Efficacy and Safety of First-Line Targeted Treatment and Immunotherapy for Patients with Biliary Tract Cancer: A Systematic Review and Meta-Analysis

Simple Summary We performed a meta-analysis of all clinical trials of first-line combination therapies for advanced biliary tract cancer and concluded that combining immunotherapy with chemotherapies could improve survival for patients with aBTCs by increasing the objective response rate. Abstract Background: Biliary tract cancer is one of the most aggressive and fatal tumours. Gemcitabine with cisplatin chemotherapy has long been the first-line treatment, but the prognosis is poor. In recent years, targeted treatment and immunotherapy have produced encouraging outcomes requiring a thorough review and meta-analysis. Method: For this systematic review and meta-analysis, we searched four databases, starting from the inception dates of databases to 11 January 2022. This study comprised randomised clinical trials and cohort studies that used immunotherapy or targeted treatment as the first line of treatment for patients with biliary tract cancer. Results: From the 888 studies extracted, 33 trials were examined and found to meet the criteria. These included 3087 patients, 16 single-arm trials, 13 RCTs, one nRCT, a prospective single-arm pilot study, and a clinical setting in the real world. From 2010 to 2020, 33 studies were conducted using targeted treatment or immunologic therapies as first-line treatments for BTC patients, and 18 of those studies had positive outcomes. Conclusion: This study demonstrates that immunotherapy combined with chemotherapy as first-line treatment can provide survival benefits by improving the objective response rate for patients with unresectable biliary tract cancer. The potential for combination therapy to become a new trend in clinical treatment is promising but needs further clinical evaluation.


Introduction
Biliary tract cancer (BTC), which includes intrahepatic cholangiocarcinoma (ICC) and extrahepatic cholangiocarcinoma (ECC), is one of the most malignant and lethal tumours (ECC). The epidemiological characteristics of ICC and ECC differ. ECC is the most common subtype of cholangiocarcinoma in general. However, ICC is more common in some East Asian countries, accounting for 15% to 20% of all liver cancers and typically invading the bile duct wall [1]. In Western countries, the incidence and mortality rate of ICC are increasing [2]. There are significant gender and ethnic differences in ICC incidence and mortality. Men are 1.5 times more likely than women to develop ICC. Asians have a 2.0 times higher incidence rate than whites and blacks. Southeast Asia and China have the highest prevalence of ICC worldwide. Indian Americans, Alaska Natives, and Asians have the highest ICC mortality rates, while Caucasians and blacks have the lowest [3].
We focused on RCT and cohort studies. Case-control studies, letters, reviews, case reports, and articles that do not provide raw data were not included. When data came from different phases of the same experiment, only studies with the most complete and up-to-date data were retained. No restrictions were placed on the language of the article. We also searched CNKI but did not find any more additional Chinese studies.

Population
This study targeted populations who needed to meet all the following criteria: (1) Unresectable gallbladder cancer and cholangiocarcinoma diagnosed by histopathology or cytology.
(2) The patient has not received systemic treatment for unresectable biliary cancer. (3) The patient has at least one measurable lesion.

Interventions
Targeted and immunologic agents used alone or in combination with chemotherapeutic agents were included in this study. The drug targets include MEK1/2, EGFR, VEGF, mTOR, PD-1/PD-L1, MET, and CTLA-4 (as shown in Table 1). The primary outcomes were the objective response rate (ORR), disease control rate (DCR), overall survival (OS), and progression-free survival (PFS).
Overall Survival (OS) is the time between randomisation and the onset of death by any cause. Objective Response Rate (ORR) is the proportion of patients whose tumour volume decreases to a predetermined value. ORR equals the ratio of complete response (CR) to partial response (PR), or ORR = CR + PR. ORR excludes both stable disease (SD) and the effect of the disease's natural progression. Smaller sample sizes and shorter follow-up periods are required. Partial response (PR) is defined as a volume reduction of at least 30 percent in all tumours that can be measured. The Disease Control Rate (DCR) is the proportion of patients whose cancer diminishes or stabilises over time. DCR equals the sum of the rates of complete remission, partial remission, and stable disease. PFS is the time between randomisation and the onset of objective tumour progression or death from any cause, which is a surrogate endpoint for OS.

Secondary Outcomes
Secondary outcomes were treatment-related adverse events (TRAE).

Data Sources and Search Strategy
Two investigators searched PubMed, Web of Science, Embase, Cochrane Library, and clinicaltrials.gov, starting from the inception dates of databases to 11 September 2022. The terms "biliary tract cancer", "cholangiocarcinoma", gene mutation type such as "EGFR", and drug names such as "pembrolizumab" were used as keywords to search titles or abstracts (See Appendix A for a detailed search strategy).

Literature Selection
Endnote 20 software was used to import all the search results. First, duplicates were removed. Second, two researchers independently screened titles and abstracts to determine inclusion eligibility. Third, full manuscripts of potentially eligible trials were read to determine which studies should be included. During the literature selection process, disagreements were resolved through discussion between the two researchers with the assistance of a third researcher as needed [14].

Data Extraction
Two researchers extracted and compiled data. The first author, study method, publication time, journal of publication, follow-up time, number of patients, baseline level of patients, observed indicators, and interventions were all extracted from the study data. If there were any disagreements, the researchers discussed them. If two researchers cannot reach an agreement, a third researcher makes the final decision. We recorded the data in Microsoft Excel.

Quality Assessment
The PRISMA guidelines were followed when conducting the systematic review. The quality of the included literature was assessed using Version 2 of the Cochrane risk-of-bias tool for randomised trials and plotted in the quality evaluation table [15]. Non-randomised intervention studies were evaluated using ROBINS-I (Risk Of Bias In Non-randomised Studies of Interventions) [16].

Statistical Analysis
We used the "meta" package in R environment to perform the meta-analysis using frequentist approach [17]. Heterogeneity was assessed with I 2 . A fixed-effects model was used for small heterogeneity (I 2 < 25%), and a random-effects model was used for a large one (I 2 > 25%) [18]. Bias is tested using funnel plots if the heterogeneity between studies Cancers 2023, 15, 39 5 of 29 in the meta was too large (I 2 > 75%) [19]. We determined the rank of all interventions by using the 'netrank' function in the 'netmeta' package in R to obtain P-scores. The P-score ranged from 0 to 1, showing a progressive rise in the efficacy of the included medications based on the estimation result and confidence interval of the effect value [18].

Study Inclusion
We initially identified a total of 888 articles from four databases. After excluding 179 duplicates, 709 articles remained. Then, we screened titles and abstracts to exclude the types of articles that did not meet the criteria, leaving 128 articles. Full-text screening was performed to exclude 73 articles with immunotherapies or targeted therapies as second-line therapy, Seven articles with a mix of first-line and second-line therapy, and eight articles whose studies included other diseases were excluded, leaving 33 included. After searching clinicaltrial.gov and checking the reference of the included literature, 17 additional studies were identified, resulting in a final 33 studies that met all the inclusion criteria ( Figure 1). The total number of patients included in the trials is 3087. We used the "meta" package in R environment to perform the meta-analysis using frequentist approach [17]. Heterogeneity was assessed with I 2 . A fixed-effects model was used for small heterogeneity (I 2 < 25%), and a random-effects model was used for a large one (I 2 > 25%) [18]. Bias is tested using funnel plots if the heterogeneity between studies in the meta was too large (I 2 > 75%) [19]. We determined the rank of all interventions by using the 'netrank' function in the 'netmeta' package in R to obtain P-scores. The P-score ranged from 0 to 1, showing a progressive rise in the efficacy of the included medications based on the estimation result and confidence interval of the effect value [18].

Study Inclusion
We initially identified a total of 888 articles from four databases. After excluding 179 duplicates, 709 articles remained. Then, we screened titles and abstracts to exclude the types of articles that did not meet the criteria, leaving 128 articles. Full-text screening was performed to exclude 73 articles with immunotherapies or targeted therapies as secondline therapy, Seven articles with a mix of first-line and second-line therapy, and eight articles whose studies included other diseases were excluded, leaving 33 included. After searching clinicaltrial.gov and checking the reference of the included literature, 17 additional studies were identified, resulting in a final 33 studies that met all the inclusion criteria ( Figure 1). The total number of patients included in the trials is 3087.

Targeted Therapies
This study included 22 trials focusing on targeted therapies, with a total of 1658 patients. The basic characteristics of patients are listed in Table 2.
Among the 22 studies evaluating targeted therapy, 11 studies were single-arm phase II trials; 10 studies were randomised parallel phase II trials, and one study was a phase III trial. The publication years ranged from 2010 to 2021, with the earliest trial starting in 2006. There were 11 single-arm trials, 11 controlled trials, one with three groups (Valle et al., 2020 [20]), and three trials were blinded (Santoro et al., 2015 [21]; Valle et al., 2015 [22]; Moehler et al., 2014 [23]). One nRCT (Factorial assignment) and 10 RCTs were included.

Targeted Therapies
This study included 22 trials focusing on targeted therapies, with a total of 1658 patients. The basic characteristics of patients are listed in Table 2.
Among the 22 studies evaluating targeted therapy, 11 studies were single-arm phase II trials; 10 studies were randomised parallel phase II trials, and one study was a phase III trial. The publication years ranged from 2010 to 2021, with the earliest trial starting in 2006. There were 11 single-arm trials, 11 controlled trials, one with three groups (Valle et al., 2020 [20]), and three trials were blinded (Santoro et al., 2015 [21]; Valle et al., 2015 [22]; Moehler et al., 2014 [23]). One nRCT (Factorial assignment) and 10 RCTs were included. Seven trials were completed in the United States; five were multicenter trials, and the others were distributed in Italy, Australia, Austria, France, Korea, Taiwan, Denmark, and Germany. The primary tumour sites are shown in Table 3 below. 694 patients with IHC, 384 patients with gallbladder cancer, 235 patients with EHC, 44 patients with hilar cholangiocarcinoma, and 17 patients with Vater ampulla carcinoma were included. Three studies (consisting of 219 patients) did not specify the disease type in patients with BTC, and 48 patients were classified as "other disease types" in the original literature, which included patients with liver metastases.

Immunotherapies
As shown in Table 4, the included immunotherapy-related studies were published between 2018 and 2021. The trial of Oh et al., 2020 [41] of Bintrafusp alfa is a phase 2/3 trial. Oh et al., 2022 [42] is a phase 3 trial, demonstrating the effectiveness of Durvalumab in combination with chemotherapy therapies. No immunotherapy was used alone in any of the study's regimens, which all combined both immunotherapy and chemotherapy as the first-line treatment. A total of 1286 patients were included.  The primary tumour sites of immunotherapy in the studies are listed in Table 5, among which most were ICC, 460 cases, accounting for 50.9% of the total.

Quality Assessment
Figures 2 and 3 depict the quality of the included RCTs. Overall, the included randomised controlled trials were of high quality. El-Khoueiry (2012) [24] was terminated after the first phase of accrual because the primary objectives were not met. Iyer (2018) [38] is a meeting abstract lacking detailed data.  As shown in Table 6, ROBINS-1 assessed the quality of 16 single-arm studies observational studies. All studies documented the definition of controls and the com rability of cases and controls, except for one study that did not report case defini (Khoueiry, 2012 [24]).   As shown in Table 6, ROBINS-1 assessed the quality of 16 single-arm studies and observational studies. All studies documented the definition of controls and the comparability of cases and controls, except for one study that did not report case definitions (Khoueiry, 2012 [24]).  As shown in Table 6, ROBINS-1 assessed the quality of 16 single-arm studies and observational studies. All studies documented the definition of controls and the comparability of cases and controls, except for one study that did not report case definitions (Khoueiry, 2012 [24]).  [52] showed that patients with BTC treated with bevacizumab + GEMOX obtained a median PFS of 7.0 months (95% CI 5.3-10.3) and a 6-month PFS of 63% (47-79), below the set target rate of 70%, with objective responses recorded in 19 patients and an overall disease control rate of 80%. The trial by Gruenberger et al., 2010 [27] had an overall disease control rate of 80%, progression-free survival of 8.8 months (95% CI 5.1-12.5), and median overall survival of 15.2 months (9.9-20.5) for all treated patients. The trial by Lau et al., 2018 [28], using Everolimus 10 mg/d alone, had a median PFS of 5.5 months (95% confidence interval (CI: 2.1-10.0 months) and a median OS of 9.5 months (95% CI: 5.5-16.6 months). Notably, gallbladder cancer had a significantly worse DCR at 12 weeks than other anatomic sites and a trend toward worse PFS and OS, but the treatment was well tolerated. Sohal et al., 2013 [30] added Irinotecan and Panitumumab to Gemcitabine, which had a median PFS of 9.7 months and a median OS of 12.9 months, showing encouraging efficacy and good tolerability of this regimen. The trial by Borbath et al., 2013 [31] met the primary endpoint with a median PFS time of 5.8 months (95% CI 3.6-8.5 months), median OS time of 13.5 months (95% CI 9.8-31.8 months), and 53.7% of patients remained alive at 1 year, suggesting that Gemcitabine-Cetuximab has activity in BTC and that KRAS status is not associated with PFS and, unlike cutaneous toxic effects, may serve as a surrogate marker of efficacy. Lubner et al., 2010 [39] showed that 87% of patients showed disease progression, with a median time to disease progression of 4.4 months and a median OS of 9.9 months. This study concluded that the combination of bevacizumab and erlotinib demonstrated significant activity in treating advanced BTC, with few adverse events of grades 3 or 4. Bevacizumab and Erlotinib demonstrated significant activity in advanced BTC with few Grade 3 or 4 adverse events. The trial by Leone et al., 2016 [36] added Panitumumab to GEMOX, and the results confirmed a marginal effect of anti-EGFR therapy in WT-KRAS-selected BTC.
Lee et al., 2012 [33] combined erlotinib with GEMOX and showed no significant difference in progression-free survival between the groups. Still, adding erlotinib to gemcitabine and oxaliplatin showed antitumour activity: significantly more patients had objective responses in the chemotherapy plus erlotinib group than in the chemotherapy alone group (40 patients versus 21 patients; p = 0.005), but median overall survival was the same in both groups. The trial by Chen et al., 2015 [34] showed a trend toward improved PFS was observed, but the addition of cetuximab did not significantly improve the ORR of GEMOX chemotherapy in advanced BTC, and KRAS mutations did not affect the trend in ORR and PFS differences between C-GEMOX and GEMOX. In the study by Jensen et al., 2012 [35], the addition of Panitumumab to chemotherapy resulted in a 6-month progression-free survival (PFS) rate of 31/42 [74%; 95% confidence interval (CI) 58% to 84%], a disease control rate of 86%, a median PFS of 8.3 months (95% CI 6.7-8.7 months), and a median overall survival of 10.0 months (95% CI. 7.4-12.7 months). Hezel et al., 2014 [26] used a combination of gemcitabine, oxaliplatin, and panitumumab for KRAS wild-type metastatic BTC and achieved a remission rate of 45% and a disease control rate of 90%. Its median PFS was 10.6 months (95% CI 5-24 months), and median overall survival was 20.3 months (95% CI 9-25 months).
Other trials did not meet the expected endpoints but were still informative. Khoueiry et al., 2012 [24], as a phase II study of sorafenib in patients with advanced BTC based on the role of the RAS-RAF-MEK-ERK pathway and the VEGF axis in BTC, was terminated after phase I due to failure to meet the primary objective. The trial by Khoueiry et al., 2014 [25] to study sorafenib and erlotinib was also terminated after Phase I enrollment. Lee (2012) [33] added sorafenib to gemcitabine and cisplatin for biliary tract adenocarcinoma, which did not improve efficacy compared with historical data and had increased toxicity. In the trial of Santoro et al., 2015 [21], patients were randomised in a 1:1:1. The results showed no statistical difference between secondary endpoints except for ORR, and the V/G combination was slightly outperformed by the other treatments. Patients in the three groups reported similar rates of adverse effects. Malka et al., 2014 [29] concluded that adding cetuximab to gemcitabine and oxaliplatin did not appear to enhance chemotherapeutic activity in patients with advanced BTC, although it was well tolerated.
The trial by Valle et al., 2015 [22] showed that Cediranib did not improve progressionfree survival in patients with advanced BTC in combination with cisplatin and gemcitabine. Valle et al., 2020 [20] added Ramucirumab or Merestinib to GEM + CIS standard chemotherapy and showed no improvement in PFS, OS, or ORR. Lowery et al., 2019 [37] demonstrated that Binimetinib, combined with Gemcitabine and Cisplatin, had no effect on PFS-6-month or RR. Moehler et al., 2014 [23] similarly demonstrated that adding Sorafenib to Gemcitabine did not improve outcomes in patients with advanced BTC, but biomarker subgroup analysis suggested that some patients may benefit from the combination. Iyer et al., 2018 [38] demonstrated that adding Bevacizumab to Gemcitabine/Capecitabine did not improve prognosis in unselected patients with advanced BTC compared to historical controls. Vogel et al., 2018 [40] concluded that combining Panitumumab with chemotherapy did not improve ORR, PFS, or OS in patients with KRAS wild-type advanced BTC.

Immunotherapies
The trial of Yu et al., 2021 [43] resulted in an ORR of 14.3% (95% CI: 1.8 to 42.8), a DCR of 64.3% (95% CI: 41.7 to 86.9), a median PFS of 6.5 months (95% CI: 3.8 to 9.2), PFS rates of 61.6% and 12.3% at 6 and 12 months, respectively, and a median OS of 9.9 months (95% CI: 7.6 to 12.2), concluding that Camrelizumab in combination with chemotherapy as firstline treatment for metastatic BTC demonstrated acceptable safety and efficacy. Chen et al., 2015 [34] also concluded that Camrelizumab plus GEMOX as first-line treatment for patients with advanced BTC looked promising, with a median PFS that was 6.1 months and median OS that was 11.8 months. Sun et al., 2018 [44] concluded that the combination of PD-1 antagonist plus chemotherapy or targeted therapy was effective and tolerable as firstline treatment for advanced BTC. OS was significantly longer in the group treated with the combination drug than in the chemotherapy group (median, 8. Gou et al., 2021 [45] yielded results that in advanced BTC, anti-PD-1 therapy plus chemotherapy prolonged PFS compared to chemotherapy alone, and AE was tolerable. Oh et al., 2020 [46] showed that adding D + T immunotherapy to chemotherapy was tolerable and showed promising efficacy. Chiang et al., 2021 [48] concluded that Nivolumab in combination with a modified GS (gemcitabine and S-1) is a promising regimen with a good safety profile. Oh et al., 2020 [41] showed that Bintrafusp alfa was clinically active in Asian patients with BTC and had a durable response. Oh et al., 2022 [42] concluded that Durvalumab in combination with chemotherapy for BTC significantly improved OS compared to chemotherapy alone (D: 12.8 (11.1-14.0) vs. placebo: 11.5 (10.1-12.5)). The combination also greatly improved progression-free survival compared with chemotherapy alone. Median progression-free survival with durvalumab combined with gemcitabine and cisplatin was 7.2 months compared with 5.7 months with chemotherapy alone (HR = 0.75; p = 0.001). The proportion of progression-free patients was 34.8% and 24.6% at 9 months and 16.0% and 6.6% at 12 months, respectively. The ORR also improved, with an overall efficacy rate of 26.7% with durvalumab/chemotherapy compared with 18.7% with chemotherapy alone, with a superiority ratio of 1.60 for efficacy (p = 0.011). The trial by Sahai et al., 2020 [49] concluded that the combination of nivolumab with chemotherapy drugs failed to improve efficacy.

Combined Therapies
In the study by J. Zhou  Twelve patients experienced disease progression, and four patients died. Median PFS and OS had not been reached. The median duration of response has not been determined, and the 6-month OS rate was 90%. A quantity of 43% (13/30) of patients experienced grade 3 or higher adverse events (AEs). This study showed that ORR was significantly associated with PD-L1 expression and mutations associated with DNA damage repair (DDR) in tumour samples. In patients with advanced ICC, the combination of toripalimab, lenvatinib, and GEMOX chemotherapy was well tolerated and showed an encouraging ORR.

Meta-Analysis for OS of Targeted Therapy
As the heterogeneity was considerably high (I 2 = 63%), a random effects model was used to obtain a pooled OS of 10.65 months for the targeted drug treatment group (Figure 4). As evidenced by the funnel plot with a more symmetrical distribution, these studies have less publication bias. Egger's test for a regression intercept gave a p-value of 0.1377 > 0.05, indicating no evidence of publication bias ( Figure 5). A subgroup meta-analysis of the chemotherapy and combined therapy with targeted therapy, yielded I 2 = 65%, p-value = 0.21 ( Figure 6). So, a difference between the two groups could not be demonstrated. The asymmetry of the funnel plot indicates a possible publication bias (Figure 7). The Egger's test could not be applied because the sample size was less than 10.  As evidenced by the funnel plot with a more symmetrical distribution, these studi have less publication bias. Egger's test for a regression intercept gave a p-value of 0.137 > 0.05, indicating no evidence of publication bias ( Figure 5). A subgroup meta-analysis of the chemotherapy and combined therapy with targete therapy, yielded I 2 = 65%, p-value = 0.21 ( Figure 6). So, a difference between the two group could not be demonstrated. The asymmetry of the funnel plot indicates a possible publication bias (Figure 7). Th Egger's test could not be applied because the sample size was less than 10. A subgroup meta-analysis of the chemotherapy and combined therapy with targeted therapy, yielded I 2 = 65%, p-value = 0.21 ( Figure 6). So, a difference between the two groups could not be demonstrated. As evidenced by the funnel plot with a more symmetrical distribution, these studies have less publication bias. Egger's test for a regression intercept gave a p-value of 0.1377 > 0.05, indicating no evidence of publication bias ( Figure 5). A subgroup meta-analysis of the chemotherapy and combined therapy with targeted therapy, yielded I 2 = 65%, p-value = 0.21 ( Figure 6). So, a difference between the two groups could not be demonstrated. The asymmetry of the funnel plot indicates a possible publication bias (Figure 7). The Egger's test could not be applied because the sample size was less than 10. The asymmetry of the funnel plot indicates a possible publication bias (Figure 7). The Egger's test could not be applied because the sample size was less than 10.

Meta-Analysis for OS of Immunotherapy
The pooled overall survival is 15.62 months for the immunotherapy group, becau the heterogeneity I 2 = 84% > 50%, random effects model was applied (Figure 8).

Meta-Analysis for PFS of Targeted Therapies
Data on PFS were available for six studies involving immunotherapy and 17 studi involving targeted treatment. Plots of each study's individual PFS and its confidence i tervals were made for the two groups ( Figure 9). In the targeted therapy group, the m dian PFS was 6.02 months (95% CI: 5.01-7.03) (range: 2.0-13.5 months).

Meta-Analysis for OS of Immunotherapy
The pooled overall survival is 15.62 months for the immunotherapy group, because the heterogeneity I 2 = 84% > 50%, random effects model was applied (Figure 8).

Meta-Analysis for OS of Immunotherapy
The pooled overall survival is 15.62 months for the immunotherapy group, because the heterogeneity I 2 = 84% > 50%, random effects model was applied (Figure 8).

Meta-Analysis for PFS of Targeted Therapies
Data on PFS were available for six studies involving immunotherapy and 17 studies involving targeted treatment. Plots of each study's individual PFS and its confidence intervals were made for the two groups ( Figure 9). In the targeted therapy group, the median PFS was 6.02 months (95% CI: 5.01-7.03) (range: 2.0-13.5 months).  3.6. Meta-Analysis for PFS 3.6.1. Meta-Analysis for PFS of Targeted Therapies Data on PFS were available for six studies involving immunotherapy and 17 studies involving targeted treatment. Plots of each study's individual PFS and its confidence intervals were made for the two groups ( Figure 9). In the targeted therapy group, the median PFS was 6.02 months (95% CI: 5.01-7.03) (range: 2.0-13.5 months).

Meta-Analysis for OS of Immunotherapy
The pooled overall survival is 15.62 months for the immunotherapy group, because the heterogeneity I 2 = 84% > 50%, random effects model was applied (Figure 8).

Meta-Analysis for PFS of Targeted Therapies
Data on PFS were available for six studies involving immunotherapy and 17 studies involving targeted treatment. Plots of each study's individual PFS and its confidence intervals were made for the two groups ( Figure 9). In the targeted therapy group, the median PFS was 6.02 months (95% CI: 5.01-7.03) (range: 2.0-13.5 months).   As evidenced by the funnel plot with a more symmetrical distribution, these studies have less publication bias ( Figure 10). In addition, Egger's test for a regression intercept gave a p-value of 0.3583 > 0.05, indicating no evidence of publication bias. As evidenced by the funnel plot with a more symmetrical distribution, these studies have less publication bias ( Figure 10). In addition, Egger's test for a regression intercept gave a p-value of 0.3583 > 0.05, indicating no evidence of publication bias. Further subgroup analysis was performed to divide the studies of targeted therapies into two groups, EGFR and VEGF, and the pooled PFS was obtained from the forest plot (Figures 11 and 12). After Egger's test, the p-value of EGFR was 0.9437 > 0.05, and the pvalue of VEGF was 0.3214 > 0.05, indicating no publication bias.  Data on progression-free survival were available for seven studies in the immunotherapy group. Individual PFS and their confidence intervals were plotted for each study within the two groups ( Figure 13). The median of PFS was 8.56 months (range: 2.5-12.3 Further subgroup analysis was performed to divide the studies of targeted therapies into two groups, EGFR and VEGF, and the pooled PFS was obtained from the forest plot (Figures 11 and 12). After Egger's test, the p-value of EGFR was 0.9437 > 0.05, and the p-value of VEGF was 0.3214 > 0.05, indicating no publication bias. As evidenced by the funnel plot with a more symmetrical distribution, these studies have less publication bias ( Figure 10). In addition, Egger's test for a regression intercept gave a p-value of 0.3583 > 0.05, indicating no evidence of publication bias. Further subgroup analysis was performed to divide the studies of targeted therapies into two groups, EGFR and VEGF, and the pooled PFS was obtained from the forest plot (Figures 11 and 12). After Egger's test, the p-value of EGFR was 0.9437 > 0.05, and the pvalue of VEGF was 0.3214 > 0.05, indicating no publication bias.

Meta-Analysis of PFS of Immunotherapies
Data on progression-free survival were available for seven studies in the immunotherapy group. Individual PFS and their confidence intervals were plotted for each study within the two groups ( Figure 13). The median of PFS was 8.56 months (range: 2.5-12.3 months) (95% CI: 6.40-10.73) in the immunotherapy group. As evidenced by the funnel plot with a more symmetrical distribution, these studies have less publication bias ( Figure 10). In addition, Egger's test for a regression intercept gave a p-value of 0.3583 > 0.05, indicating no evidence of publication bias. Further subgroup analysis was performed to divide the studies of targeted therapies into two groups, EGFR and VEGF, and the pooled PFS was obtained from the forest plot (Figures 11 and 12). After Egger's test, the p-value of EGFR was 0.9437 > 0.05, and the pvalue of VEGF was 0.3214 > 0.05, indicating no publication bias.

Meta-Analysis of PFS of Immunotherapies
Data on progression-free survival were available for seven studies in the immunotherapy group. Individual PFS and their confidence intervals were plotted for each study within the two groups ( Figure 13). The median of PFS was 8.56 months (range: 2.5-12.3 months) (95% CI: 6.40-10.73) in the immunotherapy group.

Meta-Analysis of PFS of Immunotherapies
Data on progression-free survival were available for seven studies in the immunotherapy group. Individual PFS and their confidence intervals were plotted for each study within the two groups ( Figure 13). The median of PFS was 8.56 months (range: 2.5-12.3 months) (95% CI: 6.40-10.73) in the immunotherapy group. The meta-analysis yielded a pooled PFS of 6.20 for the targeted drug treatment and 6.33 for the targeted combination chemotherapy group. Still, the p-value for the subgroup analysis was 0.60, which was not statistically different, with a considerable heterogeneity of I 2 = 68% ( Figure 14). The funnel plot is asymmetric with more results falling on the left side, suggesting possible publication bias ( Figure 15). The pooled PFS for obtaining immune combination chemotherapy drug treatment was 9.89 months, p-value = 0.6 for subgroup analysis, concluding that there was no significant difference in PFS between the two groups ( Figure 16). The meta-analysis yielded a pooled PFS of 6.20 for the targeted drug treatment and 6.33 for the targeted combination chemotherapy group. Still, the p-value for the subgroup analysis was 0.60, which was not statistically different, with a considerable heterogeneity of I 2 = 68% ( Figure 14). The meta-analysis yielded a pooled PFS of 6.20 for the targeted drug treatment and 6.33 for the targeted combination chemotherapy group. Still, the p-value for the subgroup analysis was 0.60, which was not statistically different, with a considerable heterogeneity of I 2 = 68% ( Figure 14). The funnel plot is asymmetric with more results falling on the left side, suggesting possible publication bias ( Figure 15). The pooled PFS for obtaining immune combination chemotherapy drug treatment was 9.89 months, p-value = 0.6 for subgroup analysis, concluding that there was no significant difference in PFS between the two groups ( Figure 16). The funnel plot is asymmetric with more results falling on the left side, suggesting possible publication bias ( Figure 15).

Meta-Analysis of DCR
The meta-analysis yielded a pooled DCR of 76.6% (I 2 = 29%) for the immune-combination chemotherapy regimen. In contrast, the pooled disease control rate for targeted agents was 79.1% (see Figures 20 and 21).

TRAE of Targeted Therapies
The results of treatment-related adverse events are summarised in Table 7. The top ten of these were neutropenia, thrombocytopenia, anemia, fatigue, diarrhea, leukopenia, neuropathy, rash, hypertension, and hand-foot skin reactions. In the erlotinib group, TRAE was less in the group targeted therapy combined chemotherapy than in the control group with chemotherapy agents alone, especially hand-foot syndrome, which occurred The pooled PFS for obtaining immune combination chemotherapy drug treatment was 9.89 months, p-value = 0.6 for subgroup analysis, concluding that there was no significant difference in PFS between the two groups ( Figure 16). The funnel plot is asymmetric, with more results falling on the right side, suggesting possible publication bias ( Figure 17).

Meta-Analysis of ORR
The meta-analysis heterogeneity of chemotherapy drugs combined with immune drugs versus chemotherapy drugs alone was 0%, yielding an OR of 1.622 ( Figure 18). So, the combination of chemotherapy immune drugs yielded a higher objective response rate than chemotherapy alone. The pooled objective response rate for the targeted therapy group was 32.1%, I 2 = 78%, p-value < 0.01 ( Figure 19). The ORR values of the three trials in immunotherapy, which had immunotherapy drugs in combination with chemotherapy drugs to compare with chemotherapy drugs alone, found no significant difference. The funnel plot is asymmetric, with more results falling on the right side, suggesting possible publication bias ( Figure 17). The funnel plot is asymmetric, with more results falling on the right side, suggesting possible publication bias ( Figure 17).

Meta-Analysis of ORR
The meta-analysis heterogeneity of chemotherapy drugs combined with immune drugs versus chemotherapy drugs alone was 0%, yielding an OR of 1.622 ( Figure 18). So, the combination of chemotherapy immune drugs yielded a higher objective response rate than chemotherapy alone. The pooled objective response rate for the targeted therapy group was 32.1%, I 2 = 78%, p-value < 0.01 ( Figure 19). The ORR values of the three trials in immunotherapy, which had immunotherapy drugs in combination with chemotherapy drugs to compare with chemotherapy drugs alone, found no significant difference.

Meta-Analysis of ORR
The meta-analysis heterogeneity of chemotherapy drugs combined with immune drugs versus chemotherapy drugs alone was 0%, yielding an OR of 1.622 ( Figure 18). So, the combination of chemotherapy immune drugs yielded a higher objective response rate than chemotherapy alone. The funnel plot is asymmetric, with more results falling on the right side, suggesting possible publication bias ( Figure 17).

Meta-Analysis of ORR
The meta-analysis heterogeneity of chemotherapy drugs combined with immune drugs versus chemotherapy drugs alone was 0%, yielding an OR of 1.622 ( Figure 18). So, the combination of chemotherapy immune drugs yielded a higher objective response rate than chemotherapy alone. The pooled objective response rate for the targeted therapy group was 32.1%, I 2 = 78%, p-value < 0.01 ( Figure 19). The ORR values of the three trials in immunotherapy, which had immunotherapy drugs in combination with chemotherapy drugs to compare with chemotherapy drugs alone, found no significant difference. The pooled objective response rate for the targeted therapy group was 32.1%, I 2 = 78%, p-value < 0.01 ( Figure 19). The ORR values of the three trials in immunotherapy, which had immunotherapy drugs in combination with chemotherapy drugs to compare with chemotherapy drugs alone, found no significant difference.

Meta-Analysis of DCR
The meta-analysis yielded a pooled DCR of 76.6% (I 2 = 29%) for the immune-combination chemotherapy regimen. In contrast, the pooled disease control rate for targeted agents was 79.1% (see Figures 20 and 21).

TRAE of Targeted Therapies
The results of treatment-related adverse events are summarised in Table 7. The top ten of these were neutropenia, thrombocytopenia, anemia, fatigue, diarrhea, leukopenia, neuropathy, rash, hypertension, and hand-foot skin reactions. In the erlotinib group, TRAE was less in the group targeted therapy combined chemotherapy than in the control group with chemotherapy agents alone, especially hand-foot syndrome, which occurred in up to 20 cases in grade 3/4 but not in the combination group. A total of 27 cases of rash occurred in the group with cetuximab and none in the group with chemotherapy alone. Moreover, with cetuximab, the incidence of TRAE (n = 72) was greater than in the placebo group (n = 33). TRAE incidence was higher in both Ramucirumab and Merestinib compared to the placebo group. TRAE was also higher with panitumumab (n = 82) compared

Meta-Analysis of DCR
The meta-analysis yielded a pooled DCR of 76.6% (I 2 = 29%) for the immune-combination chemotherapy regimen. In contrast, the pooled disease control rate for targeted agents was 79.1% (see Figures 20 and 21).

Meta-Analysis of DCR
The meta-analysis yielded a pooled DCR of 76.6% (I 2 = 29%) for the immune-combination chemotherapy regimen. In contrast, the pooled disease control rate for targeted agents was 79.1% (see Figures 20 and 21).

TRAE of Targeted Therapies
The results of treatment-related adverse events are summarised in Table 7. The top ten of these were neutropenia, thrombocytopenia, anemia, fatigue, diarrhea, leukopenia, neuropathy, rash, hypertension, and hand-foot skin reactions. In the erlotinib group, TRAE was less in the group targeted therapy combined chemotherapy than in the control group with chemotherapy agents alone, especially hand-foot syndrome, which occurred in up to 20 cases in grade 3/4 but not in the combination group. A total of 27 cases of rash occurred in the group with cetuximab and none in the group with chemotherapy alone. Moreover, with cetuximab, the incidence of TRAE (n = 72) was greater than in the placebo group (n = 33). TRAE incidence was higher in both Ramucirumab and Merestinib compared to the placebo group. TRAE was also higher with panitumumab (n = 82) compared to chemotherapy alone (n =33), where skin toxicity was 36 cases. In the trial of

Meta-Analysis of DCR
The meta-analysis yielded a pooled DCR of 76.6% (I 2 = 29%) for the immune-combination chemotherapy regimen. In contrast, the pooled disease control rate for targeted agents was 79.1% (see Figures 20 and 21).  The results of treatment-related adverse events are summarised in Table 7. The top  The results of treatment-related adverse events are summarised in Table 7. The top ten of these were neutropenia, thrombocytopenia, anemia, fatigue, diarrhea, leukopenia, neuropathy, rash, hypertension, and hand-foot skin reactions. In the erlotinib group, TRAE was less in the group targeted therapy combined chemotherapy than in the control group with chemotherapy agents alone, especially hand-foot syndrome, which occurred in up to 20 cases in grade 3/4 but not in the combination group. A total of 27 cases of rash occurred in the group with cetuximab and none in the group with chemotherapy alone. Moreover, with cetuximab, the incidence of TRAE (n = 72) was greater than in the placebo group (n = 33). TRAE incidence was higher in both Ramucirumab and Merestinib compared to the placebo group. TRAE was also higher with panitumumab (n = 82) compared to chemotherapy alone (n =33), where skin toxicity was 36 cases. In the trial of panitumumab, the combination was higher than chemotherapy alone. In the trial of Moehler et al. [23], there were 18 cases of grade 3/4 TRAE in the sorafenib group, while there were 30 cases in the chemotherapy alone group, which is more than in the combination group.   Table 8 below. Sun (2018) [44] showed no significant difference in TRAEs between the monotherapy and combination groups. the TOPAZ-1 trial by D. Y. Oh [50] concluded that the incidence of grade 3/4 adverse reactions was lower in the durvalumab group than in the placebo group. The incidence of grade 3/4 treatment-related adverse events (TRAEs) was 62.7% in the durvalumab-treated group and 64.9% in the placebo-treated group. The rates of TRAEs leading to treatment discontinuation were 8.9% and 11.4%, respectively.  [53] 84.2% of patients had one TRAE. Fatigue (n = 14), anorexia (n = 8), increased alanine aminotransferase (ALT) (n = 7) or aspartate aminotransferase (AST) (n = 7), rash (n = 6), hypertension (n = 5), and hoarseness (n = 5), were the most common TRAEs of any grade. An amount of 34.2% of patients had grade 3 TRAEs, the most prevalent of which were fatigue (n = 5) and hypertension (n = 3). One patient experienced a grade 4 cerebral hemorrhage as a result of hypertension, while five (13.9%) and one (2.8%) patient experienced dose reductions and treatment suspensions as a result of TRAEs. Due to Lenvatinib-related adverse effects, the dose of four individuals was reduced from 8 mg to 4 mg per day. Due to treatment-related cerebral bleeding, one patient terminated Lenvatinib plus PD-1 inhibitor therapy. There were six postoperative problems among patients who had resection, including two cases of biliary leakage, two cases of pleural effusion, one case of delayed liver function recovery, and one incidence of upper gastrointestinal haemorrhage (Table 9). observed efficacy signal was modest and not superior to using gemcitabine plus cisplatin alone. The combination of panitumumab with chemotherapy did not improve ORR, PFS, or OS in patients with advanced BTC with KRAS WT. According to a meta-analysis by Vogel (2018) [40], EGFR receptor antagonists did not reveal a benefit on sustained patient survival compared to chemotherapy alone only benefit. Therefore, no further studies investigating the combination of chemotherapy with anti-EGFR antibodies are needed. In contrast, Lau (2018) [28] showed that Everolimus showed clinical activity as first-line monotherapy for advanced BTC in unselected patients.
In three phase 2 trials, sorafenib did not improve survival in this setting. Similarly, other EGFRs, such as Vandetanib and Cetuximab, also failed to improve the prognosis of patients. All these trials suggest that adding EGFR-targeting agents to GEMOX is feasible and safe but ineffective. The lack of efficacy may be related to the heterogeneity of the target population for advanced BTC, the suboptimal treatments explored, or the need for alternative endpoints after survival.
Two studies on the combination of immunotherapy and targeted therapy were included. Preliminary data showed that lenvatinib in combination with PD-1 inhibitors showed some efficacy in patients with advanced ICC. Both pembrolizumab and nivolumab showed antitumour effects when combined with lenvatinib. The effect of this combined regimen on overall survival in individuals with advanced ICC is still being studied in clinical trials.
Despite the low response rate of targeted and immunotherapy in BTC and the scarcity of clinical trial data, more research is needed, and better individualised therapy as well as drug combinations may be the way forward for such promising antitumour agents.
It needs to indicate that, in the included studies, outcome data were not counted separately according to the patient's site of tumour development. They therefore could not be compared based on differences in the anatomical characteristics of the biliary tract.
BTC has an immunogenic profile, implying that immunotherapy is promising. However, current studies show that immune checkpoint inhibitors have limited activity in first-line therapy. With three drugs approved for marketing as second-line therapy, targeted agents have shown some success, but evidence of efficacy as first-line therapy is still lacking. More high-quality RCTs based on patient target genotyping are required to investigate the efficacy of using targeted agents as first-line BTC therapy. This article summarises and analyses current clinical trials in which immune or targeted agents have been added to standard BTC treatment as first-line therapy. Most clinical trials for targeted or immune agents as first-line treatments for BTC are still in the early stages, and future results will provide more evidence for future research. Future multi-institutional clinical trials should allow for large-scale studies that stratify patients based on anatomical subtype and genetic drivers to predict response and prognosis to new treatment regimens.
The drawback of this study is that our conclusions are based on some unadjusted analyses and may be influenced by additional confounding factors, including gender, age, genotypic mutation status, prior systemic medication, and other characteristics. Second, we were unable to conduct further subgroup analyses to assess the efficacy and safety of immunotherapy or targeted therapies in patients with various conditions due to a lack of etiological data.

Conclusions
This study demonstrates that immunotherapy combined with chemotherapy as firstline treatment can provide survival benefits by improving the objective response rate for patients with unresectable BTC. The potential for combination therapy to become a new trend in clinical treatment is promising. However, because the research design of existing clinical trials is insufficient, more comparable and high-quality investigations of regimens based on immunotherapies or targeted therapy as first-line treatment are required.
Author Contributions: X.Y. contributed to the study selection and data extraction. X.Y. and H.Z. contributed to data analysis. X.Y., H.Z., Y.L., C.O.L.U. and H.H. participated in study design, drafting and revision of the manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: This work is supported by the University of Macau (MYRG2020-00230-ICMS).

Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.