Colorectal cancer (CRC) continues to be an important health concern in North America.^1?3 In both men and women, CRC is the third most commonly diagnosed cancer and the third leading cause of cancer death.¹ Once disease is metastatic, chances of survival reduce.^4,5 Surgical resection is the treatment of choice for CRC liver metastasis (CRCLM); however, only about 20% of the patients are candidates for resection.⁶ Other alternative therapeutic options for metastatic disease to liver include systemic or localized chemotherapy, thermal ablative therapy and transarterial therapies such as trans-arterial chemoembolization (TACE), drug eluting bead TACE (DEB-TACE), trans-arterial yttrium-90 (TAY-90) therapy beads and stereotactic body radiation therapy (SBRT).⁷ Localized thermal ablative therapies include radiofrequency ablation (RFA), cryoablation (CA) or cryosurgery, microwave ablation (MWA) and laser induced thermotherapy.

Although indications for local ablative therapies for CRCLM are more or less restricted to unresectable tumors, as well as resectable tumors in patients with medical co-morbidities or as a local control for widespread disease, there is no agreed consensus as to which local ablative therapy should be used. Patients with CRCLM are often discussed in multidisciplinary hepatobiliary case-conferences: optimal treatment determination is often very challenging. Each institution has its own policy and guidelines, which usually depend on a multidisciplinary team agreement including oncologists, hepatologists and interventional radiologists based on experience and available equipment at that institution. The purpose of this study is to compare the effectiveness and complication rate of various thermal ablative therapies using systematic review, and where applicable, meta-analysis.

The study protocol was not registered. No funding source is declared. We used the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement to guide reporting of this study.⁸

Study characteristics

For the purpose of this systematic review, our patient group-intervention-comparator-outcomes (PICO) question was the following: In patients with unresectable CRCLM, what is the comparative effectiveness and what are complication rates of the various thermal ablative therapies (with or without partial surgical resection)?

 ? Patient group/subjects: Patients with unresectable CRCLM.

 ? Intervention: One specific interventional treatment, which could include any one of the following; RFA, MWA, or CA.

 ? Comparator: One or both of the other 2 ablative modalities that are not the intervention being evaluated.

 ? Outcomes: 1) Efficacy/effectiveness outcomes; 1-, 3-, and 5-year survival rates and local tumor progression rates of hepatic metastatic lesions. 2) Harms outcomes; major complications related to the intervention; complications characterized as per the society of interventional radiology guidelines.⁹

All reported durations and follow-up periods will be considered. Also any study setting and study type will be considered.

Eligibility criteria

Inclusion criteria:

 1. Patients must have had a thermal ablative therapy (RFA, MWA, CA) for treatment of unresectable CRCLM by either percutaneous or intra-operative routes.

 2. Patients who also had associated hepatic surgery (partial resection of metastasis) at the time of the intra-operative therapy were included, though this was not a requirement.

 3. Considers at least one measurable outcome defined in our PICO question.

 4. Data is presented in a manner that allows extraction.

 5. The study is not a qualitative review article.

 6. The number of subjects is greater than or equal to 10.

 7. The study is not a duplicate of data from another study.

Report characteristics

All study designs were considered including comparative experimental (randomized controlled trial, RCT) and observational analytic study designs (cohort, case-control, etc.). We included studies published between 2000 and 2012. It was determined prior to beginning the study, that the thermal ablative procedures are believed to have been well-established techniques in CRCLM treatment during this time frame. We have chosen 2000 as a start date for our systematic review search to minimize the effect of early operator learning curves. Only studies published in the English language were included since the study team did not have the resources to translate articles. Only published studies were evaluated since the study team did not have the resources to search the grey literature.

Where direct comparative studies are lacking, we undertook indirect comparisons between different bodies of evidence evaluating various interventions of interest. If significant heterogeneity were encountered between study groups, study design, and outcomes between the different bodies of evidence, comparisons would be mainly qualitative.

Search strategy

In consultation with the hospital librarian, multiple databases were queried including Embase, ERIC, MEDLINE, Google Scholar, and PubMed. Separate search strategies were conducted for each intervention (RFA, CA, and MWA) and comprised of looking for CRC metastasis, liver, hepatic metastasis and image-guidance modalities (such as computed tomography and sonography) as keywords. These words were combined with the appropriate Boolean operators (AND, OR). Results were combined from the different databases and duplicates were removed. The latest search was performed February 27, 2012. Search results were screened in duplicate with disagreement resolved by consensus to determine potentially eligible studies. These were retrieved and inclusion determined from full text (in duplicate with disagreement resolved by consensus). Only studies meeting all of the inclusion criteria were included. The references of the included studies were hand-searched for any further eligible studies.

Regarding data collection; patient population (including the age, sex, population size, tumor stage, the carcinoembryonic antigen levels, size and number of lesions, previous chemotherapy and equipment used) and outcomes (including 1-, 3-, and 5-year survival rates, local and distant recurrence rates of metastatic lesions, major and minor complications related to the intervention) were collected. The data was extracted in tabular form for each modality utilizing a preliminary Microsoft Office Excel version 2010 (Microsoft, Redmond, WA, USA) sheet and then a standard data extraction Microsoft Office Word version 2010 (Microsoft) form later (Appendix 1).

Risk of bias in individual studies

This was assessed in duplicate using a customized “risk of bias” assessment tool. Given that no standardized tool exists for qualitative data, the basis for this customized risk-of-bias tool was the Cochrane risk-of-bias tool.¹⁰ Modifications were made based on the design of this study. The “risk of bias” tool includes assessment of selection bias, performance bias, detection bias, attrition bias, financial conflict-of-interest bias, and selective outcome reporting bias, where applicable. Each item was categorized into “yes”, “no”, or “unclear”. The overall risk of bias for each included paper was categorized as low, medium and high. Risk of bias appraisal was routinely done during data extraction and added to the same standard Microsoft Office Word form. This was done independently, in duplicate with disagreement resolved by consensus.

The risk of bias across studies could not be assessed formally because there is no agreed upon method with which to reliably detect publication bias in diagnostic accuracy studies as the number of included studies was small in some groups (MWA) and because the studies were heterogeneous in the 3 groups.¹¹ So funnel plots and Egger test could not be performed. Heterogeneity was not quantified by I² statistics since it is expected in systematic reviews of diagnostic test accuracy and since tests for heterogeneity in sensitivity and specificity and estimates of the I² statistics do not account for heterogeneity explained by phenomena such as positivity threshold effects.¹¹ Instead, heterogeneity was assumed, and sources for heterogeneity were explored in details in the discussion.

Data analysis

If data were deemed appropriate for pooling (moderate to low heterogeneity between studies), pooled 1-, 3-, and 5-year survival rates, as well as the minor, major and total complication rates would be calculated by using the random-effects model weighted by the inverse variance.¹² Data that was not amenable to pooling would be presented in a qualitative and tabular manner. All statistical analyses were conducted using statistical software (Comprehensive Meta-Analysis version 2.2; Biostat, Engle-wood, NJ, USA).

Searching the databases resulted in 632 RFA, 309 CA, and 50 MWA records (Fig. 1). Finally 30 RFA, 11 CA, and 5 MWA studies were included in the qualitative synthesis. Fig. 1 is a study flow diagram documenting the search and inclusion process.

Of all the studies included in our systematic review, there was only one RCT in the MWA group¹³ and one case-control prospective study in the CA group.¹⁴ The other included studies were observational cohort studies; either retrospective (35 studies) or prospective (9 studies). A summary of the study design of the included studies is demonstrated in Table 1.

Risk of bias appraisal is summarized in Table 2. Six studies out of the included 46 studies (13.0%) showed overall low risk of bias all of them in the RFA group, while 21 studies (45.7%) showed medium risk of bias and 19 studies (41.3%) showed high risk of bias. None of the included CA or MWA studies showed overall low risk of bias. The percentage of studies that showed high risk of bias in each group was not significantly different in the three ablative modalities, being highest in the CA group (5 studies, 45.5%), followed by the RFA group (12 studies, 40.0%) and the MWA group (2 studies, 40.0%).

Characteristics of the included studies for each modality are summarized in Table 3,^15,43Table 4,^14,26,44,52 and Table 5.^13,30,53,54 Significant heterogeneity encountered on looking at the patient characteristics including the primary tumor stage, number and size of hepatic metastases, extra-hepatic metastasis and the accompanying chemotherapy. Outcomes of the included studies are demonstrated in Table 6,^15?43Table 7,^14,26,44,52Table 8.^13,30,53,54 Not all the targeted outcomes were demonstrated or described in each of the included studies: The available data was collected. A compiled summary of the characteristics and outcomes of different modalities is shown in Table 9. The number of patients included from all the studies (all types of interventions) was 3,107 patients; 2,021 patients in the RFA group (65.0%), 988 in the CA group (31.8%), and 98 in the MWA group (3.2%). The range of median survival was longest in the RFA group (22 40 months) and the MWA group (28 36 months), and shortest in the CA group (13 32 months). The range of post-intervention major complications was lowest in MWA group (7% 19%), followed by RFA group (0%?27%) and highest with the CA group (6%?30%). The range of rate of local tumor progression was widest in RFA group (2%?51%), followed by CA group (6% 44%) and then the MWA group (6%?13%).

Representative forest plots of the 1-, 3-, and 5-years survival rates, major complication rates and local recurrence rates are demonstrated in Fig. 2?6. As expected, not all of the included studies provided data for every outcome. The forest plots confirm the significant heterogeneity of the included studies. Visual assessment of forest plots, as well as qualitative analysis of included papers suggested that between studies heterogeneity was too great and thus, pooling through meta-analysis was not appropriate.

Our observation is that the 1-year survival rates were less heterogeneous in all groups likely due to the relatively short interval. Visual assessment of the 3-year and 5-year survival rates in the RFA group showed significant heterogeneity. Less heterogeneity was encountered in the CA group with the exception of one study⁴⁴ in the 3-year survival rates. Only 1-year and 3-year survival rates were available in the included MWA studies.

Again, significant heterogeneity was clearly observed on looking at the forest plot of major complication rates. In the RFA group, 3 studies showed relatively higher major complication rates as compared to the rest of the included RFA studies.^16,17,28

The forest plot of the local tumor progression rates confirmed the marked variability and heterogeneity of the ranges in each group except in the MWA group. However only 2 studies in the latter group demonstrated local tumor progression results.

This systematic review has identified considerable heterogeneity among the included studies, which rendered meta-analysis implausible. This led us to explore this important heterogeneity and discuss its possible causes and ramifications.

First, the number of included studies for each modality is variable, with the majority of articles included being related to RFA. This difference in the number of included studies is reflected in the ranges of outcomes demonstrated in Table 9. The second factor affecting heterogeneity is the variability in study design, risk of bias, available resources, follow-up imaging schedules and differences in the purposes and conclusions for each included study. This is complicated by the lack of uniformity of reporting the pre and post intervention imaging utilizing standardized terminology as suggested by Goldberg et al.⁵⁵ A third factor affecting heterogeneity is the underlying patient population which varies between studies; some variable factors include patient age, number and size of hepatic lesions, presence of extra-hepatic metastasis, prior hepatic surgery or intervention and significant co-morbidities. Fourth, the differences in type and duration of adjuvant chemotherapy, including use of artery infusion pump therapy in some patients act as confounding factors for the outcome. The types of administered therapies were described in detail in some studies and were not mentioned at all in others. Additionally, the types of chemotherapy or targeted therapy used, has changed significantly over the past 12 years, also possibly affecting patient outcomes. Finally, the method of guidance of each modality and the approach (percutaneous, laparoscopic or intraoperative) can lead to differences in complications rates, primary technical success and long term success; MWA and CA have been shown to adequately ablate larger volumes of tissues while RFA has a more limited size range.^3,56 It is also hypothesized that larger tumors may be associated with worse overall patient outcomes.^7,56

The forest plots demonstrated the significant heterogeneity of the outcomes of each modality between the studies. However, the RFA group showed less heterogeneity in the major complication rates (Fig. 5) with exceptions of three studies.^16,17,28 On further evaluation, it was determined that in these three studies, RFA was combined with partial hepatic resection, which may explain the higher rates of complications. These rates of complication were calculated for the overall therapy including the surgical component which itself has inherent risks of complications.

Less heterogeneity was encountered in the 3-year survival rates in the CA group with the exception of one study.⁴⁴ This study by Bageacu et al⁴⁴ demonstrated a lower 3-year survival rate compared to the other included studies in the same group. However, some patients included in this study had extra-hepatic metastasis. In addition, the size of some of the cryoablated tumors was up to 10 cm, which is relatively large and may preclude primary technical ablative success.

Marked outcome variability and heterogeneity were clearly demonstrated in the forest plot for local tumor progression rates in RFA and CA groups. As discussed earlier, this is likely related to the significant heterogeneity of the patient population, tumor burden and method of guidance. The relatively higher complications rates in the CA group can be attributed to the relatively lager tumor burden in this group. A syndrome of multi-organ failure, severe coagulopathy, and disseminated intravascular coagulation following hepatic CA, has been described and referred to as the cryoshock phenomenon of the liver.⁵⁷ Though it is fairly rare, it is associated with high morbidity and mortality when it occurs. This could have contributed to apparent higher complications rates in the CA arm of this review. Overall, we qualify the interpretation of apparently higher complication rates using CA as guarded due to heterogeneity of the populations studied.

In addition to the differences in the patient populations and their described treatment algorithms, another limitation of this systematic review was the small amount of available literature and relatively small sample sizes; this was most significant for MWA likely because this is a newer therapeutic option and not available at many centers.

We sought to perform a systematic review with meta-analysis when possible specifically for patients with CRCLM undergoing local thermal ablative therapies. Of the published studies on this topic till February 2012, some have systematically reviewed RFA or other ablative therapies for hepatic neoplasms including primary tumors (e.g., hepatocellular carcinoma) in conjunction with metastatic lesions. Pathak et al,⁵⁸ published a systematic review similar in design to ours which focused exclusively on CRCLM except that some of the studies included in their systematic review were excluded from ours as they lacked any measurable outcome related to any of the chosen intervention to specifically treat CRCLM. In addition, Pathak et al⁵⁸ included studies dating back to 1994; at this time, RFA was just beginning to be used for ablations and as it was a new technology, the “learning curve” could have contributed to increased heterogeneity of patient populations. Their conclusion stated that ablative therapies offer significantly improved survival compared with palliative chemotherapy alone with 5-year survival rates of 17% to 24%. Their study also concluded that complication rates among the commonly used ablative therapies are low.⁵⁸ Stang et al⁵⁹ reviewed the clinical benefit of RFA and its role in CRCLM treatment. Similar to our results, they found significant heterogeneity in the available literature. They concluded that RFA prolongs time without toxicity and survival as an adjunct to hepatectomy and/or chemotherapy in well-selected patients, but not as an alternative to resection.

It is clear from the available literature that RFA is the most widely used ablative modality, while MWA is likely the least commonly one used at that time. Despite more ubiquitous use of RFA, the way patient are selected for this therapy as well as chemotherapy use pre, during and post ablation varies significantly from center to center. TACE, DEB-TACE, TAY-90, and SBRT were not compared in this systematic review due to significant differences in patient populations including higher tumor burden and extent of disease, when compared to the 3 local thermal ablation therapies.

Based on this systematic review, more published outcomes of MWA for CRCLM treatment are needed. This modality is promising from the few available results in the literature. This review showed lowest local tumor progression rates in MWA group even though some of the treated tumors were larger than those in the RFA group. This technology is described as being more efficient as compared to standard RFA with consistently higher temperatures measured within the treated tissues, larger tumor ablation volumes and faster ablation times.⁶⁰ In addition, studies have shown that heat sink effects do not affect MWA as much as RFA.³ This may allow successful treatment of patients who may not benefit from RFA due to the location of a liver metastasis next to a prominent vessel.

Future prospective studies may overcome the currently encountered heterogeneity. For example, studies directly comparing the 3 ablative techniques in controlled experiments with comparable patient groups and standardized reporting requirements may be able to more definitively answer questions about efficacy, survival and complications rates.

This systematic review should serve as an important indicator that data regarding the complication rates and relative efficacy of ablative therapies for CRCLM is very heterogeneous. RFA is the most widely used modality to treat CRCLM, yet MWA is a promising technology. More publications about CRCLM treated with MWA are needed. Answering the question; what is the best ablative modality to treat CRCLM necessitates further research comparing techniques with carefully controlled patient populations and standardized outcomes reporting in order to have a better understanding of these potentially beneficial therapies. This may be achievable by creating a multinational working group that could establish parameters in which patients could be enrolled in different arms depending on their concurrent therapies and overall disease burden.

We appreciate the help of Brian Hutton, MSc, PhD, Ottawa Hospital Research Institute (OHRI) and Mohammed T. Ansari, MD, MMedSc, MPhil, Ottawa Hospital Research Institute (OHRI).