Immunotherapies in clinical development for biliary tract cancer

ABSTRACT

Introduction: Cancers of the biliary tract (BTC) are aggressive malignancies with limited treatment options and an overall dismal prognosis. In recent years, two concepts, namely precision oncology and immune oncology (IO) have profoundly influenced and, in some cancers, even revolutionized tumor treatments. While positive data from randomized trials have led to the incorporation of targeted concepts for genetically select BTC patients, IO is not yet implemented in clinical practice.

Areas covered: We discuss published results from completed, as well as from ongoing studies on IO in BTC, based on a literature search on Pubmed and information provided by clinicaltrials.gov in October 2020. Apart from monotherapy, we outline IO-based combination approaches and highlight pivotal studies whose results will likely influence the future development of relevant concepts in BTC.

Expert opinion: Despite partially positive signals, IO thus far disappointed in unselected BTC populations and should currently not be considered as a preferred systemic treatment in patients with microsatellite stable disease outside of clinical trials. In the coming years, a better understanding of the molecular mechanisms underlying resistance to checkpoint inhibition, and the identification of positive predictive biomarkers will be important for the successful integration of IO into treatment concepts for BTC patients.

KEYWORDS:

• Cholangiocarcinoma
• chemotherapy
• PD-1/PD-L1
• CTLA-4
• targeted therapies

1. Introduction

Cancers of the biliary system (BTC) are highly aggressive tumors that either originate within the liver (intrahepatic cholangiocarcinoma (iCCA)), in the perihilar or distal bile ducts (perihilar or distal CCA), or in the gallbladder [1,2]. In most countries, CCA is considered a ‘rare’ cancer with incidence rates below 6/100.000. However, due to the demographic distribution of different risk factors and probably influenced by ethnic factors, the incidence of CCA ranges from 0.1/100.000 in Australia to more than 110/100.000 in Northeast Thailand. In Europe, most biliary tumors occur sporadically after the age of 50, with a slight predominance of male patients. Risk factors include obesity, viral hepatitis B and C, primary sclerosing cholangitis (PSC) and, for gallbladder carcinoma, gallstones. In Asian countries, infestation with parasitic liver flukes is considered an important risk factor. Surgical resection is the only potentially curative approach and should be offered to patients who are diagnosed at an early stage. However, due to the late manifestation of clinical symptoms, most patients suffer from locally advanced or metastatic disease at the time of diagnosis, and even after successful resection early recurrence is frequent. Most patients are eventually bound to receive palliative treatments, and the dismal median overall survival (mOS) of 11–13 months under systemic palliative therapy with gemcitabine and cisplatin (GemCis) highlights the urgent need to expand the limited therapeutic measures available to date for patients with advanced BTC [3–5].

In the last years, treatment concepts for several solid malignancies were genuinely influenced by two novel ‘approaches’ to anti-cancer therapy, namely precision oncology and immune oncology (IO). With the broader access to tumor genetic diagnostics, the therapeutic promise of targeted therapies was realized for the treatment of BTC, and we will briefly summarize the two most relevant developments with direct clinical implications below. In addition, the advent of IO put the host immune response under the spotlight as a target for anti-tumor therapies – a novel concept that revolutionized the treatment of several solid malignancies, but thus far has not excelled in biliary tract cancers. In this review, we summarize the recent clinical data on the efficacy of IO in BTC. Furthermore, we discuss ongoing phase-III studies, the potential prospect of predictive biomarkers, and possible future directions.

2. Precision oncology in BTC

In patients with intrahepatic CCA (iCCA), mutations in isocitrate dehydrogenase (IDH) 1 or 2 occur with a frequency of 10–20%. As the first and thus far still the only positive phase-III trial in precision oncology in BTC, the ClarIDHy study investigated the efficacy of the IDH1 inhibitor ivosidenib, compared to placebo, in 185 IDH1 mutant CCA patients that progressed on first-line therapy [6]. The primary endpoint – an improvement in median progression-free survival (mPFS) – was achieved with an impressive HR of 0.37 (95% CI 0.25–0.54; p < 0.001), but with an mPFS of only 2.7 months for ivosidenib compared to 1.4 months for placebo. Regardless of the minor absolute PFS difference and the low radiological response rate of 2.4%, PFS rates at 6 and 12 months were clinically relevantly improved (32.0% and 21.9% in the ivosidenib arm at 6 and 12 months, respectively; 0% at both time points in the placebo group). Following statistical adjustment for the cross over design, median overall survival (mOS) of 10.8 months under ivosidenib was significantly improved compared to the mOS of 6 months in the placebo arm (HR 0.46; p = 0.0008). mOS was not significantly improved in the intention to treat (IIT) population, most likely to the permitted cross over (35/61 in the placebo arm). Until now, however, U.S. Food and Drug Administration (FDA) or European Medicines Agency (EMA) approval for ivosidenib for the treatment advanced BTC has not yet been applied for.

One of the most frequent genetic alterations that are highly promising targets for targeted therapies are FGF-receptor 2 (FGFR2) alterations. Oncogenic FGFR2 fusions occur in up to 15% of iCCA patients. Among others, the FIGHT-202 phase-II trial investigated the efficacy of the FGFR inhibitor pemigatinib in progressed fusion-positive iCCA patients and was thus far the largest study conducted in this genetically selected population. A disease control rate (DCR) of 85.1% and an overall response rate (ORR) of 36%, with a clinically meaningful mPFS and mOS of 6.9 and 21.1 months, respectively, established a new benchmark in the second or higher lines of therapy [7]. Based on these results, pemigatinib received FDA-approval for pre-treated CCA patients with FGFR2-fusions in April 2020.

3. Mono immunotherapy in BTC

The success of immunotherapy in several solid malignancies, including lung cancer, renal cell cancer, or malignant melanoma, sparked the enthusiasm for this novel treatment approach and fostered considerable hope that these encouraging results could also be recapitulated in patients with biliary cancers. Accordingly, several trials were initiated in BTC in the 1^st and 2^nd line setting (Table 1).

Table 1. Ongoing ICB-based clinical trials in BTC

Trial number	Phase	Pathways targeted	Drug	Indication	Line of treatment	Primary outcomes	Secondary outcomes	Country
Neoadjuvant
NCT04308174	2	PD-1, chemotherapy	Durvalumab, Gemcitabine, Cisplatin	rBTC	neoadjuvant	R0 resection rate	OS, EFS, AEs, ORR	Asia
NCT04506281	2	PD-1, TKI, chemotherapy	Toripalimab, Lenvatinib, GEMOX	iCCA	neoadjuvant	EFS	OS, ORR, pathological remission rate, AEs	Asia
NCT03867370	1/2	PD-1, Placebo	Toripalimab, Placebo	HCC, iCCA	neoadjuvant	Path. response rate	Time to operation	Asia
Adjuvant
NCT04295317	2	PD-1, chemotherapy	SHR-1210, Capecitabine	iCCA	adjuvant	RFS	OS, side effects	Asia
Palliative – Phase 1/2
Pembrolizumab
NCT03937895	1/2	Allogenic NK cell, PD-1	SMT-NK, Pembrolizumab	aBTC	I/II	DLT, ORR	TTP, safety	Asia
NCT03110328	2	PD-1	Pembrolizumab	aBTC	II	ORR, PFS, OS	-	Asia
NCT03260712	2	PD-1, chemotherapy	Pembrolizumab, Cisplatin, Gemcitabine	aBTC	I	PFS	ORR, toxicity, PFS	Spain
NCT03695952	2	PD-1	Pembrolizumab, Nivolumab	aBTC, HCC	I/II	ORR	AEs	Asia
NCT04550624	2	PD-1, TKI	Pembrolizumab, Lenvatinib	aCCA	II	ORR	AEs, DOR, PFS, OS	Asia
NCT03111732	2	PD-1, chemotherapy	Pembrolizumab, Oxaliplatin, Capecitabine	aBTC	II+	PFS	Safety, ORR, OS	US
NCT03849469	1	PD-1, CTLA-4, LAG-3	XmAb®22,841/ Pembrolizumab	iCCA (advanced solid tumors)	II	Safety and tolerability	NA	US
NCT03797326	2	PD-1, VEGF	Pembrolizumab/ Lenvatinib	Pretreated Solid Tumors	II	ORR, AEs	DCR, PFS, OS	US
NCT03895970	2	PD-1, VEGF	Pembrolizumab/ Lenvatinib	aBTC	II	ORR, DCR, PFS	OS, DoR	Asia
NCT03937895	1/2	PD-1, SMT-NK	Pembrolizumab/ Allogeneic NK Cell	aBTC	II	DLT, ORR	TTP, toxicity	Asia
Toripalimab
NCT04010071	2	PD-1, VEGF	Toripalimab, Axitinib	aBTC, HCC	II	ORR, PFS	DCR, OS, DoR, SD, PFS rate	Asia
NCT04361331	2	PD-1, TKI, chemotherapy	Toripalimab, Lenvatinib, GEMOX	a(i)CCA	I/II	ORR	Safety, OS, PFS	Asia
NCT03982680	2	PD-1, chemotherapy	Toripalimab, Gemcitabine, 5-FU	aCCA	I	PFS, toxicity	ORR, DCR, OS	Asia
NCT04027764	2	PD-1, chemotherapy	Toripalimab/S1/ Albumin Paclitaxel	aBTC	I	ORR	PFS, DCR, OS	Asia
NCT03796429	2	PD-1, chemotherapy	Toripalimab/ Gemcitabine-S1	aBTC	I	PFS, OS	ORR, Safety	Asia
NCT04191343	2	PD-1, chemotherapy	Toripalimab/GEMOX	aBTC	I	ORR	NA	Asia
NCT04211168	2	PD-1, VEGF	Toripalimab/ Lenvatinib	aBTC	II	ORR, AEs	OS, PFS, OS, CBR	Asia
Nivolumab
NCT04172402	2	PD-1, chemotherapy	TS-1, Gemcitabine, Nivolumab	aBTC	I	ORR	-	Asia
NCT03785873	2	PD-L1, chemotherapy	Nivolumab/Nal-irinotecan/5FU, leucovorin	aBTC	II	Phase 1b: DLT; phase 2:PFS	AE ORR, OS	US
NCT02834013	2	PD-1, CTLA-4	Nivolumab/ Iipilimumab	aBTC (rare tumors)	II	ORR	Safety, OS, PFS, DCR	US
NCT04057365	2	PD-1, DKK1	Nivolumab/ DKN-01	aBTC	II	ORR (CR, PR)	PFS, OS	US
NCT03250273	2	PD-1, Histone deacetylase Inhibitor	Nivolumab/ Entinostat	aBTC	II	ORR	PFS, OS, DoR	US
Durvalumab
NCT03257761	1	PDL-1, DNMT inhibitor	Durvalumab/ Guadecitabine (SGI-110)	aCCA	II	DLT, ORR	Safety and tolerability, PFS, OS	US
NCT03473574	2	PD-1, CTLA-4, chemotherapy	Durvalumab, Tremelimumab, Gemcitabine, Cisplatin	aCCA	I	ORR	OS, AEs, QoL	Germany
NCT03704480	2	PD-1, CTLA-4, chemotherapy	Durvalumab, Tremelimumab, Paclitaxel	aBTC	II (after platin)	PFS	AEs, OS, QoL, toxicity, predictive markes of response	France
NCT04298008	2	PD-1, ATR	Durvalumab, AZD6738	aBTC	II (after immunotherapy)	DCR	ORR, PFS, DoR, OS, safety, QoL	Asia
NCT04301778	2	PD-1, CSF-1 R	Durvalumab, SNDX-6532	a(i)CCA (locally)	I/II	ORR, toxicity	OS, PFS, DOR	US
others
NCT04300959	2	PD-1, RTK, chemotherapy	Anlotinib, Sintilimab, Gemcitabine, Cisplatin	aBTC	I	OS (12 m)	OS (24 m), PFS, ORR, AE	Asia
NCT02091141	2	PD-1	Atezolizumab	Solid tumors, including aBTC, elevated tTMB	I/II	ORR, ORR by tTMB	DCR, PFS, DoR, 1-year-survival, AEs	US
NCT03833661	2	PD-1, TGF-ß	M7824	aBTC	II	ORR	DOR, PFS, AE, OS, pharmacogenetics	US
NCT04004234	1/2	PD-1, chemotherapy	Manganese Chloride, nab-paclitaxel, Gemcitabine, anti-PD-1 antibody	aBTC	I/II	AEs, PFS,	DCR, ORR, OS, serum cytokines and chemokines	Asia
NCT04454905	2	PD-1, VEGF	Camrelizumab, Apatinib	a(i)CCA	I/II	PFS	ORR, OS, DCR, DoR, TTP, AEs
NCT04066491	2/3	PD-L1; TGFβ Trap fusion Protein, chemotherapy	Gem-Cis with or without Bintrafusp Alfa (M7824)	aBTC	I	DLT, OS	DoR, AE, Pharmacokinetics	US
NCT03475953	1/2	PD-1, MAPK	Avelumab/ Regorafenib	metastatic solid tumors	I/II	RP2D, antitumour activity	MTD, DLT, toxicity	France
NCT03475953	1/2	PD-1, MAPK	Avelumab/ Regorafenib	Metastatic solid tumors	I/II	RP2D, antitumour activity	MTD, DLT, toxicity	France
Palliative – Phase 3
NCT03478488	3	PDL-1, chemotherapy	KN035/Gemcitabin/ Oxaliplatin	aBTC	I	OS	PFS, ORR	Asia
NCT04003636 KN-966	3	PD-1, chemotherapy	Pembrolizumab/ GemCis	aBTC	I	PFS, OS	ORR, DOR, AE	global
NCT03875235 TOPAZ-1	3	PDL-1, chemotherapy	Durvalumab or Placebo and GemCis	aBTC	I	OS	PFS, ORR, DoR	global

Ongoing clinical trials were identified by searching ClinicalTrials.gov using the terms ‘Biliary Cancer’, ‘cholangiocarcinoma’ and “biliary tract. The results were manually curated for inclusion of an immunotherapy arm. Trials were included with status of ‘Recruiting’, ‘Not yet recruiting’, ‘Active, not recruiting’, ‘Completed’ or ‘Enrolling by invitation’

DLT = Dose Limiting Toxicity, CBR = Clinical benefit rate, EFS = event free survival, AE = adverse events, RP2D = recommended phase II dose, MTD = maximum tolerated dose, TTP = time to progression, QoL = quality of life, NA = not available.

In 2020, some of these early prospective studies conducted in patients with progressed BTCs were published (Table 2); in contrast to the promising data with targeted therapies in molecularly defined BTC patients, IO has so far disappointed in unselected advanced BTC patients without microsatellite instability (MSI)/deficient mismatch repair (dMMR). The BTC subgroup from the global phase-II multicohort study KEYNOTE-158 (NCT02628067) is thus far the largest study population that was treated with pembrolizumab, and it included microsatellite stable (MSS), as well as MSI patients for individual evaluation [8]. The ORR was 5.8% (all partial responses (PR); 95% CI 2.1%–12.1%) in 104 MSS pretreated patients with an mOS and PFS of 7.4 months (95% CI 5.5–9.6) and 2.0 months (95% CI 1.9–2.1), respectively. 16.3% of patients reached disease stabilization resulting in a disease control rate (DCR) of 22.1%. The safety profile appeared manageable, and high-grade treatment-related adverse events were reported in 13.5% of patients (14/104, of which 13 were considered grade 3), causing treatment discontinuation in 5.8% of the study population. Paralleling observations in other cancers, in the (few) responding patients, responses were durable, all lasting > 6 months, and at least 50% with a duration of response (DOR) > 24 months. As expected, MSI/dMMR CCAs clearly outperformed the MSS counterparts, and the results of this separate subgroup analysis will be discussed in further detail below.

Table 2. Published IO-based clinical trials

Trial number	Phase	Pathways targeted	Drug	Indication	Line of treatment	Primary outcomes	Secondary outcomes	Country
Mono immunotherapy
NCT02829918	2	PD-1	Nivolumab	aBTC	II/III	ORR (CR, PR) 22% (10/46), SD 37% (17/46), DCR 59%	mOS 14.24 months, mPFS 3.68 months (PD-L1+ mPFS 10.4, mOS NR)	US
NCT02628067 (KN-158)	2	PD-1	Pembrolizumab	Subgroup aBTC (n = 104)	II+	ORR 5,8% (16% SD), DCR 22,1%	DoR n.r. (95% CI 6.2–15.7+), mPFS 2 months, mOS 7.4 months	US
Dual checkpoint inhibition
NCT02923934	2	PD-1, CTLA-4	Nivolumab/ Ipilimumab	rare tumors	I/II	ORR 23% (9/39), DCR 44%	mPFS 2.9 months (95% CI, 2.2–4.6, OS 5.7 months (95% CI, 2.7–11.9))	Australia
IO in combination with chemotherapy
JapicCTI-153,098	1	PD-1, chemotherapy	Nivolumab ± GemCis	aBTC	I/II	Nivo-mono: ORR 3%; Combination: ORR 37%	Nivo-mono: mPFS 1.4 months, mOS 5.2 months Combination: mPFS 4.2 months, mOS 15.4 months	Asia
NCT03101566	2	PD-1, CTLA-4, chemotherapy	Nivolumab +GemCis (Arm A) or Nivolumab +Ipilimumab (Arm B)	aBTC	I	PFS rate: Arm A 70%, Arm B 18.6%	mPFS Arm A 7.4 months (95% CI, 6.1–11.3), Arm B 4.1 months (95% CI, 2.4–5.2) mOS Arm A 10.6 months, Arm B 8.3 ORR n.a.	US
NCT03046862	2	PD-1, CTLA-4, chemotherapy	GemCis ± Durvalumab ± Tremelimumab	aBTC	I	ORR (GemCis 50%; GemiCis+IO 73%)	DCR (Gem/Cis 96,7% vs. +D 100% vs. +D + T 97,8%), PFS (Gem/Cis 13 m vs. +D 11 m vs. +D + T 11,9 m), DoR (Gem/Cis 11 m vs. +D 9,8 m vs. +D + T 9,1 m m), OS (Gem/Cis 15 m vs. +D 18,1 m vs. +D + T 20,7 m)	Asia
IO in combination with targeted therapies
NCT02443324	1	PD-1, VEGR	Pembrolizumab, Ramucirumab	aBTC	I/II/III	Toxicity: Hypertension	ORR 4%, mPFS 1.6 months, mOS 6.4 months	US
NCT03895970	2	PD-1, TKI	Pembrolizumab, Lenvatinib	aBTC	II+	ORR 25%, DCR 78.1%, mPFS 4.9 months (95% CI: 4.7–5.2),	mOS 11.0 months (95% CI: 9.6–12.3), AEs: Fatigue, hypertension, elevated aminotrasnferase levels	Asia
IO + intervention
NCT01853618	1/2	CTLA-4, ablation	Tremelimumab, RFA	BTC (Arm E)	II+	AEs	Best response (0% CR, 10% PR, 25% SD, 11% PD) PFS, OS	US

n.r. = not reported

In a conceptually similar trial that evaluated the efficacy of programmed cell death protein-1 (PD1) inhibition in a multicenter phase-II setting, 45 partially heavily pre-treated patients received nivolumab (NCT02829918) [9]. Here, results as assessed by central review appeared slightly more favorable with an ORR of 11% (95% CI 5–46%, central review) and a DCR of 50%. mPFS was 3.68 months (95% CI 2.30–5.69) and mOS was 14.24 months (95% CI 5.98 to not reached (NR)). The investigator review yielded an even superior ORR of 22% with a DCR of 59%, and the difference in outcome parameters was attributed to the discrepancy in the assessment of reactive vs metastatic lymphadenopathy. Of note, consistent with other studies, 4/10 responders (investigator-assessed) had a durable response that lasted at least one year, and no unexpected AEs were reported. Overall, the relatively small cohorts, particularly in the nivolumab trial, put a limitation to these studies, and the interpretation is further complicated by the possibility that response to checkpoint inhibition may differ by anatomical site.

4. Combination therapies

4.1. Dual checkpoint inhibition

Following pioneering work in other cancers, dual checkpoint inhibition also entered into (early) clinical studies in BTC [10,11]. Due to the low patient numbers in these studies, results again need to be interpreted with caution and should be regarded as early signals (Table 2).

The CA209-538 trial (NCT02923934) is a multicenter phase-II study that was conducted in Australia, and enrolled patients with various advanced cancers after one or more lines of therapy, including a BTC subgroup [12]. The dosing corresponded to the ‘Nivo3/Ipi1’ regimen, in which nivolumab at 3 mg/kg and ipilimumab at 1 mg/kg is administered every 3 weeks for 4 doses, followed by nivolumab monotherapy every two weeks [12]. The ORR in the BTC subgroup analysis was 23% (9/39), with a DCR of 44% (17/34). The mPFS and mOS were 2.9 (95% CI 2.2–4.6) and 5.7 months (95% CI 2.7–11.9), respectively. Of note, only 15 patients in this study entered the maintenance phase, after several of the partially heavily pretreated patients experienced early disease progression, and due to adverse events during the combination phase. Interestingly, responses were limited to patients with intrahepatic CCA or gallbladder cancer, and no responses were reported in patients with extrahepatic CCA. The rate of grade 3 or higher AE was 15%, which is lower than what was reported in other trials on dual checkpoint inhibition. However, as the authors also critically discuss and as mentioned before, this might be accounted for by the limited drug exposure in those patients who experienced rapid disease progression.

With a single-arm design, no valid conclusion can be drawn regarding the potential additive value of dual checkpoint inhibition. In addition to the general caveats associated with cross-trial assessments, a comparison with the results of NCT02829918 (nivolumab monotherapy) is especially challenging not only because of the discrepancies in investigator- and central review in the latter trial, but also due to the use of different nivolumab dosing schemes in the studies.

4.2. IO in combination with chemotherapy

A pronounced desmoplasia is a histological hallmark of BTC. This tumor reactive stroma is a site of intensive crosstalk between tumor cells and cells of the adaptive and innate immune system. There is accumulating evidence that chemotherapy might enhance the efficacy of immunotherapy by overcoming immunosuppressive effects of the microenvironment, increasing cross-presentation of tumor antigens, and supporting penetration of immune cells into the tumor core [13,14].

One of the first open-label multicenter phase-I trials evaluated the clinical efficacy of nivolumab alone and in combination with the standard of care, GemCis, as first-line treatment in mainly pre-treated Japanese patients [15]. In the monotherapy cohort, ORR was 3% with a mPFS of 1.4 months and an mOS of 5.2 months. Of note, the only patient in the monotherapy cohort (1/30) with a durable response and long-term survival was MSI. In the combination cohort, ORR was 37% with an mPFS of 4.2 months and mOS of 15.4 months. In both cohorts, most patients had a programmed cell death 1 ligand 1 (PD-L1) expression in less than 1% of tumor cells, whereas the relative proportion of PD-L1 expressing tumor-associated immune cells exceeded 1% in most samples (please see discussion on PD-L1 below).

The two-arm BilT-01 phase-II study evaluated the combination of GemCis with nivolumab and the combination of nivolumab and ipilimumab in a first-line setting. In the chemo-IO combination arm, mPFS was 8.8 months and mOS 10.6 months, and thus did not suggest a strong additive value for nivolumab when compared to historical controls [16]. Moreover, results in the IO combination arm disappointed with an mPFS of 4.13 and an mOS of 8.3 months.

First data for a combination with GemCis and the PD-L1 antibody durvalumab with and without the CTLA-4 antibody tremelimumab were presented at ASCO 2020 [17]. In this Korean population, the combinations yielded a very promising efficacy with a DCR of 100%, an mPFS of 11.0 (95% CI: 7.0–15.0) and an mOS of 18.1 (95% CI: 11.3–24.9) months in the durvalumab arm (n = 45) and with a DCR of 98%, an mPFS of 11.9 (95% CI: 10.1–13.7) and an mOS of 20.7 (95% CI: 13.8–27.6) months in the durvalumab/tremelimumab arm (n = 46) in combination with GemCis, respectively. The very similar outcome data in both arms do not suggest an additional benefit for the dual checkpoint inhibition over durvalumab alone in combination with chemotherapy. While PD-L1 expression prior to treatment was not associated with immunotherapy efficacy, increased PD-L1 expression in biopsy samples obtained after the 1^st cycle of GemCis correlated with improved mPFS. Interestingly, tumor mutational burden (TMB) failed to correlate with mPFS or mOS.

Whether these promising data can also be extrapolated to Caucasian patients needs to be shown in ongoing studies such as TOPAZ-1 (NCT03875235) and KEYNOTE-966 (NCT04003636), in which patients are treated with GemCis either alone or in combination with durvalumab or pembrolizumab, respectively (Table 1). Overall, however, there are not yet sufficient data to justify these combinations outside clinical trials in MSS patient.

4.3. IO in combination with targeted therapies

Antiangiogenic therapies targeting vascular endothelial growth factor receptor (VEGF) or VEGF receptor-2 (VEGFR2) have been shown to increase T-cell trafficking into tumors, and to reduce immunosuppressive cytokines and regulatory T-cells. VEGF(R) directed therapies may thus help to overcome resistance to checkpoint inhibitors, and the combination of atezolizumab and the VEGF-antagonist bevacizumab demonstrated convincing efficacy in hepatocellular carcinoma (HCC) [18].

Ramucirumab is an IgG1 monoclonal antibody that binds to the extracellular domain of VEGFR2 and has shown anti-tumor activity both as single agent and in combination in multiple tumor types, but not in BTC when combined with GemCis [3]. Similarly, whereas a multicohort, single-arm phase-I a/b trial reported promising activity for the combination of pembrolizumab and ramucirumab in gastric, lung, and urothelial carcinomas, [19], the efficacy in BTC was low with an ORR of 4%, an mPFS of 1.6 months and an mOS of 6.4 months [20].

In contrast to the focused targeting of VEGFR2, multityrosine kinase inhibitors with broader activity inhibit additional pathways, such as FGF signaling, which may augment the activity of checkpoint inhibitors. Lenvatinib in combination with pembrolizumab was recently awarded ‘breakthrough designation’ by the FDA for HCC, endometrial carcinoma, and renal cell cancer [21], and promising data were also reported from a Chinese BTC cohort. In this phase-II trial that included 32 patients after progression on chemotherapy, DCR was 78.1% with an mPFS of 4.9 months and an mOS of 11.0 months, suggesting that this combination could be further explored in BTC [22]. Of note, efficacy was higher in the PD-L1 positive population, which might help to select the target population for this approach.

Similar to VEGF signaling, transforming growth factor β (TGF-β) plays an immunosuppressive role in the tumor microenvironment. Based on encouraging efficacy observed in a phase-I study [23], M7824, a first-in-class bifunctional fusion protein composed of two extracellular domains of TGF-βRII (a TGF-β ‘trap’) fused to a human IgG1 mAb against PD-L1, is currently evaluated in combination with chemotherapy as first-line therapy in BTC (NCT04066491).

4.4. IO in combination with local therapies

Local therapies such as transcatheter arterial chemoembolization (TACE), radiofrequency ablation (RFA), radiotherapy, or selective internal radiation therapy (SIRT) are alternative treatment options for selected unresectable iCCA with reasonable effectiveness (mOS durations of 11–22 months), specifically for patients with liver limited disease.

Accumulating evidence suggests that the combination of radiotherapy and immunotherapy could be more effective than either treatment alone. Mechanistically, the excess antigen release following local tumor destruction is thought to trigger an adaptive immune response that is not limited to the lesion treated loco-regionally, but can also affect distant metastatic lesions – the so-called ‘abscopal effect’. In a case series of BTC patients with low TMB, pMMR, MSS, and negative PD-L1 expression, strong abscopal effects were observed in non-irradiated lesions in response to the combination of radiotherapy and immunotherapy [24]. Similarly, a significant increase in CD4⁺ and CD8⁺ T cells has been observed in resected HCC that were treated with SIRT prior to surgery, suggesting that the combination of SIRT with immunotherapy may represent an attractive therapeutic strategy for treatment of liver tumors [25]. This concept – a combination of SIRT and durvalumab/tremelimumab – is currently evaluated in the IMMUWHY study in patients with liver-dominant iCCA, among other trials (Table 3).

Table 3. Ongoing immunotherapy trials combined with locoregional intervention

Trial number	Phase	Pathways targeted	Drug	Indication	Line of treatment	Primary outcomes	Secondary outcomes	Country
IO + intervention
NCT03482102	2	PD-1, CTLA-4, radiation	Durvalumab, Tremelimumab, radiation therapy	aBTC, aHCC	II	ORR	AEs, OS, DCR, PFS, DOR, TTP	US
NCT02866383	2	PD-1, CTLA-4	Nivolumab, Ipilimumab, Radiotherapy	PC, BTC	II+	CBR (SD, CR, PR)	AEs, CBR (irRC), ORR, PFS, OS, QoL	Denmark
NCT04217954	2	PD-1, VEGF, chemotherapy	Toripalimab (i.v.); Oxaliplatin, 5-FU, Bevacizumab (HAIC)	aBTC	I	PFS, ORR	OS, AEs, CA19-9 level, change in imaging	Asia
NCT04238637	2	PD-1, CTLA-4, SIRT	Durvalumab, Tremelimumab, SIRT	a(i)CCA	I	ORR	Safety, DoR, PFS, OS	Germany
NCT03898895	2	PD-1, radiotherapy, chemotherapy	Anti-PD-1, radiotherapy, Gemcitabine, Cisplatin	a(i)CCA	I	PFS	OS, AEs, tumor response	Asia
NCT04299581	2	PD-1, ablation	Camrelizumab, Cryoablation	a(i)CCA	II	ORR	DoR, PFS, OS, DCR, AEs	Asia
NCT02821754	2	PD-1, CTLA-4, ablative therapies	Durvalumab, Tremelimumab, TACE, RFA, SIRT, Cryoablation	aBTC, HCC	II+	PFS	Safety	US
IO+G-CSF
NCT02703714	2	PD-1, GM-CSF	Pembrolizumab, Sagramostim	aBTC	I/II	ORR	Safety, PD-L1 positive status, PFS, DCR, OS	US

HAIC = Hepatic arterial infusion chemotherapy, PC = pancreatic cancer, irRC = immune-related Response Criteria

In addition to radiotherapy, local ablation is frequently used to treat small liver tumors of up to 3–5 cm in size. Two pilot studies demonstrated the feasibility, safety, and efficacy of the combination of tremelimumab and microwave ablation or radiofrequency ablation in patients with advanced BTC and HCC [26,27]. Increased accumulation of intratumoral CD8 + T-cells in the HCC cohort and an increased number of circulating activated CD8 + T-cells as well as an expansion of the TCR repertoire in the BTC cohort induced by tremelimumab may have contributed to the treatment benefit.

Overall, these early data support the concept of ongoing studies that evaluate the combination of immune therapy with local therapies (Table 3).

5. Biomarkers

In BTC, the initial enthusiasm about IO was dampened by the observation that only few patients derive a durable clinical benefit from checkpoint inhibition. Therefore, an active search for predictive biomarkers is under way. Several studies attempted to correlate clinical response to the presence or absence of candidate markers established in other solid tumor entities, such as PD-L1 expression, MSI status, or TMB.

5.1. PD-L1 expression

Expression of the PD-L1 protein was the first biomarker (Table 4) that was associated with response to anti – PD-1 therapy in several tumor entities, and multiple PD-L1 immunohistochemistry assays have been developed as a companion or complementary diagnostic for patients with lung cancer, melanoma or gastric cancer to help select patients for immune therapy [28]. However, the predictive value of PD-L1 expression in certain tumor types does not necessarily mean that it can be used as a marker to distinguish responders and non-responders in other entities. Major obstacles for the reliable implementation of PD-L1 testing in daily clinical use is the availability of multiple assays, the divergent predictive significance of tumor versus immune cell expression in different tumor types, the insufficiently defined thresholds for expression levels, and the not yet fully established scoring system for pathologists, with a poor inter-observer reproducibility. Accordingly, a recent systematic review and meta-analysis involving tumor specimens from more than 8,000 patients in multiple solid tumors including melanoma, NSCLC, urothelial carcinoma and GI tumors revealed that PD-L1 immunohistochemistry (IHC) has a poor ability to predict efficacy of IO therapy [29]. In contrast, multiplex immunohistochemistry/immunofluorescence (mIHC/IF) enables the simultaneous detection of multiple markers, such as PD-L1, PD-1, and CD8. Thereby, specific cellular subsets can be identified and assessed under consideration of their spatial relationships and their localization within different tumor compartments. Overall, this retrospective analysis suggests that mIHC/IF outperforms the use of uni-dimensional biomarkers such as PD-L1 IHC or TMB alone in their ability to predict response to immune therapies.

Table 4. Efficacy data in phase-II trials according to PD-L1 expression

Study	n	Prior lines	PD-L1	ORR	PFS	OS
NCT02628067 KN-158 Pembrolizumab	104	52% 3 L	61 pos	6.6%	2.1	9.3
			34 neg	2.9%	1.9	7.2
JapicCTI-153,098 Nivolumab	30	53% 3 L	5 pos.	20%	2.8	11.6
			23. neg.	0%	1.4	5.2
NCT02054806 KN-028 Pembrolizumab	24	50% 3 L+	24 pos.	13%	1.8	6.2
NCT02829918 Nivolumab	45	56% 2 L	13 pos.	46%	10.5	n.r.
			15 neg.	6%	2.3	10.3

PD-L1 expression in BTC is highly heterogeneous and the prognostic and predictive value is not yet completely clear[30]. Various clinicopathological variables have been associated with PD-L1 positivity, including venous invasion, extrahepatic metastases, tumor grade, clinical stage, and survival, even though the results remain controversial. A comprehensive meta-analysis of 15 independent studies with 1.776 patients concluded that high PD-L1 expression was moderately associated with poor overall survival (HR 1.79, 95% CI 1.55–2.07, p < 0.001), but not with other outcome parameters in patients with BTC [31].

One of the first prospective studies on IO in BTC that is currently awaiting data maturation, the KEYNOTE-028 study (NCT02054806), evaluates the efficacy of pembrolizumab in 24 patients with PD-L1-positive BTC. Positivity was defined as membranous PD-L1 expression in ≥1% of tumor and associated inflammatory cells or positive staining in the stroma [32]. The ORR was 13.0% (3/23, all PR; 95% CI 2.8%-33.6%), mPFS was 1.8 months (95% CI 1.4–3.1) and mOS was 5.7 months (95% CI 3.1–9.8) [8]. In contrast, the KEYNOTE-158 study with 104 patients and 58.7% PD-L1-positive of cases showed an ORR of only 5.8% [33]. Among 95 patients evaluable for PD-L1 expression, ORR was 6.6% (4/61, all PR; 95% CI 1.8%-15.9%) and mOS was 7.2 months (95% CI 3.7–10.8) in patients with PD-L1-expressing tumors and 2.9% (1/34, PR; 95% CI 0.1%-15.3%) and 9.3 months (95% CI 4.2–11.5) in patients with PD-L1 negative tumors.

The value of PD-L1 expression as a biomarker has also been the topic of several retrospective analyses and small single-center trials [34,35]. Overall, these reports, which used different assays and cut offs for PD-L1 expression, suggest a higher efficacy of checkpoint inhibitors in PD-L1 positive BTC, but additional prospective studies are required to better define the predictive value of PD-L1 expression on tumor and/or immune cells not only for response to IO therapy, but also for survival.

5.2. MSI

MSI is a molecular marker of defective DNA mismatch repair that can arise from either germline mutations in components of the mismatch repair (MMR) machinery (MSH2, MSH6, MLH1, PMS2) in patients with Lynch syndrome, or following epigenetic silencing of the MLH1 promoter. Diagnostic guidelines relating to MSI status have evolved over the past decades, and a combination of immunohistochemistry for MMR proteins (5% to 11% of MSI cases may however not show MMR protein loss) and a PCR-based assay with a highly sensitive mononucleotide marker panel (BAT25, BAT26, CAT25), which specifically detects DNA mismatch repair deficiency, are currently regarded as standard of care. Of note, very recently it has been shown that artificial intelligence-based pattern recognition can predict MSI directly from H&E histology, which might enable an even more efficient identification of MSI tumors in the future [36].

Whole-exome sequencing of tumor tissue revealed an average of 1,782 somatic mutations in cancers with MSI versus 73 somatic mutations in cancers without MSI [37]. Several studies have subsequently shown that increased TMB in MSI tumors creates neoepitopes that can be recognized by the immune response and that are potential targets for an efficient immune response. Importantly, tumors with MSI display a significant upregulation of immune checkpoint proteins, including PD-1 and PD-L1, on tumor and immune cells, allowing them to escape from immune-mediated clearance. In 2015, Le and colleagues reported highly favorable data for anti–PD-1 inhibitors in patients with MSI after failure of conventional therapies. The activity of IO therapy in MSI patients has subsequently been confirmed in multiple studies independent of tumor type and drug used, and established MSI as a pivotal predictive biomarker for immunotherapy [37,38].

Beyond its value as a biomarker for IO therapy, MSI status might have additional prognostic and predictive implications. MSI colorectal cancer for example is associated with a lower stage at diagnosis and improved prognosis in earlier stages, most likely resulting from a significant immunologic response elicited by neoepitopes. In addition, emerging data suggest that MSI gastric and colorectal tumors do not benefit from chemotherapy specifically in the peri-operative setting [39,40].

In BTC, MSI status is rare (less than 1% of patients) but is currently the only reliable marker that appears to be predictive of clinical response to immunotherapy [41,42]. Robust anti-tumor activity was shown with pembrolizumab in the phase-II KEYNOTE-158 study in previously treated advanced non-colorectal MSI/dMMR cancer. Overall, ORR was 34.3% for the 27 different tumor types included in the study, and approximately one-third of patients with an objective response had a complete response. From the 22 MSI/dMMR BTC patients, 40.9% showed a clinical response with an mPFS of 4.2 months, and an mOS of 24.3 months [43]. The results are similar to the efficacy in colorectal cancer and appear to be better than in pancreatic cancer. Based on the trial results, pembrolizumab monotherapy was approved in several countries, including the USA, Japan, and Australia, for use in patients with previously treated, MSI, advanced solid tumors.

Despite the high tumor immunogenicity, however, not all MSI patients uniformly respond to immune therapy, and roughly half are refractory to monotherapy with checkpoint inhibitors, indicating that a more granular understanding of the mechanisms underlying IO sensitivity in these tumors is required. Recently, it has been shown that the degree of MSI, the extent of TMB, and particularly the associated accumulation of exonic frameshifting insertion-deletion mutations strongly correlate with the immune infiltration and the response to immunotherapy in preclinical models as well as MSI GI-cancer patients [44,45]. Of note, response to IO in MSI patients might be further enhanced by dual checkpoint inhibition. Data from the Checkmate-142 study indicate an even higher efficacy with a combination of nivolumab and ipilimumab in MSI colorectal cancer (CRC) patients with respect to ORR and mOS compared to a monotherapy with either nivolumab or pembrolizumab [46–48]. Additional studies should therefore also evaluate dual checkpoint inhibition in non-CRC MSI patients, including BTC patients.

5.3. TMB

TMB defined as the total number of somatic mutations per coding area of a tumor genome, is an emerging clinical biomarker for IO therapy. TMB can be quantified by a number of NGS-based sequencing technologies. While whole-exome sequencing is considered the gold standard, targeted gene panels of various genomic sizes are frequently used for TMB assessment in the clinical setting. Of note however, the lack of harmonization in TMB quantification, of adequate methods to compare TMB estimates across different panels, and of robust predictive cutoffs, currently challenge the implementation of TMB as a reliable biomarker in clinical practice [49].

Because most IO-based phase-II studies are single-arm studies, the prognostic impact of high TMB is not yet clear. A recent analysis evaluated the prevalence of TMB^high status and its potential association with OS in a real-world data set [50]. Nearly 3,000 patients with 10 different tumor types, including BTC, with available panel sequencing data (Foundation Medicine CDx®), and without prior immunotherapy treatments, were evaluated. Prevalence of TMB^high was highest in patients with SCLC, NET and cervical cancer, whereas prevalence of TMB^high was lowest in patients with mesothelioma, thyroid cancer and BTC; only 28 out of 706 BTC patients (4%) belonged to the TMB^high group. Overall, mOS across all cancer types was similar among patients with TMB^low and TMB^high with mOS of 8.4 months and 10.5 months in the respective groups, suggesting that TMB is not a strong prognostic biomarker in cancer. In addition to the marked differences in the TMB^high prevalence across the different tumors, the median TMB varied markedly within the different entities. Median TMB was 2.6 for BTC patients compared to 8.7 in lung cancer. A universal cutoff for ‘TMB^high’ would therefore enrich for tumor entities with higher mutation load, and stratifying tumors by selecting the highest mutation load quintile (top 20%) per histology might be better suited to correlate TMB with clinical outcome data. The varying distributions of TMB were recently further confirmed in a study across tumor histologies, suggesting that TMB thresholds need to be defined on a per-entity basis, and that the optimal threshold to identify patients for IO is likely to vary for different cancers [51].

In line with the observation that highly mutated tumors can produce many tumor-specific neoantigens, some of which can be recognized by the immune system, high TMB has been shown to correlate with clinical benefit of antibodies targeted against CTLA-4, PD-1, and PD-L1 in multiple tumors including melanoma and NSCLC [52,53]. In an exploratory analysis of the aforementioned KEYNOTE-158 study, the association between TMB and clinical outcomes with pembrolizumab monotherapy across 10 different tumor types was analyzed to determine whether TMB can be used as a tumor-agnostic biomarker [54]. In this study, a cutoff of at least 10 mutations per megabase defined a TMB^high tumor on the basis of the FoundationOne CDx® assay (Foundation Medicine, Cambridge, MA, USA). Of the 805 patients, 13% had TMB^high status, most of them with lung and cervical cancer and none with BTC. In the efficacy population, 29% of the patients in the TMB^high group and 6% in the non-TMB^high group had an objective response, and the median DOR was not yet reached. Despite the significantly higher response rate, mPFS was 2.1 months in both the TMB^high and non-TMB^high groups and there was only a minor difference in the overall survival when stratified according to TMB status, which might be complicated by confounding factors such as the inclusion of multiple tumor types. Of note, the predictive value appears to be independent of other biomarkers such as PD-L1 expression or tumor type. Together, these data indicate that TMB^high status can identify a subgroup of patients that are more likely to derive a clinically meaningful benefit in terms of tumor response to IO monotherapy, but that this subgroup appears to be a very small in BTC.

5.4. Genetic alterations

As discussed above, BTC frequently harbors potentially druggable genetic alterations and there is increasing preclinical evidence that genetic alterations can affect PD-L1 expression. Accordingly, a recent study found a statistically significant association between PD-L1 expression and mutations in BRAF, BRCA2, RNF43, and TP53 [55].

The association with mutations in the DNA damage repair pathway is of particular interest as these tumors might be more prone to accumulate mutations, and subsequently increase neoantigen release, making them promising candidates for IO. Of note, patients with DNA damage repair deficiencies might benefit from a combination of PARP inhibitors and immune therapy, a concept that has already demonstrated efficacy predominantly in BRCA-mutated tumors including breast, ovarian, and prostate cancer. Ongoing trials currently evaluate this concept also in BTC (Table 5).

Table 5. Ongoing immunotherapy trials in BTC combined with targeted therapies

Trial number	Phase	Pathways targeted	Drug	Indication	Line of treatment	Primary outcomes	Secondary outcomes	Country
NCT04298021	2	PD-1, ATR, PARP	Durvalumab, AZD6738, Olaparib	aBTC	II	DCR	ORR, DoR, OS, safety and tolerability, QoL, PFS	Asia
NCT04306367	2	PD-1, PARP	Pembrolizumab, Olaparib	aCCA	II	ORR	DoR, PDS, OS, safety and tolerability	US
NCT03991832	2	PD-1, PARP	Durvalumab, Olaparib	Solid tumors (IDH-mutated)	I/II+	ORR, DCR	PFS, OS, AEs	Canada
NCT04298021	2	DDR, PD-1, PARP	AZD6738, Durvalumab, Olaparib	aBTC	II	DCR	ORR, PFS, DoR, OS, safety and tolerability, QoL	Asia
NCT03684811	1/2	IDH1, PD-1, chemotherapy	FT-2102, Nivolumab, Gemcitabine, Cisplatin	aBTC, HCC (IDH1-mutated)	II	DLT, dose recommendation, ORR	Pharmacokinetic, AEs, PFS, TTP, DOR, OS, TTR	US
NCT03639935	2	PD-1, PARP	Nivolumab/Rucaparib	aBTC	II	4-month survival, response	CR, PR, SD, PFS, OS	US
NCT03201458	2	PD-1, MEK	Atezolizumab, Cobimetinib	aCCA	II/III	PFS	AEs, OS, ORR, change in CD8+	US

TTR = time to response

Further evidence exists that activation of the Raf- MEK-ERK pathway in the tumor can lead to an immunosuppressive tumor microenvironment. Preclinical data in a melanoma model suggest that BRAF/MEK inhibition increases T-cell recruitment into the tumors and thereby improves therapeutic efficacy in combination with immunotherapy [56]. Nevertheless, none of the recurrent genetic alterations in BTC has been validated as a marker of resistance to immunotherapy yet. Potential candidates, however, include mutations leading to an activation of tumor-intrinsic WNT/b-catenin signaling, which are enriched in non-T-cell-inflamed tumors across multiple entities [57], or mutations in STK11, that are associated with a lack of clinical response to IO in lung cancer [58].

Currently, FGFR2 fusions are the most advanced and promising genetic targets for molecular therapies, together with therapies directed against HER2 and IDH1. Positive signals from first trials in solid cancers that investigate combinations of precision- and immune-oncology approaches have recently been reported, for instance in gastric cancer. A first-line triplet regimen including pembrolizumab, trastuzumab, and chemotherapy showed impressive clinical activity with a DCR of 100%, an mPFS of 13.0 months, and an mOS of 27.3 months, regardless of PD-L1 status [59]. Consequently, the concept is now further developed in the phase-III KEYNOTE-811 study. A recent study revealed that ERBB2/3 mutations were associated with the upregulation of PD-L1 and suppressed anti-cancer immunity in a model of gallbladder cancer. Either PD-L1 blockade or anti-ERBB2/3 therapies reversed these immunosuppressive effects, and combined therapy further enhanced therapeutic efficacy [60]. Considering that activating mutations or amplifications of ERBB2/3 can be found in 7–20% of BTC and gallbladder patients, and are associated with a poor survival in BTC, this combination might also become an interesting option in BTC.

With respect to FGFR2 fusions, recent preclinical studies suggest that FGFR inhibition impacts the tumor immune microenvironment. Erdafitinib, a pan-FGFR small molecule inhibitor, led to a survival advantage compared to the respective monotherapies when combined with PD-1 blockade in an FGFR2-driven murine autochthonous lung cancer model [61]. The synergistic antitumor effect of this combination was dependent on erdafitinib-induced tumor cell killing, de novo priming, and enhancement of antitumor T-cell responses via PD-1 blockade. More recently, it was shown that erdafitinib treatment drives T-cell infiltration and causes de novo priming and broadening of the TCR repertoire, likely via an indirect mechanism that depends on tumor cell killing. The addition of PD-1 blockade to erdafitinib treatment led to focusing of the T-cell repertoire through expansion of specific T-cell clones that were likely critical to induce productive antitumor immune responses [61]. The ADVANCE phase-II study currently evaluates the efficacy of the anti- PD-L1 antibody atezolizumab in combination with the FGFR inhibitor derazantinib in patients with advanced BTC with FGFR2 fusions/rearrangements (EudraCT 2020–004938-38).

6. Conclusion

In summary, there seems to be clinically significant activity with checkpoint inhibition in patients with BTC, but only few patients derive a convincing benefit. Similar to mono-immunotherapy, the clinical efficacy of dual checkpoint inhibition appears to be inferior in BTC when compared to other tumor types. Apart from combining PD-1/PD-L1 inhibitors with CTLA4-antagonists, the combination of immunotherapy with targeted therapies and local treatments is pursued in ongoing studies. With respect to biomarkers, only the MSI status currently appears to be sufficiently validated to select patients for immunotherapy, but the prevalence is below 1% in BTC. Similarly, the frequency of TMB^high tumors is very low in this tumor type and – in contrast to MSI status – the available data do not yet allow a definitive conclusion concerning its potential use as a predictive biomarker for IO-based therapies. PD-L1 expression on tumor and immune cells has been reported with variable frequencies in BTC, but again the data published do not yet support the use of PD-L1 as reliable predictive biomarker for treatment selection.

7. Expert opinion

BTCs are highly aggressive tumors with a very poor prognosis. After several years without significant progress, however, a number of important studies were successfully completed or have been initiated that significantly impact the use of systemic therapies already today. Despite the fact, that BTC is a chemo-resistant tumor (and that more data from ongoing studies are still required), we now have sufficient evidence to consider adjuvant therapy with capecitabine based on the BILCAP study and to recommend 2^nd line therapy based on the ABC-06 study [62,63]. With respect to targeted therapies, we must conclude that all studies in unselected BTC cohorts were negative, including several trials evaluating the efficacy of anti-angiogenetic agents [3,64]. Similarly, anti-epidermal growth factor receptor (EGFR) antibodies and tyrosine kinase inhibitors targeting the EGRF-RAS-MEK pathway failed to show any clinical benefit in the 1^st and 2^nd line setting [65–67]. In recent years, the detailed characterization of the molecular landscape of biliary tumors has revealed that approximately 20–40% of patients harbor genetic alterations, which are potential targets for molecular therapies [42,68,69]. The better understanding of the mutational spectrum in BTC and the implementation of clinical trials in genetically defined subgroups helped to realize the promise of targeted therapies in this tumor type. Based on the published data and the ongoing clinical trials, we strongly recommend molecular testing using diagnostic assays that are capable of detecting BRAF and IDH1/2 mutations, FGFR2- and NTRK fusions, HER2/neu amplifications and mutations, as well as MSI, at an early stage in the palliative setting to ensure that clinically relevant therapeutic options are not withheld from BTC patients. In order to further expand the spectrum of evidence-based therapies in the near future, it is also important to systematically screen BTC patients for their eligibility to be included in ongoing clinical trials.

To exploit the full potential of immunotherapy in BTC, a better understanding of the potential mechanisms underlying primary resistance or sensitivity will be important to develop rational approaches and guide the design of meaningful trials in this rare cancer entity. Biomarkers are needed to define the molecular subgroups that are most likely to benefit from immunotherapy, or from IO-based combinations.

The data published so far strongly suggest that BTC is less sensitive to checkpoint inhibition than HCC, the most frequent primary liver cancer. Even in HCC, monotherapy approaches have failed despite promising signals in single-arm phase-II studies, and only the combination of a checkpoint inhibitor with a targeted agent has shown superior efficacy over the previous standard of care, sorafenib [70–74]. Despite the lack of clinically established molecular biomarkers, systemic therapy in HCC is dominated by tyrosine kinase inhibitors (TKI), which might explain to some extent the large number of clinical trials evaluating IO/TKI combinations. As a note of caution , we still need to acknowledge that it is not yet completely clear to which extent the IO/TKI combinations display truly synergistic activity in cancer or whether they are more the result of additive activity of the respective agents. In contrast to HCC, systemic treatment in BTC is dominated by chemotherapy, which might explain why all phase-III studies which evaluate IO-based combinations in BTC include a chemotherapy backbone. Of note however, the combination of chemotherapy with checkpoint inhibitors has so far failed in most phase-III studies in gastrointestinal (GI) oncology in all-comer populations and only the most recent data in gastric cancer provided evidence for a clinical meaningful benefit in combined positive score (CPS) positive patients [59,75–77]. Data from phase-II studies in Asian countries are very promising, but due to the high genetic heterogeneity in BTC it remains to be seen whether the ongoing global phase-III studies will validate these findings. Unfortunately, the most promising biomarkers such as MSI and TMB^high have a very low prevalence in BTC, which makes their validation in prospective, randomized trials challenging considering the overall low incidence of BTC. The signal of PD-L1 expression as a positive predictive biomarker is moderate at best, but a systematic evaluation of its expression in tumor and immune cells within the ongoing clinical trials is strongly encouraged. At the moment, a routine testing of PD-L1 expression to guide treatment decisions cannot be recommended.

In addition to checkpoint inhibition, adoptive cell therapies utilizing T-cells expressing chimeric antigen receptors (CARs) have achieved convincing responses in B-cell hematologic malignancies and several studies have also shown the feasibility of this approach in solid tumors, including BTC [78]. However, the results in early phase clinical studies so far demonstrated only limited anti-tumor efficacy despite targeting a variety of target antigens and tumor types. The disappointing results are most likely multifactorial and include a limited number of targetable antigens and heterogeneous antigen expression, as well as inefficient homing to and penetration of the immunosuppressive tumor microenvironment [79].

Overall, immunotherapy should currently not be considered as a preferred systemic treatment outside of clinical trials in MSS BTC patients. However, there is a clear signal that a subgroup of BTC patients can derive a benefit from IO-based systemic therapies and we strongly encourage inclusion of patients in the ongoing clinical trials, specifically those with a broad translational program.

Article highlights

• Cancers of the biliary system (BTC) are highly aggressive tumors with a dismal prognosis and only a few established options for systemic therapy.

• In recent years, treatment concepts for BTC were genuinely influenced by two novel ‘approaches’ to anti-cancer therapy, namely precision oncology and immune oncology (IO).

• The initial enthusiasm about immunotherapy in BTC was dampened by the observation that only few patients derive a durable clinical benefit from checkpoint inhibition. However, there is a clear signal that a subgroup of BTC patients benefits from IO-based systemic therapies.

• The most promising data on IO in BTC have been reported in phase-II trials in combination with chemotherapy.

• MSI and TMBhigh, as well as high PD-L1 expression, might serve as valuable biomarkers, but their prevalence is very low in BTC.

• Inclusion of BTC patients into ongoing clinical trials is strongly encouraged.

This box summarizes key points contained in the article.

Reviewer disclosures

One reviewer has received honoraria from Ono Pharmaceutical.

Peer reviewers on this manuscript have no other relevant financial or other relationships to disclose.

Declaration of interest

Honoraria for speaker, consultancy, and advisory roles was awarded to A Vogel from Roche, Bayer, Sanofi, BMS, Lilly, Novartis, EISAI, AstraZeneca, Merck, Medac, Ipsen, PierreFabre and MSD.

The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Additional information

Funding

A Vogel and A Saborowski have received funding from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – [SFB/TRR 209 - 314905040, and Vo959/9-1 (to AV)]. A Vogel is supported by the European-Latin- American ESCALON consortium, funded by the EU Horizon2020 program. A Saborowski is supported by Deutsche Krebshilfe (DK) −70114101.