Surgical resection (SRS) is generally regarded as the gold standard therapy for patients with early stage hepatocellular carcinoma (HCC) (Reference Lau and Lai16;Reference Marlow19;Reference Sutherland, Williams, Padbury, Gotley, Stokes and Maddern22). For patients who are not suitable candidates for curative surgical resection because of the number or anatomic distribution of tumors, the presence of extrahepatic disease, or poor liver function, percutaneous radiofrequency ablation (PRFA) is the preferred treatment (Reference Lau and Lai16;Reference Lencioni and Crocetti17;Reference Marlow19;Reference Sutherland, Williams, Padbury, Gotley, Stokes and Maddern22). However, recent studies have reported comparable clinical outcomes with either PRFA or SRS for early stage, potentially resectable, HCC (Reference Garrean, Hering, Saied, Helton and Espat9).
Radiofrequency ablation (RFA) was introduced in the early 1990s (Reference Hasegawa, Makuuchi and Takayama12). It can be administered at the time of open surgery, by means of laparoscopy or percutaneously, but the latter is the most common approach (Reference Amersi, Elrath-Garza and Ahmad1). Under computed tomography (CT) or ultrasound guidance, the radiologist inserts a needle through the skin into the tumor. Electrodes at the tip of the needle deliver energy created by radiofrequency waves, and the energy induces thermal injury in proportion to the temperature achieved and the duration of heating (Reference Lencioni and Crocetti17). To destroy tumor tissue, the temperature is typically maintained at 50° to 100°C for 4 to 6 minutes, according to the size of the tumor (Reference Lencioni and Crocetti17). The destroyed tumor tissue shrinks and slowly forms a scar. Traditionally, PRFA has been used for tumors of less than 3 cm (Reference Cormier, Thomas, Chari and Pinson8), but some studies now describe its use in tumors of 3–5 cm (Reference Buell, Thomas and Rudich3;Reference Cormier, Thomas, Chari and Pinson8;Reference Khan, Poon and Ng14;Reference Zhai, Xu, Chen, Li, Liu and Wu24), and even larger (Reference Zhai, Xu, Chen, Li, Liu and Wu24). An indication for PRFA is poor hepatic function, while contraindications include tumor proximity to major vascular or biliary structures, and the presence of more than three tumors (Reference Kooby, Egnatashvili and Graiser15).
The primary objective of this study is to compare the clinical effectiveness of PRFA with surgical resection (SRS), based on a systematic review of the literature. In addition we compared the costs of these two therapeutic approaches from the point of view of the third party (the Canadian healthcare system). This article is based on a recent Health Technology Assessment report prepared by the Technology Assessment Unit of the McGill University Health Centre (Available at www.mcgill.ca/tau/).
METHODS
Literature Search
We first searched the databases maintained by the International Network of Agencies for Health Technology Assessment (INAHTA), Cochrane Collaboration and the Centre for Reviews and Dissemination (CRD), PubMed, MEDLINE, and EMBASE databases for relevant health technology assessment reports or systematic reviews. We used the key words (radiofrequency OR radio frequency OR radio-frequency OR catheter ablation) AND (liver carcinoma OR liver cancer OR hepatocellular carcinoma OR liver cell carcinoma). To identify relevant individual articles in English, French, and Chinese we searched the PubMed, MEDLINE, and EMBASE databases using the same keywords as above. Publications in other languages were considered, if their English abstracts/Tables/Figures were understandable and added information to the existing evidence base. A further search was conducted by tracking references in publications identified.
The most recent HTA, published by the Australian Safety and Efficacy Register of New Interventional Procedures-Surgical (ASERNIP-S), used a comprehensive search strategy ending in April 2006 (Reference Marlow19). Therefore, we commenced our literature search from January 2006. The date of the last literature search was January 15, 2009. To identify articles published before 2006, we relied on this and other HTAs (Reference Bouza, Martin and Magro2;Reference Marlow19;20), systematic reviews (Reference Garrean, Hering, Saied, Helton and Espat9;Reference Lau and Lai16), and the bibliography of articles that met our inclusion criteria.
Inclusion Criteria
We only included articles that had a mean or median follow-up time of at least 1 year, reported survival rates, included patients with early stage HCC (Child-Pugh score of A or B and tumor size <5 cm) and studies that were reported in peer-reviewed journals. Both randomized controlled trials (RCTs), and nonrandomized comparative cohort studies were considered for this review. If multiple publications were based on the same data, only one of them was included. We excluded cohort studies when the average sample size per treatment group was less than fifty.
Data Extraction
We extracted specific clinical outcomes, including survival rates and disease-free survival rates, typically at 1 year, 3 years, and 5 years, local and distant recurrence rates and major complications. If articles did not report cumulative survival rates in the text or tables, but presented them in figures, such as in a Kaplan-Meier survival curve, we estimated cumulative survival rates from the curve. When necessary, we contacted authors to clarify issues. The first author conducted the preliminary data extraction. Data extraction and results were cross-checked by the second author. All selected articles were reviewed by all three authors.
Meta-Analyses
We divided studies into subgroups based on the Child-Pugh class of the patients. The Child-Pugh score is used to measure the prognosis of patients with chronic liver disease, with scores from A to C indicating worsening expected survival (Reference Pugh, Murray-Lyon, Dawson, Pietroni and Williams21). We carried out separate meta-analyses to estimate the pooled 1-, 3-, and 5-year survival rates of each treatment strategy within two subgroups of studies—those with Child-Pugh A patients alone and those with a mix of Child-Pugh A and B patients. We used an exact likelihood approach based on the binominal distribution (Reference Hamza, van Houwelingen and Stijnen11). Meta-analyses were conducted using SAS 9.1.
Cost-Analysis
Our primary interest was to estimate the difference in cost of the two treatments, including healthcare costs and physician's fees, seen from the perspective of the third party. We obtained information on costs and resource use from the Department of Finance and relevant service departments of the McGill University Health Centre (MUHC). Because patients undergoing PRFA and SRS are managed identically during pretreatment and follow-up, the costs incurred during these phases of management are ignored. For simplicity, we also excluded the cost of complications from our analysis. All cost information is per patient and not per tumor. We carried out a Monte Carlo analysis to estimate a 95 percent confidence interval for our best estimate of the difference in cost between the two procedures. The variation in costs of each item required for surgical resection was determined based on fifty-seven procedures carried out at the MUHC during the fiscal year 2008–2009. Costs of items required for PRFA were assumed to have a variation of ±25 percent around the best estimate elicited from the physician. We also conducted one-way sensitivity analyses to identify key variables impacting the incremental cost. All costs are expressed in Canadian Dollars ($).
RESULTS
We identified three health technology assessment (HTA) reports in the INAHTA database, two of them published in 2004 (Reference Bouza, Martin and Magro2;20), and the third in 2006 (Reference Marlow19). We identified three systematic reviews published after 2006, two published in 2008 and the other one in 2009, which included comparisons of RFA and SRS. The most recent systematic review by Lau and Lai (Reference Lau and Lai16) concluded that “level II and level III-2 evidence showed the survival rates after RFA was comparable to surgical resection for tumor less than or equal to 3 cm. There was uncertain evidence for tumor sized 3.1 to 5 cm.,” and that in the absence of evidence on RFA for resectable HCC size >5 cm, surgical resection remains the primary treatment. Chen et al. concluded that there was no difference in survival between PRFA and SRS up to 4 years, but disease-free survival was significantly worse in the PRFA group in the fourth year (Reference Chen, Xiao and Zhou6). Garrean et al. concluded that level I and level II evidence show that survival rates by RFA are comparable to those of SRS, but local recurrence rates by RFA appear higher than SRS (Reference Garrean, Hering, Saied, Helton and Espat9).
We identified 838 abstracts published between January 2006 and January 2009. We found several instances where multiple publications were based on the same sample. In total, we identified six unique studies meeting our inclusion criteria consisting of one RCT (Reference Chen, Li and Zheng5), and five comparative cohort studies (Reference Cho, Tak and Kweon7;Reference Guglielmi, Ruzzenente and Valdegamberi10;Reference Hong, Lee and Choi13;Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18;Reference Takahashi, Kudo and Chung23). In addition to the references identified in the 2009 review of Lau et al., we identified two observational studies, Cho et al. (Reference Cho, Tak and Kweon7) and Takahashi et al. (Reference Takahashi, Kudo and Chung23), and one large registry-based study (Reference Hasegawa, Makuuchi and Takayama12) of PRFA versus resection in early stage hepatocellular cancer which was not included in the analyses as it had a median time of follow-up of only 10.4 months.
Details of demographic and clinical variables at baseline in each of the included studies are listed in Table 1. The three observational studies by Takahashi et al., Cho et al., and Hong et al., included only Child-Pugh Class A patients, and the average tumor size in these studies was relatively small, around 3 cm or less. However, the two observational studies by Guglielmi et al. and Lupo et al. included both Child-Pugh Class A and B patients. The median tumor size in these two studies was greater than 3 cm and the study by Lupo et al. included only patients with tumor sizes of 3–5 cm. In the five studies that reported this information, the number of tumors per patient was 1.
Table 1. Baseline Characteristics of Patients in Studies of PRFA vs. SRS
Note. ¶Median; §Mean; #Median/mean (range); &Mean ± standard deviation. †Intention to treat analysis was conducted. However, for PFRA group, descriptive demographic statistics in Table 1 were based on 71 patients who received PRFA treatment, and excluded 19 patients who withdrew their consent.
RCT, randomized controlled trial; RFA, radiofrequency ablation; PRFA, percutaneous RFA; SRS, surgical resection/partial hepatectomy/ hepatectomy/ resection; HCC, hepatocellular carcinoma tumors; Class, Child-Pugh class; TACE, transarterial chemoembolization
N, total sample size.
Survival
The RCT by Chen et al., included patients with tumors of size ≤5 cm (Reference Chen, Li and Zheng5). Although nineteen of ninety patients withdrew after randomization to the PRFA group and did not receive PRFA, the results of intention-to-treat (ITT) and efficacy analyses (EA) were very similar, and both demonstrated that PRFA is as good as the gold standard therapy, SRS, in terms of survival rate and disease-free survival rate. The 1-, 2-, 3-, and 4-year overall survival rates in the PRFA group (94 percent, 80 percent, 69 percent, 66 percent) were almost identical to those in the SRS group (93 percent, 82 percent, 73 percent, 64 percent) (Table 2; a more detailed version of Table 2 appears in Supplementary Table 1, which can be viewed online at www.journals.cambridge.org/thc2010026).
Table 2. Clinical Outcomes Reported by Studies Comparing SRS to PRFA
Note. #Originally, this study was published in Chinese in 2005 (Reference Chen, Li and Liang4). Thereafter, authors added additional patients and republished, including all patients in 2006 (Reference Chen, Li and Zheng5). However, the 2006 article did not report recurrence rates, so we derive these from the 2005 study (Reference Chen, Li and Liang4). *Estimated value, extracted from the graphs of the articles. ‡In the PRFA group, 89 out of 109 with complete necrosis were included in disease-free survival analysis.
RCTs, randomized controlled trials; NA, not applicable; PRFA, percutaneous radiofrequency ablation; SRS, surgical resection; DFS, disease-free survival; T, T1, T2, T3, time point (in years) when clinical outcomes were reported; N, total sample size; n, number with outcome; y, year/years; mortality, procedure-related mortality or hospital mortality.
Some observational studies attempted to adjust for baseline differences in PRFA and SRS groups either by study design or statistical analysis, but it is very likely that selection bias still existed (Reference Guglielmi, Ruzzenente and Valdegamberi10;Reference Takahashi, Kudo and Chung23). Nonetheless, only one study found a clinically important and statistically significantly higher survival rate following SRS (Reference Guglielmi, Ruzzenente and Valdegamberi10). This study had an obvious selection bias as at this center Child-Pugh Class A and single HCC cases were typically considered for surgery, whereas Class B and multiple HCC cases were typically considered for PFRA.
Based on the observational studies with Child-Pugh A patients, the pooled survival rates with (95 percent confidence intervals) in the PRFA and SRS groups, respectively, were 98 percent (96 percent,100 percent), and 98 percent (96 percent, 100 percent) at 1 year, and 83 percent (74 percent, 91 percent) and 82 percent (76 percent, 87 percent) at 3 years (Reference Cho, Tak and Kweon7;Reference Hong, Lee and Choi13;Reference Takahashi, Kudo and Chung23). One study that estimated survival at 5-years reported estimates of 77 percent (70 percent, 83 percent) (Reference Takahashi, Kudo and Chung23) in the PRFA group versus 71 percent (58 percent, 83 percent) in the SRS group (Reference Takahashi, Kudo and Chung23).
In the observational studies with a mix of Child-Pugh A and B, the pooled 3-year and 5-year survival rates were higher in SRS groups (Reference Guglielmi, Ruzzenente and Valdegamberi10;Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18). The pooled survival rates with (95 percent confidence intervals) in the PRFA and SRS groups, respectively, were 90 percent (81 percent,100 percent), and 88 percent (82 percent, 93 percent) at 1 year, 43 percent (35 percent, 50 percent) and 62 percent (53 percent, 70 percent) at 3 years, and 25 percent (16 percent, 33 percent) and 48 percent (40 percent, 57 percent), at 5 years of follow-up (Reference Guglielmi, Ruzzenente and Valdegamberi10;Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18).
Recurrence
In the RCT, recurrence rates following both interventions were also comparable, with 3-year cumulative recurrence rates of 23 percent and 25 percent for PRFA and SRS, respectively (Reference Chen, Li and Zheng5). In the three cohort studies that reported recurrence rates, there was a higher rate following PRFA (see Table 2) (Reference Cho, Tak and Kweon7;Reference Hong, Lee and Choi13;Reference Takahashi, Kudo and Chung23). All three studies included only Child-Pugh A patients.
Disease-Free Survival
In the RCT, disease-free survival (DFS) rates at 4-year follow-up were 48 percent and 52 percent following PRFA and SRS, respectively (p > .05) (Reference Chen, Li and Zheng5). In all of the five observational studies that reported DFS, this was longer following SRS (see Table 2), but only in Guglielmi et al. (Reference Guglielmi, Ruzzenente and Valdegamberi10) was this statistically and clinically significant. (As noted above, there was obvious selection bias in this study [Reference Guglielmi, Ruzzenente and Valdegamberi10].) In Takahashi et al. (Reference Takahashi, Kudo and Chung23), the median DFS was 23 and 25 months following PRFA and SRS, respectively (p = .075).
The pooled 1- and 3- year disease-free survival rates in the two observational studies of Child-Pugh A patients for PRFA and SRS groups, respectively, were 72 percent (95 percent CI: 64 percent, 79 percent) and 73 percent (95 percent CI: 66 percent, 80 percent); and 37 percent (95 percent CI: 30 percent, 45 percent) and 47 percent (95 percent CI: 34 percent, 59 percent), respectively (Reference Cho, Tak and Kweon7;Reference Hong, Lee and Choi13).
The pooled 1-, 3-, and 5-year disease-free survival rates in the two observational studies, which included both Child-Pugh A and B patients, survival in PRFA and SRS groups were 65 percent (95 percent CI: 57 percent, 72 percent) and 80 percent (95 percent CI: 73 percent, 87 percent); 47 percent (95 percent CI: 32 percent, 62 percent) and 73 percent (95 percent CI: 66 percent, 80 percent); and 2 percent (95 percent CI: 0 percent, 13 percent),and 22 percent (95 percent CI: 12 percent, 32 percent), respectively (Reference Guglielmi, Ruzzenente and Valdegamberi10;Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18).
Additional Treatment
Patients in the PRFA group are more likely to have multiple treatments to achieve complete necrosis. In the RCT by Chen et al., twenty-three (32 percent) of patients in the PRFA group needed additional treatment with PRFA, percutaneous ethanol injection (PEI) or transarterial chemoembolization (TACE) because the 1-month follow-up CT scan showed incomplete tumor necrosis. The authors believe that without the additional treatment the subsequent intrahepatic recurrence rate would have been very high. In the study by Guglielmi et al., patients treated with PRFA received between 1 and 4 treatment sessions, and those with smaller tumors (<3 cm) were more likely to achieve complete necrosis than those with larger tumors (93.3 percent versus 80.3 percent). The repeat procedure rate at the MUHC is roughly 8 percent.
Complications
The most common major complications following PRFA are hemorrhage, skin burn (under the electrode), pleural effusion, and hepatic abscess. Whereas skin burns are only associated with use of PRFA, the other complications also occur following surgical resection. In the RCT by Chen et al. (Reference Chen, Li and Zheng5), major complications (liver failure, gastrointestinal bleeding, ascites, persistent jaundice) were significantly more frequent in the SRS group (50 of 90 procedures [55 percent] versus 3 of 71 [4 percent]), and one patient died following SRS treatment. In the three cohort studies that reported complications the rates for PRFA and SRS were 10 percent and 36 percent (Reference Guglielmi, Ruzzenente and Valdegamberi10), 10 percent and 17 percent (Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18), and 6.5 percent and 5 percent (Reference Cho, Tak and Kweon7), respectively (see Table 2). One hospital death was observed following SRS (Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18).
Hospitalization
In the two studies that recorded hospitalization, SRS was associated with significantly longer hospital stay (Reference Chen, Li and Zheng5;Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18). In Chen et al. (Reference Chen, Li and Zheng5), the hospital stays (mean ± SD) in SRS and PRFA groups were 19.7 ± 5.6 and 9.2 ± 3.1 days, respectively; and in Lupo et al. (Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18), the median (range) hospital stays in SRS and PRFA groups were 5.5 (3–43) and 2 (1–16) days, respectively.
Patients' Attitudes Toward Treatment Options
There are surprisingly large differences in different studies in patients treatment preferences. In the RCT, nineteen of ninety patients withdrew after randomization into the PRFA group, and all received surgical resection (two of ninety patients in the surgical resection group who were found to have a disseminated tumor were treated with ethanol injection) (Reference Chen, Li and Zheng5). In the study of Hong et al. (Reference Hong, Lee and Choi13), because of refusal of surgery or general anesthesia, thirty-three patients underwent PRFA rather than SRS. (In total, fifty-five patients received PRFA treatments in this study.) Both Lupo et al. (Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18) and Guglielmi et al. (Reference Guglielmi, Ruzzenente and Valdegamberi10) mentioned that one reason for choosing PRFA treatment was refusal of surgery, but did not provide further information on how many patients refused surgery and then received PRFA.
Cost Analysis
For the purposes of cost analysis we first assumed that recurrence rates are comparable following either treatment (in accordance with the findings of the RCT). Component and total costs for the two treatments are listed in Table 3. When both options are possible, performing PRFA would cost $8,275 (Canadian dollars) less per case (95 percent confidence interval based on a Monte Carlo analysis: $8,079–$8,317). Results of univariate sensitivity analyses are shown in Table 4. The most important variables impacting the difference in cost between the two treatments are the cost of nursing and intensive care, the cost of operating room use for SRS and the recurrence rate for PRFA.
Table 3. Cost Analysis
Note. †The above costs are those pertaining at one Canadian hospital, the McGill University Health Centre, in October 2009. SRS cost estimates were based on 57 procedures, 5th to 95th percentile of each item used for sensitivity analysis; PRFA, cost estimates made in consultation with the Department of Radiology of the Royal Victoria Hospital division.
ICU, intensive care unit; PRFA, percutaneous radiofrequency ablation; $, Canadian dollars (CAD).
Table 4. Univariate Sensitivity Analysis
$, Canadian dollars (CAD).
DISCUSSION
In the evaluation of PRFA and SRS, we have had to largely rely on noncontrolled studies, and, although, to facilitate analysis, we have grouped patients into specific treatment groups, it is difficult, for many reasons, to be certain of the real role of any one therapy in the management of these liver cancers. For example, many of the observational studies did not report the criteria by which treatment was chosen (Reference Cho, Tak and Kweon7;Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18;Reference Takahashi, Kudo and Chung23). All were subject to selection bias and this was suggested in some (Reference Guglielmi, Ruzzenente and Valdegamberi10;Reference Hong, Lee and Choi13). Furthermore, irrespective of the initial therapy, patients are likely to experience many other forms of treatment as their disease progresses. Therefore, even though patients with similar characteristics can be grouped together, selection bias still cannot be excluded. By separating studies according to Child-Pugh score we found that SRS and PRFA have comparable survival rates in studies of Child-Pugh A patients where the average tumor size was 3cm or less (Reference Cho, Tak and Kweon7;Reference Hong, Lee and Choi13;Reference Takahashi, Kudo and Chung23). On the other hand, SRS showed better survival than PRFA in two Italian studies with a mix of Child-Pugh A and B patients with a median tumor size greater than 3 cm (Reference Guglielmi, Ruzzenente and Valdegamberi10;Reference Lupo, Panzera, Giannelli, Memeo, Gentile and Memeo18). However, it should be noted that all observational studies were probably affected by selection bias with patients with worse prognosis being typical candidates for PRFA.
A further study, not included in the meta-analysis because of a median follow up time of only, 10.4 months, was a retrospective review of a registry of the results of a Japanese nationwide survey in which 7,185 patients were identified with no more than three tumors (<3 cm) and a liver function of Child-Pugh class A or B (Reference Hasegawa, Makuuchi and Takayama12). After a median 10.4 months follow-up, the 2-year time-to-recurrence rates for SRS (2,857) and RFA (3,022) were 35.5 percent and 55.4 percent, respectively (p < .0001). Overall survival rates were not different.
In conclusion, there is good evidence that where both SRS and PRFA are available options, survival rates following either procedure are comparable for Child-Pugh A patients, while complications are more frequent and hospitalization longer following SRS. The evidence concerning recurrence rates and disease-free survival is less clear.
A limitation of our HTA is that it includes only one RCT that had a relatively small median follow-up time of 29 months. Despite its drawbacks, this RCT indicates comparable outcomes with either procedure. However, the results of several of the cohort studies (with possible selection bias) favor the surgical option in terms of disease-free survival and recurrences. The surgical option costs approximately $8,275 more per case than PRFA, although this could fall if the recurrence rate following PRFA is high.
Continuing doubts on this issue can only be resolved by a substantial RCT. Meanwhile, at the time of choosing which of these two options to use, for patients with good liver function (Child-Pugh A) there is now good consistent evidence of equal survival up to 3 years with either. In the presence of poor liver function (Child-Pugh B) this evidence is less robust. There is suggestive evidence that is less reliable because of possible selection bias, that recurrence is more likely following RFA. Under these circumstances patients' wishes can justifiably determine which intervention is used, some preferring the lower chance of undergoing a second procedure, others preferring the less invasive intervention with almost equivalent outcome, lower complication rates, shorter hospital stay, and lower costs.
CONTACT INFORMATION
Xuanqian Xie, MSc (shawn.xie@muhc.mcgill.ca), Biostatician, Technology Assessment Unit, Royal Victoria Hospital, McGill University Health Center, 687 Pine Avenue West, R4.14, Montreal, Quebec, Canada H3A 1A1
Nadini Dendukuri, PhD (nandini.dendukuru@mcgill.ca), Assistant Professor, Epidemiology, Biostatistics, and Occupational Health, McGill University, 1020 Pine Avenue West, Montreal, Quebec, Canada H3A 1A2; Director, Technology Assessment Unit, McGill University Health Center, 687 Pine Avenue West, R4.09, Montreal, Quebec, Canada H3A 1A1
Maurice McGregor, MD (maurice.mcgregor@mcgill.ca), Professor Emeritus, Department of Medicine, McGill University, 3655 Prom. Sir Wilfred Osler, Montreal, H3G1Y6; Chair, Technology Assessment Unit, Department of Medicine, McGill University Health Centre, 687 Pine Avenue West, Montreal, Quebec, Canada H3A 1A1
CONFLICT OF INTERESTS
All authors report having no potential conflicts of interest.
