How are you taking part in this consultation?

You will not be able to change how you comment later.

You must be signed in to answer questions

    The content on this page is not current guidance and is only for the purposes of the consultation process.

    Appendix Meta-analysis

    The 6 primary studies in the key evidence (tables 2 and 3) were included in the meta-analysis.

    Data analysis

    The effects of treatments (RFA plus stent versus stent only) on OS and stent patency were examined using pooled HR and weighted mean difference[1], respectively, with 95% CI. Data on OS and stent patency was analysed using the random-effects generic inverse variance model. Sensitivity analyses were carried out to explore the influence of different factors on the effect size for OS and stent patency (including study design, approaches to RFA, repeated RFA sessions, the use of other antitumour treatments, and stent types). Adverse events were analysed using an exploratory approach, and data was pooled in a fixed-effect meta-analysis using the Mantel-Haenszel model.

    Heterogeneity was assessed using Chi2 and I2 statistics. I2 values of 25%, 50%, and 75% were considered as low, medium, and high levels of heterogeneity. Forest plots were used to display the meta-analysis results. All analyses were performed using Cochrane Review Manager V5.

    Results

    OS

    All 6 studies (n=1,023) reported a measure of survival, with 4 studies presenting HRs and 2 studies describing median OSs. When comparing RFA plus stent with stent only, the pooled HR for OS from 6 studies (n=1,040) was 0.50 (95% CI, 0.36 to 0.69; figure 1a), favouring RFA plus stent insertion. This effect was statistically significant (p<0.0001), with moderate heterogeneity (I2=56%). This direction of effect on survival is consistent across all individual studies as shown in figure 1a.

    A sensitivity analysis for 5 RCTs versus 1 observational study showed statistically significant effects for both subgroups (RCTs: HR, 0.48 [95% CI 0.29 to 0.78], I2=63%; observational study, HR, 0.55 [95% CI, 0.44 to 0.70]) and there was no detectable subgroup difference (I2=0%, p=0.61; figure 1b).

    It is noted that 2 studies (Kang 2021a, 2021b) included a small proportion of people undergoing RFA via PTC (≤20%). Also, these 2 studies reported median OSs which were then used to calculate HRs. In contrast, the other 4 studies included people undergoing ERCP-directed RFA and reported HRs. Proportional hazards were assumed, even though proportionality was untested. Visual observation of the Kaplan-Meier curves suggested that this assumption was considered reasonable, except for Albers (2022). Although the Kaplan-Meier curves in the Albers study crossed during the first 3 months, after which the curves diverged and then showed parallel curves until the end of follow up; so, its impact on the overall result would likely to be small. A sensitivity analysis of studies with endoscopic RFA versus mixed approaches to RFA was carried out. The analysis showed a statistically significant effect on survival after endoscopic RFA (HR, 0.46; 95% CI, 0.31 to 0.67; I2=69%; p<0.0001) but not after mixed approaches to RFA (HR, 0.72; 95% CI, 0.42 to 1.25; I2=0%; p=0.24). There was no statistically significantly subgroup difference, but with moderate heterogeneity (I2=43.8%; figure 1c).

    When considering studies with versus without repeated RFA sessions, a sensitivity analysis indicated a statistically significant effect of RFA on OS for the use of repeated RFA sessions (HR, 0.43; 95% CI, 0.28 to 0.64; I2=66%; p<0.0001) but not for single RFA session (HR, 0.73; 95% CI, 0.46 to 1.14; I2=0%; p=0.17; figure 1d). There was no statistically significantly subgroup difference, but moderate heterogeneity presented (I2=65.6%, p=0.09).

    For the use of other antitumour treatments, especially chemotherapy as a key confounding factor, a sensitivity analysis of studies with versus without the use of other antitumour treatments was conducted. The effects of RFA on survival for both subgroups were statistically significant (studies with the use of other antitumour treatments: HR, 0.56 [95% CI, 0.47 to 0.66]; studies without the use of other antitumour treatments: HR, 0.18 [95% CI, 0.09 to 0.37]), but there was a subgroup difference with significant heterogeneity (I2=89.2%, p=0.002; figure 1e).

    Noticeably, only Yang (2018) excluded people who underwent systemic therapies. After separating this study, the analysis for studies with other antitumour treatments showed no observed heterogeneity, indicating a substantial reduction in the level of heterogeneity. One possible explanation would be due to a different RFA application protocol used. People in the Yang study received more RFA sessions (half of the people had 3 or more sessions of RFA) and had their stent replaced more frequently (every 3 months) than people in other studies. This difference could also contribute to the large effect of RFA on survival and prolong stent patency duration (as described in the 'stent patency' section).

    Across the 5 studies that involved the use of other antitumour treatments, the proportions of people who received other treatments were generally low (18% versus 17% across studies) but comparable between the RFA plus stent group and the stent only group within studies. This suggests that the results from this subgroup may be more representative of people who were forgoing systemic treatments. In fact, such treatments (e.g. chemotherapy) might have limited efficacy, because MBO is usually diagnosed in people with advanced CCA or pancreatic cancer (Kang 2021a). In addition, due to strict eligibility criteria, systemic treatments may not be suitable for many people with advanced cancer. Nevertheless, careful interpretation of the results is warranted and further exploration of repeated RFAs (such as its application frequency) would be valuable.

    When considering different types of stents used, a sensitivity analysis was conducted. Evidence suggested statistically significant effects of RFA on OS for the subgroups using plastic stents (HR, 0.31; 95% CI, 0.12 to 0.82; I2=84%; p=0.02) or both types of stents (HR, 0.55; 95% CI, 0.44 to 0.70; I2=0%; p<0.0001), but not for the SEMS subgroup (HR, 0.73; 95% CI, 0.46 to 1.14; I2=0%; p=0.17; figure 1f). Despite that, there was no statistically significant subgroup difference with low heterogeneity. It is noted that Kang (2021b) inserted plastic stents and then exchanged to SEMS after 3 months. The stent patency was measured from enrollment to at least 1 SEMS occlusion or death. Thus, this study was included in the subgroup with the use of both types of stents.

    Stent patency

    In term of stent patency, its duration was reported in 4 RCTs (n=317). The pooled weighted mean difference in patency duration between RFA plus stent and stent only was 0.97 months (95% CI, -0.67 to 2.62; I2=61%; p=0.25; figure 2a), but this effect was not statistically significant, with significant heterogeneity. This non-significant effect was also consistent with the finding from each individual RCT except for Yang 2018. Given the difference in the RFA application protocol used, the results need to be interpreted with caution.

    In terms of stent types, a sensitivity analysis of studies using different stents was conducted. No evidence suggested that RFA statistically significantly prolonged patency duration in all 3 subgroups (figure 2b), and there was no detectable subgroup difference (I2=0%, p=0.93).

    Stent patency is one of the factors that affect survival. The mechanism of the survival benefit of RFA may be explained by its ability to relieve biliary obstruction and prevent recurrent cholangitis. Interestingly, in this meta-analysis, although RFA fails to show a prolonged effect on stent patency, it is proven to have a survival benefit. One possible explanation would be that mechanisms other than stent patency improve survival outcomes: 1) RFA reduces the maximal tumour load differing the disease's progression (Xia 2021); 2) in situ tumour destruction provided a useful antigen source for the induction of antitumor immunity; 3) RFA may play a role through indirect antitumour effects (Cha 2021).

    Cholangitis and cholecystitis

    Cholangitis is typically inflammation of the biliary tract, commonly caused by infection. Cholangitis influences morbidity and mortality. Based on 6 studies (n=1,023), the pooled risk ratio of cholangitis between groups was 0.94 (95% CI, 0.72 to 1.22; I2=0%; p=0.64; figure 3a). A sensitivity analysis for acute cholangitis (within 30 days) was conducted. The pooled risk ratio reduced slightly to 0.86 (95% CI, 0.53 to 1.36), but also no evidence of a statistically significant difference between groups (p=0.49; figure 3b).

    For cholecystitis, which is an inflammation of the gallbladder, commonly caused by a blockage, the pooled risk ratio from 4 studies (n=910) was 9.48 (95% CI, 2.96 to 30.31; I2=7%; p=0.0002; figure 4) between groups. This indicates that RFA carries a higher risk of cholecystitis than stent insertion alone, although the estimates are very imprecise. One possible explanation for this imprecision is that RCTs are not powered for safety outcomes. Nevertheless, most individual studies favour stent only and extra precaution would be necessary during ablation. It is noted that all cases were acute cholecystitis (within 30 days) except for 1 delayed cholecystitis after RFA plus stent; so, the impact of this delayed case on the overall outcome is likely to be minimal.



    [1] Medians were used when means were not reported.