Effects of radiofrequency ablation versus other ablating techniques on hepatocellular carcinomas: a systematic review and meta-analysis

Background Percutaneous ablation has quickly arisen as one of the important alternative treatments for hepatocellular carcinoma (HCC). We aimed to compare the therapeutic effects of radiofrequency ablation (RFA) and other ablative techniques on HCCs. Methods Databases were searched to identify literature on complete tumor ablation (CTA), overall survival (OS), local tumor recurrence (LTR), and complications of RFA in the treatment of HCC, compared with those of microwave ablation (MWA), percutaneous ethanol injection (PEI), PEI plus RFA, cryoablation (CRA), laser ablation (LSA), and high-intensity focused ultrasound. Randomized controlled trials and high-quality cohort studies were included in the assessment. Results The effects of MWA and CRA appeared to be similar to those of RFA, but lower rates of LTR and higher rates of CTA in large tumors compared with RFA were reported (P < 0.05). CTA rates were lower in patients treated with PEI (odds ratio [OR] 0.16, 95% confidence interval [CI] 0.06–0.42), and higher in those treated with PEI plus RFA (OR 2.28, 95% CI 1.19–3.60), with an increased incidence of fever (P < 0.05). LSA resulted in lower CTA rates (OR 0.32, 95% CI 0.13–0.81) and OS (hazard ratio 1.47, 95% CI 1.01–2.15), with a lower incidence of complications. Conclusions Compared with RFA, identical effects were found in MWA and CRA groups. Fewer complications were observed in PEI and LSA group. PEI plus RFA appeared more effective, with a higher rate of complications. Well-designed randomized controlled trials are further needed to confirm above results.


Background
Hepatocellular carcinoma (HCC) is the most common primary malignant tumor of the liver, with a poor prognosis. Currently, nearly 78% of cases occur in Asia, and the global incidence of HCC is increasing steadily [1,2]. Although hepatic resection was previously recommended as the firstline choice for radical treatment, only about 15-20% of patients were deemed surgical candidates at the time of HCC diagnosis [3]. The application of liver transplantation is also limited, given the shortage of appropriate donors.
PEI was among the first ablative treatments for HCC; RFA has been recently employed and widely accepted for early-stage and unresectable HCC [3]. International guidelines for HCC management refer to these two methods [8]. However, other available ablative techniques have been reported to achieve effective therapeutic response, with subtle differences among them [6,7]. Thus, clinicians planning HCC treatment by local ablation must select from the various ablative options. The literature comparing the effects of various ablative methods is limited, and the available data should be summarized and clarified.
The aim of the present systematic review and metaanalysis was to compare the outcomes of various ablation methods in HCC treatment, with summarization of the evidence supporting the selection of RFA and other ablative techniques. The rates of complete tumor ablation (CTA), local tumor recurrence (LTR), overall survival (OS), and complications of RFA were assessed relative to those of other percutaneous ablative techniques.

Search strategy and selection criteria
A computerized literature review was conducted to identify articles published between January 1995 and September 2016 using the PubMed, Embase, Cochrane Library, and China Biology Medicine databases. The initial year of 1995 was chosen because the first report on the use of RFA on a patient's liver was published in this year. The search terms used were: "carcinoma, hepatocellular" (MeSH term) OR "hepatocellular carcinoma" (text) AND "radiofrequency ablation" (text) OR "high-intensity focused ultrasound ablation" (MeSH term) OR "high-intensity focused ultrasound" (text) OR "laser coagulation" (MeSH term) OR "laser ablation" (text) OR "microwave ablation" (text) OR "percutaneous ethanol injection" (text) OR "cryosurgery" (MeSH term) OR "cryoablation" (text) OR "ablation techniques" (MeSH term). The reference list of identified publications and review articles were checked manually to identify additional related articles. The last search was performed in September 2015.
The studies meeting all the following criteria were included: (1) full text available in English or Chinese, due to language limit; (2) the results providing data relative to CTA, LTR, OS, or complications, and compared between the outcome of RFA with that of other percutaneously ablative techniques on HCCs, such as MWA, PEI, LSA, CRA, HIFU, or a combination; (3) randomized controlled clinical trials; (4) high quality cohort studies after assessed by Newcastle-Ottawa scale (NOS).
The following forms of publications were excluded: (1) literature in the form of case reports, editorials, reviews, and conference abstracts; (2) the full text not written in English or Chinese; (3) lack of the required postoperative data in the results, and no response after attempts to connect with the author; (4) duplicate data if recent studies were already included; (5) the animal or in vitro research; (6) ablative techniques combined with transhepatic artery chemoembolism; (7) ablation under laparoscope or during surgical operation.
Quality assessment and data extraction NOS and Cochrane collaboration's tool for bias risk assessment were used for quality assessment of cohort studies and RCTs, respectively, by two reviewers (L.LW, Y.M). Different opinions were solved through consultation.
For observational cohort studies, assessment of quality was performed in eight items described in Table 2. Two stars for comparability was acquired and one for the other items if the condition were met. A high-quality study was defined as a study with seven or more stars in total.
For RCTs, low, unclear or high risk of bias in the items including random sequence generation, allocation concealment, blinding of outcome assessment, incomplete outcome data, and selective reporting, was assessed for each study.
Two of the authors (L.W, Zh.YF) independently extracted data, including: (1) country and publication year; (2) number of patients; (3) age; (4) number and size of the tumor; (5) number of patients in CTA; (6) rate of LTR; (7) data about OS or from the OS curve; (8) number of patients with complications. Engauge Digitizer 4.1 was used to extract survival data from OS curves.
CTA was defined as no enhancement in ablated areas on enhanced CT, MRI, or US within 4 weeks after ablation, while LTR was considered as the tumor enhancement near the ablated areas on enhanced CT or MRI during follow-up period. Major complications were mentioned as the events leading to substantial morbidity and disability increase the duration of stay in hospitals or the caring rank, such as subcapsular/intrahepatic hematoma, biliary system injury, hemathorax, liver infarction, liver compensation, symptoms of breath holding and incomplete intestinal obstruction. The other complications were considered minor, such as pain, fever, skin burns, pleural effusion, liver transaminase change, and tumor seeding.

Statistical analysis
The comparison of MWA versus RFA, PEI versus RFA, PEI plus RFA versus RFA, CRA versus RFA, LA versus RFA, and HIFU versus RFA were conducted. Observational studies and RCTs were conducted through meta-analysis, respectively. Reviewer Manager (RevMan; computer program; version 5.2; The Nordic Cochrane Center, The Cochrane Collaboration, 2012, Copenhagen, Denmark) was used for pooling data.
In the meta-analysis, hazard ratio (HR) with 95% CI was calculated for comparison of survival data. Results about CTA, LTR, and complications were compared by calculating odds ratio (OR) with 95% CI. Peto OR was used for the low incidence of events. Heterogeneity was assessed by calculating I 2 . I 2 > 50% was considered as exist of significant heterogeneity. Fixed-effect model was conducted in statistical analyses if there was no obvious heterogeneity; otherwise the random-effect model was used. Inverse variance was used for evaluation of continuous variables, such as the diameter of tumor. Publication bias was evaluated by the funnel plot and Egger's regression. In all analyses, P < 0.05 was regarded as statistically significant.

Characteristics of included studies
The initial database search yielded 2675 publications ( Fig. 1), 2633 of which were excluded upon the screening of titles and abstracts by two reviewers (Zh.XD, H.GB). The full texts of the remaining 42 studies were assessed, and those with duplicate data, lack of assessment of the required outcome, and/or poor quality were excluded. The final sample comprised 30 articles (14 cohort studies, 16 randomized controlled trials [RCTs]). The cohort studies were published between 1999 and 2015, and the RCTs were published between 2002 and 2015. Table 1 shows the main characteristics of included studies .

Quality of included studies
According to the Cochrane Collaboration tool, 13 of the RCTs were affected by performance bias (Fig. 2). Because the ablative techniques require the use of different equipment throughout the duration of therapy, blinding of physicians was difficult. Whether the outcome was assessed blindly was not indicated in any trial, but the assessment of computed tomography (CT) and magnetic resonance imaging (MRI) data was objective. Therefore, no obvious detection bias in outcome evaluation was identified. Factors such as the use of "one-shot" PEI [23], the inclusion of small (diameter < 3 cm) tumors [25], and short follow-up periods [17] resulted in the high risk of other biases.
The 16 cohort studies were given 7-9 stars upon NOS scale evaluation ( Table 2). One of these studies was from Africa, 2 were from Europe, 11 were from Asia, and 2 were from the United States of America.
No significant difference in the CTA rate, OS, 1-or 3year survival rate was detected between the MWA and RFA groups (Fig. 3, Tables 3 and 4). In the cohort studies, LTR was reported in 15.2% (60 of 394) of patients in the RFA group and 13.8% (66 of 479) of patients in the MWA group. No significant difference in LTR was detected between the MWA and RFA groups in the fixedeffect model (Table 3).
CTA rates ranged from 70.1 to 100%, and were lower in the PEI group than in the RFA group (Fig. 3). For 3 of RCTs [24,26,27], a fixed-effect model yielded an OR for lesion-related CTA in the PEI group compared with that in the RFA group of 0.16 (95% CI 0.06-0.42) (Fig. 3c). For 3 of the cohort studies, the fixed-effect model yielded an OR of 0.30 (95% CI 0.14-0.63) for lesionrelated CTA; the fourth cohort study [18] was not included in this model because patient-related CTA rates were reported (Fig. 3d).

PEI plus RFA versus RFA
Two RCTs [27,29] and one cohort study [28] were included in this assessment. All the three studies showed no significant difference in tumor diameter, number,   A study can be awarded a maximum of * for each item A maximum of ** can be given for Comparability liver function, and tumor markers between the PEI plus RFA and RFA groups [27]. In a meta-analysis of data from the two RCTs, the fixed-effect model showed a significantly higher rate of CTA in the PEI plus RFA group than in the RFA group (OR 2.28, 95% CI 1.19-3.60; Fig. 5a). A fixed-effect model yielded an OR of 0.54 (95% CI 0.28-1.03, P = 0.06) for LTR in the PEI plus RFA group compared with the RFA group. OS rates did not differ between groups ( Table 4).
The incidence of fever was higher in the PEI plus RFA group than in the RFA group (Peto OR 2.22, 95% CI 1.20-4.09), whereas the incidence of pain did not differ significantly (OR 1.36, 95% CI 0.74-2.96).  Tumor diameter ≤3 cm was reported as a significant predictor of higher CTA in two studies [27,28]. For tumors with diameters of 3.1-5 cm, Zhang et al. [29] reported a higher OS rate in the PEI plus RFA group (77.6%) than in the RFA group (48.2%) (P < 0.05).

CRA versus RFA
Four cohort studies [30][31][32][33] and one RCT [6] were included in the comparison of the effects of CRA and RFA. In all the studies, it was described that CRA and RFA group did not differ in tumor number and serum laboratory tests.
The cohort studies reported on CTA after CRA and RFA (fixed-effect model: OR 1.13, 95% CI 0.31-4.06; Fig. 5b); CTA rates ranged from 73.3 to 100% in the CRA group and from 82.4 to 100% in the RFA group. The RCT reported CTA rates near 98.3% in the CRA group and 95.6% in the RFA group (P > 0.05).
The incidence of patient-related complications ranged from 3.8 to 40.7% in the CRA group and from 3.3 to 24% in the RFA group [18][19][20]. A random-effect model yielded an OR of 5.02 (95% CI 0.33-77.07) for data on the incidence of complications from two cohort studies [18,20]. In the RCT, 7 of 180 patients in the CRA group and 6 of 180 patients in the RFA group had complications (OR 1.17, 95% CI 0.39-3.56). Various complications were mentioned in these studies; the most common were fever (154 of 195 patients in the CRA group and 145 of 197 patients in the RFA group) [6,31], pain (44 of 180 patients in the CRA group and 98 of 180 patients in the RFA group), and abscess (12 of 234 patients in the CRA group and 4 of 272 in the RFA group) [6,30]. Pleural effusion occurred in 10 patients treated with CRA and 2 patients receiving RFA [16,30,33]. In one study [32], thrombocytopenia occurred in 4 of 25 patients in the CRA group and 1 of 22 patients in the RFA group, and myoglobinemia occurred in 3 of 25 patients following CRA and in no patient after RFA.
A lower LTR rate in the CRA group than in the RFA/MWA group for tumors with diameters >2 cm (P = 0.006) [33]. Although tumors were larger in the CRA group in that study, the rates of complications did not differ significantly.

LSA versus RFA
Three RCTs were included in this assessment [7,34,35]. A fixed-effect model revealed no significant difference in tumor diameter (mean difference 0.11, 95% CI −0.08 to 0.30) between the LSA and RFA groups. And number of single lesion did not differ between two groups (OR 1.38, 95% CI 0.64-2.98) [34,35], while no record was detected in another RCT [7]. Treatment groups were homogeneous with regard to laboratory findings in three RCTs.
CTA rates ranged from 66.7 to 96.2% in the LSA group and from 86.7 to 97.4% in the RFA group (OR 0.32, 95% CI 0.13-0.81; Fig. 5c [7,34,35] MWA microwave ablation, RFA radiofrequency ablation, PEI percutaneous ethanol injection, LSA laser ablation, HR hazard ratio, RCT randomized controlled trial, NA not applicable. Those in square brackets were numbers of references Fig. 4 Forest plot of pooled rates of fever in the microwave ablation group and the radiofrequency ablation group RFA group than in the LSA group (Table 4). Ferrari FS revealed that age, gender, and tumor marker did not affect survival, but in Child-Pugh class A group better survival rates were acquired after RFA than those after LSA. The same did not apply to class B cases [35]. The Peto OR for the incidence of complications was 0.57 (95% CI 0.31-1.07). Ferrari et al. [35] reported no complication or neoplastic seeding. Orlacchio et al. [7] reported minor complications occurring in 2 of 15 patients receiving LSA and 8 of 15 patients receiving RFA, including pleural effusion in 1 LSA case and 4 RFA cases, perihepatic effusion in 1 LSA case and 3 RFA cases, and subcapsular hematoma in 1 RFA case. Di Costanzo et al. [34] recorded moderate pain in 33% of patients receiving LSA and 36% of those receiving RFA, and self-limiting fever in 35% of patients in each group (P > 0.05).

HIFU versus RFA
Only one cohort study referring to HIFU versus RFA was included [36]. No obvious different therapeutic effects were detected between two groups. As it was reported, OS rates were above 60% and CTA were more than 80% in both groups (P > 0.05). Procedure-related complications occurred in the RFA group comparable with HIFU group (P = 0.06) ( Table 1).

Publication bias
Funnel plot and Egger's test were used in the metaanalyses with more than five individual studies pooled in. No obvious asymmetry and P value over 0.05 were detected, which suggested there was no evidence of publication bias (Fig. 6).

Discussion
Although great progress has been made in RFA, and this modality is considered to be important in HCC management, many physicians have used other ablative techniques in attempts to achieve better outcome. Currently, the selection of an ablation technique depends mainly on the doctor's experience and the patient's consent, as detailed indications for each modality have not been defined clearly. In the present study, we compared the therapeutic effects of multiple ablative techniques using data available in the literature to provide insight into modality selection. MWA employs electromagnetic waves from electrodes to induce high temperatures in local areas, coagulating tissue; this principle is similar to that of RFA. In the current study, CTA rates exceeded 80% for MWA and RFA, OS and LTR rates were similar, and complication rates were similarly low. Only the incidence of fever was higher in the MWA group than in the RFA group, which may indicate that MWA was more invasive. In subgroups defined according to tumor size, no difference was detected between groups, but the results did indicate a possible trend toward an advantage of MWA for larger tumors.
As a non-thermal ablation, the use of PEI for the treatment of HCC was popular before the widespread use of RFA. In this study, we observed lower CTA rates and higher LTR rates in the PEI groups than in the RFA groups. The intratumor fibrous septum may interfere with the injection of ethanol [37]. Thus, the better outcome of RFA may be related to the greater predictability of ablation than in PEI. On the other side, few minor complications were described in PEI cases, whereas serious events, including skin burns, hepatic infarction, and hemothorax, occurred after RFA. The combined methods showed enhanced effects. A trend suggests that PEI-RFA may be utilized in 3.1-5-cm tumors. Kazutaka et al. [38] reported larger areas of coagulated necrosis in the PEI-RFA group (34 ± 29.3 cm 3 ) than in the RFA group (6.5 ± 3.6 cm 3 ; P < 0.0001). The amount of ethanol injected into the tumor was significantly and positively correlated with the volume of coagulated necrosis, but not with the energy requirement. Therefore, in the application of PEI-RFA in high-risk areas, such as near vessels or other important organs, ethanol injection helps to decrease the RFA energy, thereby protecting the surrounding tissues. However, in this study, a higher incidence of fever was found after PEI-RFA than after RFA.
CRA caused the tumor cell death through ice crystal formation during rapid freezing. We observed no obvious difference between CRA and RFA, while in two studies the tumor size in CRA was noted to be obviously larger than that in the RFA group [32,33]. On the other hand, in a Japanese study [33], the tumor in CRA patients was located in close proximity to the hollow viscera such as the gallbladder, or important structure such as the hepatic hilum, and such location was not mentioned in the thermal ablation group. In addition, in the same study, lower LTR in the CRA group was reported than that in the RFA/ MWA group for tumors with diameter greater than 2 cm. These outcome data may provide some information supporting that cyroablation may be more suited for larger HCCs or those in high risk areas, but more evidence is needed to make definitive conclusions. For complications, myoglobulinemia was not observed following RFA, but did occur in 3 of 33 procedures in CRA [32]. Thrombocytopenia and myoglobulinemia after CRA occurred in more procedures than that after RFA, but more data are needed for further statistical analysis [32,33].
In the current study, it was shown that lower CTA rates and higher LTR rates were observed in the LSA patients, whereas higher OS rates in RFA patients were seen, particularly among larger HCCs (P < 0.05). The tendency of fewer complications was indeed detected in the LSA group. Thin needle for LSA may improve the ablative effects on the nodules with irregular shape or in high-risk location. In addition, in an Italian study, a lower cost of LSA was mentioned but had no statistical analysis [34].
There were some limitations in this study. Many studies have demonstrated that tumor size, number of single lesion, and Child-Pugh class were important prognosis factors [28,35]. But in the above literature, data on these factors were not enough for meta- Fig. 6 Funnel plot of pooled data. a completed tumor ablation between the microwave ablation (MWA) group and the radiofrequency ablation (RFA) group. b local tumor recurrence between the MWA group and the RFA group. c overall survival between the percutaneous ethanol injection group and the RFA group analysis in subgroup. Another limitation was that the device of techniques may have influence on its overall therapeutic effects. In the current study, we did not perform an analysis about the device, because the device varied greatly among different therapy centers. Finally, the number of papers included for CRA, LSA, and HIFU was small, especially in HIFU the meta analysis were not conducted and more evidence was needed in the future.

Conclusions
Above all, the availability of multiple ablative techniques is the reason for optimism. It was pleasing that there have been many kinds of ablative techniques for treating HCCs. The combination of PEI and RFA apparently yielded a better prognosis than a single RFA. When the outcome of MWA appeared identical to that of RFA, a potential role of CRA used in larger tumors needs to be investigated. Less complications in PEI and LSA implied their application in high-risk areas for protecting the important organs. Indeed, well-designed randomized controlled trials are further needed to confirm the above results. With the different characteristics, multiple ablative techniques may be combined in one treatment procedure, so as to achieve better effects and avoid adverse events.