Key Findings and Strength of Evidence

This review addressed the comparative effectiveness of local hepatic therapy for the treatment of unresectable HCC in patients who are not otherwise eligible for transplantation and do not have extrahepatic spread. Forty-eight studies met our inclusion criteria and included six RCTs, four nonrandomized comparative studies, 35 observational case series, and three case reports.

We assessed the strength of evidence for our primary health outcomes of overall survival and quality of life; the intermediate outcomes of TTP, local recurrence, LOS, and days of work missed for KQ1; and adverse events for KQ2 (Table 91). In addition, we reviewed the effect of patient subgroups on the comparative effectiveness of the included comparisons for our population of interest for KQ3.

Table 91

Summary GRADE strength of evidence for KQ1 and KQ2.

For the comparison of RFA to PEI/PAI, three RCTs^50-52 were pooled in a meta-analysis (Figure 3), and risk differences were calculated. The pooled estimate was 0.16 (95 percent confidence interval [CI], 0.03 to 0.28), a statistically significant result that favored RFA. The wide range of effect across the three trials and a moderate level of statistical heterogeneity in this pool of studies (I²=48 percent) led to the classification of the results as inconsistent. We judged the strength of the body of evidence on overall survival in favor of RFA compared with PEI/PAI as moderate. The strength of the body of evidence was downgraded from high, the starting point when multiple RCTs are available, to moderate for the lack of consistency in the results across studies. In addition to overall survival, two RCTs^51,52 reported on the outcomes of TTP, local control, and LOS. Due to the lack of blinding, the risk of bias was high, the results were consistent and precise, and all three are indirect measures of a final health outcome. Based on the high risk of bias and indirect measurement, we judged the strength of evidence on TTP and local control in favor of RFA compared with PEI/PAI to be low. Also based on the high risk of bias due to a lack of blinding, the strength of evidence is graded low for a longer LOS following treatment with RFA compared with PEI/PAI. All three RCTs^50-52 performed subgroup analyses to determine if overall survival was superior among specific patient subgroups. There is a low strength of evidence to support increased overall survival for RFA compared with PEI/PAI in patients with larger lesions (defined variably as >2cm, 2-3cm, and 3.1-4cm) with a high risk of bias. The evidence is insufficient to assess the effects lesion size on other outcomes of interest in this report and of other patient subgroups on any outcome of interest in this report.

We judged the strength of evidence to be insufficient to draw conclusions for effectiveness outcomes (overall survival, quality of life, disease progression, local recurrence, LOS, and days of work missed) and for adverse events for patients considered for all other comparisons.

Data were judged to be insufficient due to high risk of bias, imprecision of estimates, and lack of comparative data for some outcomes (e.g., quality of life, days of work missed).

Evaluation of comparative effectiveness requires an intervention and a comparator. Case-series do not use comparators. Therefore, comparative effectiveness cannot be assessed using this type of literature. Further, factors that may affect the effectiveness of the interventions within these populations were not controlled for in the included studies. Control may be achieved either though randomized design or statistically though careful adjustment in the analysis. Studies that aim to determine the effectiveness or comparative effectiveness of local treatment for unresectable HCC should use randomized designs. If randomization is not possible, care should be taken to control for covariates such as size and number of hepatic lesions and performance status through regression analysis.

Findings in Relationship to What Is Already Known

There is a large range of unique comparisons of various local hepatic therapies for HCC. We are not aware of any systematic review that has examined all comparisons. We identified seven previously published comparative systematic reviews, each examining a single comparison of local hepatic therapies. Two systematic reviews compared RFA to PEI,^98,99 three compared TACE-percutaneous ablation (PA; either RFA or PEI) to RFA or TACE monotherapy,^100-102 and one compared PEI to PAI.¹⁰³

Consistent with our findings, the three systematic reviews ^98,99,104comparing the ablative therapies RFA and PEI found that RFA demonstrated a significantly better overall survival rate than PEI. These reviews included the three RCTs that met the inclusion criteria for our evidence review, in addition to one or more trials that were not included in this review due to differences in inclusion criteria. The review by Bouza et al.⁹⁸ included three additional trials in which the study intervention was given prior to the year 2000 or the patient sample included those who refused surgical treatment of HCC, both of which are included in our exclusion criteria. The reviews by Cho et al.⁹⁹ and Salhab et al.¹⁰⁴ included patients refusing surgery in one and two trials, respectively. The pooled patient population in these two systematic reviews was similar to the population for this comparison in our review, that is, early stage HCC patients with up to three nodules less than 3 or 4 cm in size.

The three systematic reviews of TACE-PA combination therapy ^100-102 included studies of varying patient populations that were collectively broader than that included in our evidence review. For example, the reviews included studies in patients with more advanced disease or those with unclear Child-Pugh status, as well as studies in which the treatment was given prior to 2000. As such, these reviews included studies that reported comparisons not examined in our review (e.g., TACE-PEI vs. TACE). However, given the heterogeneity across studies and the paucity of high-quality comparative data from randomized clinical trials, the overall strength of evidence is insufficient to permit conclusions regarding these comparisons. Comparing RFA-TACE combination therapy to RFA monotherapy in a meta-analysis, Yan et al.¹⁰² reported that the combination therapy was associated with higher survival rates. However, the majority of included studies in that review were of low quality with small sample sizes, and, therefore, Yan et al. judged the overall strength of evidence as low, indicating uncertainties around the pooled estimate of effect. Wang et al.¹⁰⁰ conducted a meta-analysis of TACE-PEI combination therapy versus TACE monotherapy and found an improved overall survival with the combination therapy. The included trials in this review were of generally poor quality, with unclear baseline patient characteristics (e.g., Child-Pugh class and HCC lesion characteristics) and unclear or inadequate blinding and allocated concealment. As such, the authors of the review acknowledged the limited reliability of their conclusion. In another meta-analysis of TACE-PA combination therapy versus PA monotherapy,¹⁰¹ the combination therapy was shown to improve overall survival compared with the monotherapy. However, in a sensitivity analysis of TACE-RFA versus RFA alone, the authors found that the survival benefit of the combination therapy was not robust, which is in agreement with the inconclusive evidence base identified in our review. This systematic review also included studies in which the treatment was given prior to 2000. The authors noted the limited availability of high-quality data in their pooled analysis; therefore, the findings of this review are limited as well.

A 2009 Cochrane Review¹⁰³ compared PEI and PAI, two similar ablative techniques with different chemotherapeutic agents for injection, and found no significant difference with regard to overall survival. This finding supports our approach of combining the PEI and PAI groups in our meta-analysis of the RFA versus PEI/PAI comparison.

The strength of the present review is that it addresses all local hepatic therapies for the included indications and includes comparisons not previously examined in published systematic reviews. Table 92 displays the corresponding comparisons between this review and the previously published reviews we identified. In addition, this report also recognizes that distinct patient groups exist within the population receiving local hepatic therapies. Specifically, we addressed a single patient population, those patients who are eligible for local hepatic therapy but are not otherwise eligible for resection or transplantation. Because we focused on a patient group rather than a specific intervention, we were able to present the outcomes for a wide range of local hepatic therapies for the target population.

Table 92

Comparisons made by current report and identified recent systematic reviews.

Implications for Clinical and Policy Decisionmaking

The goal of any local hepatic therapy for unresectable HCC is to prolong life by eliminating the tumor if possible or to palliate symptoms such as pain. This report has reviewed the literature on local hepatic therapies targeting these goals.

For the comparison of RFA to PEI/PAI, our conclusions suggest that for these patients treatment with RFA confers a survival benefit at 3 years compared with PEI/PAI. In addition, TTP and local recurrence may be improved in patients treated with RFA compared with PEI/PAI. Patients treated with RFA also seem to have longer LOS after treatment compared with those treated with PEI/PAI. Beyond this evidence on the comparative effectiveness of these procedures was insufficient. Subsequent comparisons had only one or no comparative studies on a given treatment comparison. For these comparisons, evidence was insufficient for all outcomes; thus, there is no comparative evidence base to support decisionmaking. In cases where comparative evidence existed, data were judged to be insufficient due to high risk of bias and/or imprecision of estimates.

Limitations of the Comparative Effectiveness Review Process

Determination of the scope of this review was derived from a lengthy process that began in topic development and continued to be refined even as the CER was underway. The topic was initially broader, encompassing other primary tumors metastasizing to the liver and HCC. During the scoping process, this review was narrowed to focus solely on unresectable HCC, and then further by excluding transplant eligible patients and those who were treated in an effort to downstage them for resection. Based on the refined scope, the literature search revealed an evidence base with limited comparative data. When examining the comparative efficacy of local hepatic therapies it is important to establish that patient groups are comparable. In general, patients treated with ablative therapies and those treated with transarterial strategies represent two distinct patient populations, and as a result, when considering comparisons for this review we compared only ablative therapies to one another, embolization therapies to one another, and external-beam therapies to one another. Combinations of therapies were presented together, but none utilized the same interventions and could not be synthesized. Nonetheless, the evaluation of the quality of the body of literature to assess our KQs and the identification of research needs is a valuable contribution to the field.

Limitations of the Evidence Base

Limitations of the present review are related largely to two factors: (1) the lack of comparative evidence and (2) clinical heterogeneity of patient populations across studies. With the exception of six RCTs, the vast majority of the evidence base included in this review derived from observational—mostly single-arm—studies. The clinical heterogeneity was most evident in the description of patient and tumor characteristics. For example, the size of lesions being treated with RFA ranged from 4 cm or smaller in the trial by Lin⁵¹ to up to 10 cm in a study by Minami et al.⁵⁶ Often, studies failed to report on these patient and tumor characteristics, which potentially impact treatment-related outcomes. For example, only 17 out of 48 (35.4%) included studies reported both the number and size of lesions in the study patient population. Authors varied in how these tumor characteristics were described including: mean number and size of tumors, median number and size of tumors, range of number and size of tumors, percent solitary and nonsolitary tumor, interquartile range of size and number, or other categorizations. Full descriptions of the patient population is important, as those with—for example—higher ECOG score (i.e., worse functioning status), higher HCC stage, higher Child-Pugh class cirrhosis, or multinodular disease, generally experience poorer outcomes than those without. For this reason, it is ideal to stratify the studies by patient groups (e.g., BCLC stage A versus BCLC stage B) and to compare studies of equivalent patient populations. However, the poor patient characterization in the studies precluded stratification by patient groups as well as indirect comparison of interventions across studies. To maintain clinical relevance, comparisons were only made within category of intervention (e.g., ablative therapy vs. ablative therapy). This stratification is because patients with different disease characteristics are candidates for different treatments (e.g., patients with small accessible tumors are candidates for ablation whereas more extensive disease would undergo embolization therapy). Exceptions to this were two cross category comparisons of RFA and TACE and RFA versus TACE+RFA. The patient populations in these studies were patients eligible for ablative therapy. Chok and colleagues compared RFA to TACE in a patient population with tumor diameters less than 5cm with less than four nodules.⁵³ This cross-category comparison was included under the ablative therapies section because Chok et al. assessed the performance of TACE in these patients to determine if selection bias (caused by advanced disease and poor liver functional reserve) contributed to the perceived benefit of RFA compared to TACE.

The comparative data were limited even further in terms of important subgroups such as those based on age, sex, ECOG score, disease etiology, Child-Pugh class, presence of PVT, HCC stage, lesion size, and multifocal versus single-nodule HCC. Overall survival was examined by subgroup in three RCTs; however, none of these analyses were prespecified, thereby limiting their utility beyond hypothesis generation.

Given the limited number of patients and clinical heterogeneity, we did not systematically review the treatment-specific characteristics such as treatment regimens and techniques used. A very large sample size with uniform data collection of these variables would be required to assess whether specific treatment characteristics were associated with survival differences.

None of the studies included in this review used blinded outcome assessment. It can be a challenge to blind participants and outcome assessors in these studies due to the differences in treatment delivery and the appearance of the liver after treatment. This is a particular limitation for the assessment of intermediate outcomes such as progression and local recurrence.

In addition to the RCTs meeting our inclusion criteria, this review included four nonrandomized comparative studies. These studies did not use statistical adjustment to reduce confounding; such adjustment for confounding should be consistently used in nonrandomized studies. Regardless of the study design, we suggest that studies examining the effectiveness or comparative effectiveness of local hepatic therapies address potential confounders and effect measure modification that could obscure the results. This is particularly important for patient characteristics such as size and number of the lesions, Child-Pugh classification, and performance status, which could serve as both modifiers of the effectiveness and factors that are considered when choosing the best local hepatic therapy.

Although RCTs may not be possible for all comparisons in all centers, well done multivariate analyses from existing case series can aid in identifying additional factors that should be documented and potentially controlled for in the comparative analysis of these data. These analyses can enhance the design of future RCTs or observational studies.

Research Gaps

This systematic review attempted to compare outcomes of local hepatic therapies for patients treated for unresectable HCC without evidence of extrahepatic spread who are not eligible for transplant. Evidence on patient outcomes is limited. There was a moderate strength of evidence to support that RFA improved 3-year overall survival compared with PEI/PAI. There was low strength of evidence to support higher TTP, less local recurrence, and a longer LOS for RFA compared with PEI/PAI. For all other comparisons and outcomes, strength of evidence was judged to be insufficient.

We identified four broad evidence gaps during this review:

• There is no evidence on quality of life. Quality-of-life outcomes are particularly important for a population of patients in which palliation is often the focus of therapy. For all comparisons, collection and reporting of quality-of-life data using standard measurement tools is needed.
• An objective of CER is to understand the comparative effects for different subgroups. RCTs should prespecify subgroup analyses to assess the effects of characteristics such as lesion size, Child-Pugh class, and ECOG score on treatment outcomes. The subgroups of interest must be delineated using systematic definitions of patient subgroups. Further, studies should present data by these subgroups so that evidence can be interpreted accordingly.
• Future studies should employ a standard or uniform set of outcome definitions (e.g., overall survival, local recurrence) as well as patient characteristics to report (e.g., BCLC stage, Child-Pugh class, lesion number and size). Such uniformity would allow for a more accurate and level comparison of patient populations across studies which the current evidence base precludes.
• During the Peer Review process of this CER, we received the following suggested comparisons for future research: (1) RFA versus other ablative therapies (e.g., MWA, cryoablation), (2) RFA versus TACE-RFA combination therapy, (3) RFA versus radiotherapies (e.g., SBRT), and (4) between transarterial therapies (e.g., TACE versus RE or TACE versus DEB). Such comparative evidence, based on well-designed randomized studies in the patient population included in this review, is needed.

Conclusions

This review included 13 local hepatic therapies and their combinations for unresectable HCC. There was a moderate strength of evidence demonstrating better overall survival at 3 years, a low level of evidence supporting improved overall survival for patients with larger lesion sizes, and a low strength of evidence for improved TTP and local control for RFA compared with PEI/PAI for the treatment of unresectable HCC. A low level of evidence also supports a longer LOS following RFA compared with PEI/PAI. For all other outcomes and comparisons, there is insufficient evidence to permit conclusions on the comparative effectiveness of local hepatic therapies for unresectable HCC. Important direct health outcomes of therapy include overall survival, adverse effects, and quality of life. Progression-free survival is an important intermediate health outcome, as progression often marks a change in therapy. Future RCTs comparing RFA with other ablative therapies and comparisons between transarterial therapies (e.g., TACE versus RE) are needed to close the existing gap in the comparative evidence.