critically reviewed the manuscript.
Janus Jakobsen performed the analysis for the summary of finding tables and critically reviewed the manuscript.
Declarations of interest
Sources of support
- None, Not specified.
- None, Not specified.
Differences between protocol and review
- To date the term radiofrequency thermal ablation (RFTA) is not routinely used. It was replaced with the more common term radiofrequency ablation (RFA).
- The authoring team changed.
- Random-effects models were applied.
- CancerLit evolved into MEDLINE in 2003. The database was not searched again. Current Contents is represented by ISI Web of Science and was not searched again.
- Trial sequential analysis (TSA) and report of dichotomous outcomes for the outcome overall survival was included following the requirements of The Cochrane Hepato-Biliary Group.
Medical Subject Headings (MeSH)
Acetic Acid [administration & dosage]; Administration, Cutaneous; Carcinoma, Hepatocellular [mortality; *therapy]; Catheter Ablation [*methods; mortality]; Ethanol [administration & dosage]; Hepatectomy [mortality]; Liver Neoplasms [mortality; *therapy]; Microwaves [therapeutic use]; Randomized Controlled Trials as Topic; Treatment Outcome
MeSH check words