Key challenges for drugs in clinical development for cholangiocarcinoma

1. Introduction: current landscape of advanced cholangiocarcinoma treatment

Cholangiocarcinoma (CCA) is a highly morbid cancer with rising incidence and few approved standard therapies. The main anatomical sites of cholangiocarcinoma are intra-hepatic (iCCA), perihilar, and distal, with the latter two classes comprising extra-hepatic (eCCA) cholangiocarcinoma. While CCA is an uncommon cancer overall, with an incidence of 0.3–6 per 100,000 people, the mortality attributed to CCA, particularly iCCA, is rising globally [1–4]. Most patients present with or progress to advanced, incurable disease and require systemic therapy. First-line therapy for advanced cholangiocarcinoma consists of the combination of cisplatin and gemcitabine, based on the ABC-02 trial which showed median overall survival of 11.7 months for the combination by comparison to 8.1 months for gemcitabine alone (HR 0.64, p < 0.001) [5]. A recent single-arm phase II study of a triplet combination of gemcitabine, cisplatin, and nab-paclitaxel achieved objective responses in 45% with median progression-free survival and overall survival of 11.8 and 19.2 months, respectively [6]. These encouraging results have prompted an ongoing randomized, phase III trial evaluating the efficacy of this taxane-based combination versus standard gemcitabine plus cisplatin (NCT03768414). Following progression on gemcitabine plus cisplatin, a variety of other chemotherapy regimens had been previously used without prospective, randomized trial data [7]. In a phase 3 trial, the combination of fluorouracil (5-FU) and oxaliplatin (FOLFOX) plus active symptom control improved survival over active symptom control alone, though extent of improvement was modest (median overall survival of 6.2 versus 5.3 months, HR 0.69, p = 0.031) [8].

Targeted therapies are emerging as an important therapeutic option for subsets of patients based upon the prevalence of potentially actionable mutations in cholangiocarcinoma (table 1). Activating mutations in the fibroblast growth factor receptor (FGFR) pathway, most commonly FGFR2 fusions or other rearrangements, are present in approximately 20% of iCCA and have shown robust and durable responses to targeted FGFR inhibition [9,10]. Pemigatinib, an ATP-competitive inhibitor of FGFR1, FGFR2, and FGFR3, achieved objective responses in 35.5% of patients with cholangiocarcinoma harboring FGFR2 fusions or rearrangement after failure of standard chemotherapy [11], leading to the first FDA approval for a targeted therapy in a genomically defined subset of cholangiocarcinoma. Other FGFR inhibitors with demonstrated activity in phase II studies of FGFR2-translocated cholangiocarcinomas include the pan-FGFR inhibitor, infigratinib (NCT# 02150967), and the covalent pan-FGFR inhibitor, futibatinib (NCT# 02052778). IDH1 mutations, present in approximately 13% of iCCA [12], are another therapeutic target. In a phase III study (ClarIDHy), the IDH1 inhibitor, ivosidenib, demonstrated significantly prolonged progression-free survival over placebo [13]. Additional mutations which have shown response to targeted therapies in cholangiocarcinoma include BRAF V600E mutations, NTRK fusions, and activating ERBB2 mutations or amplification [14–16].

Table 1. Actionable mutations in advanced cholangiocarcinoma and associations with anatomic sub-type

Molecular target	Prevalence in cholangiocarcinoma	Prevalence by anatomic sub-type
FGFRfusionsrearrangements	8–14% [9,14]	5.5–13% iCCA, 0–5% eCCA [9,14,17,18]
IDH1, IDH2	13% (IDH1), 3% (IDH2) [12]	5.8–29% iCCA, 0–5% eCCA (IDH1) [9,12,14,17]; 1.7% iCCA, 0.6% eCCA (IDH2) [23]
BRAF V600E	3–5% [1]	3.4–5% iCCA, 2.4–3% eCCA [18,23]
ERBB2	3.9% to 8.5% [23]	1.8–3% iCCA, 11–20% eCCA [17]
KRAS	11–25% [1,3,9]	7–24% iCCA, 10–42% eCCA [9,14,17,18,23]
BAP1, BRCA1/2, other DDR gene mutations	8–20% (BAP1), lower frequency of other individual DDR gene mutations [9,14]	9.1–19% iCCA, 0–10% eCCA (BAP1) [9,14,17,23]; 1.0% iCCA, 2.4% eCCA (BRCA1) [23]; 3.1% iCCA, 4.2% eCCA [23]
NTRK	4% [16]	Not reported [16]
MSI-high/dMMR	5–10% [3]	10% iCCA, 5% eCCA [3]

*Abbreviations: DDR = DNA damage repair, MSI = microsatellite instability, dMMR = deficient mismatch repair

Though immune checkpoint inhibition (CPI) has demonstrated activity in a variety of solid tumor types, clinical trials in cholangiocarcinoma have shown generally low objective response rates in unselected populations of patients with cholangiocarcinoma, ranging from 5.8% to 13% in the Keynote-158 and Keynote-028 studies of pembrolizumab and in a phase II study of nivolumab [19,20]. For patients with microsatellite instability (MSI) or high tumor mutational burden (TMB) across tumor types including cholangiocarcinoma, CPI shows much higher rates of response, leading to a tumor-agnostic FDA approval for pembrolizumab in tumors of any histology with these features [21,22]. Ongoing clinical trials are examining the potential to augment the immune response to CPI by combination with other immunotherapies, chemotherapy, or locally ablative therapies.

2. Challenges to CCA clinical trial design and enrollment

There are numerous challenges to drug development in cholangiocarcinoma, stemming from its complex anatomy and heterogeneous biology, and amplified by the rarity of individual tumor subgroups.

2.1. Difficulties in diagnosis and response assessment

Cholangiocarcinoma can be difficult to diagnose due to its anatomical location within the biliary tract. The complexity of the biliary tract anatomy can obfuscate diagnosis, with some extrahepatic tumors infiltrating the bile duct without a discrete mass apparent on imaging, while intrahepatic lesions can be misdiagnosed as metastases or cancer of unknown primary [2]. Delays in diagnosis are common owing to the difficulty of radiographic determination of site of primary tumor or inaccessible or limited tumor tissue for pathologic diagnosis. Tumors in the biliary tract can be difficult to access for biopsy, with histology often limited to fine needle aspirates or cytology brushings from endobiliary procedures.

Beyond diagnostic delays, the scarcity of adequate tumor tissue samples due to frequent inaccessibility within the biliary tract also poses a challenge to molecular profiling. Fine needle aspiration or cytology brushings often yield insufficient tumor content for complete molecular profiling, which can lead to treatment delays if a repeat biopsy is required and can contribute to underpowered biomarker analyses in clinical trials. The challenge of insufficient tumor tissue underscores the importance of employing broad molecular profiling panels to test as many potential therapeutic targets as possible with a single small sample, as well as validating circulating tumor DNA and other noninvasive molecular profiling strategies as a surrogate for tumor tissue.

Depending on anatomic location within the biliary tract, some cholangiocarcinoma tumors, particularly eCCA, may not be readily measurable by standard Response Evaluation Criteria in Solid Tumors (RECIST) v.1.1, interfering with measurement of standard clinical endpoints in clinical trials such as objective response rate and progression-free survival. Variability in objective response rates according to anatomic subsite, as was observed in the ABC-02 trial, may reflect these differences in measurability by RECIST 1.1, though differing biologic susceptibility could also contribute [5].

2.2. Competing comorbidity and disease complications

The complex biliary tract anatomy also can lead to comorbidities and complications that can limit treatment options, such as infection, biliary obstruction, or hepatic dysfunction, and can introduce competing causes of mortality. Prior surgeries for resectable disease that has recurred can be accompanied by comorbid complications, such as pancreatic insufficiency or diabetes after a Whipple procedure for distal cholangiocarcinomas.

The etiology of cholangiocarcinoma in the majority of cases is unknown, though a subset of cases are associated with underlying chronic inflammatory processes including hepatitis B or C virus, liver fluke infections, primary sclerosing cholangitis, and nonalcoholic fatty liver disease (NAFLD). Furthermore, the etiology of cholangiocarcinoma impacts the extent of underlying liver dysfunction and other co-morbidities, such as immunosuppression, pancytopenia, hepatic dysfunction, or cirrhosis. While there are available therapeutic options for some causes of hepatitis, damage to the liver at the time of cancer diagnosis can be irreversible and complicate treatment options, as many therapies have a paucity of safety data in patients with advanced liver dysfunction. Collectively, these prevalent competing comorbidities contribute to poor prognosis, limiting the window of time a patient may have to achieve treatment response and confounding interpretation of time-to-event outcomes in clinical trials.

2.3. Heterogeneous biology and rarity of individual subgroups

Cholangiocarcinoma biology is heterogeneous between individuals and within individual tumors, and differences in disease biology can be associated with anatomic subsite, etiology, and prognosis [3,23]. The prevalence of specific targetable mutations is associated with the anatomic subsite of tumor, particularly FGFR2 fusions and IDH1 mutations which are nearly exclusively present in iCCA and extremely rare in eCCA, while ERBB2 amplifications may be more common in eCCA [14,23] (table 1). Four biological sub-groups of cholangiocarcinoma were defined in a study of liver fluke (Opisthorchis viverrini and Clonorchis sinensis) and non-fluke-related cases [23]. Exposure to liver fluke, a significant risk factor for cholangiocarcinoma in certain parts of the world, was associated with tumors enriched for ERBB2 amplifications and TP53 mutations, whereas two sub-types not associated with fluke infection were found to have higher incidence of BAP1, FGFR, and IDH1/2 alterations. The distinct gene expression profiles of mutation-defined subgroups of CCA highlight the potential for differences in response to standard cytotoxic therapies as well as to molecularly targeted therapies, and introduces the potential for confounding in unselected clinical trials. The rarity of individual subgroups, however, poses an enormous challenge to development of biomarker-selected trials in an uncommon tumor type.

Cholangiocarcinoma exists in a microenvironment characterized by a reactive stroma containing both immune and nonimmune cell types such as fibroblasts and endothelial cells [1]. In the majority of iCCA, the tumor microenvironment has ‘immune desert’ or mesenchymal features which are associated with poor prognosis, while only a subset demonstrates T-cell infiltration and immune-reactivity [24]. While the desmoplastic stroma can physically impair the delivery of anti-cancer therapies, it also supports the recruitment and polarization of stromal cells with a pro-tumorigenic phenotype such as cancer-associated fibroblasts and myeloid-derived suppressor cells [3,25]. Cross-talk between tumor cells and stromal cells can enhance oncogenic signaling pathways, promote tumor growth and metastasis, and inhibit effector T cell responses directly and by the secretion of suppressive cytokines [1,3]. Furthermore, gut microbial populations can interact with both immune and tumor cells with the tumor microenvironment to promote pro- or anti-cancer responses [26]. Targeting the microbiome, suppressive immune cells such as tumor-associated macrophages, or stromal elements all represent possible avenues forward in tumors without driver mutations or with upregulation of suppressive immune signaling, factors which may contribute to inherent therapeutic resistance.

3. Conclusions

Many of the challenges to drug development in CCA, including clinical and genetic heterogeneity, comorbidity, and tumor tissue scarcity, can be mitigated by cholangiocarcinoma-specific clinical trial eligibility, stratification, and design approaches as well as a broadening the biomarker platforms used for clinical trial eligibility.

4. Expert opinion

A variety of strategies are needed to overcome these challenges to drug development in cholangiocarcinoma, including clinical trial eligibility and stratification optimized for the target population and mechanism of action of the investigational therapy; development of multicenter registries to develop natural history data for rare subgroups within CCA; and broadening the biomarker platforms used in CCA clinical trials.

4.1. Clinical trial eligibility and stratification

Clinical trial design approaches can help to mitigate the challenges of comorbidity and complex anatomy in cholangiocarcinoma. Due to the rarity of cholangiocarcinoma overall, subgroups within CCA defined by clinical or molecular features (table 1) are comprised of even smaller numbers of patients, limiting the feasibility of biomarker-enriched or selected clinical trial designs. Conversely, all-comer trials without selection according to clinical or molecular features risk confounding by prognostic heterogeneity and potential for differential response.

Depending on the mechanism of action of the investigational therapy under study, disease-specific eligibility criteria can help to encompass the appropriate cholangiocarcinoma population while minimizing heterogeneity and risk (table 2). In general, cytotoxic therapies expected to have activity across subgroups are best suited to randomized, unselected clinical trials with stratification for key prognostic covariates [3] (table 3), while molecularly targeted therapies may require selection for the presence of the target under study if activity is limited to subgroups with presence of that molecular target.

Table 2. Selected clinical trial eligibility criteria considerations in advanced cholangiocarcinoma

Eligibility Criterion	Considerations in CCA Populations and Trials
Tumor factors
Presence of a specific molecular target	Integral biomarker positivity may be required for clinical trials of molecularly-targeted therapy with selection or enrichment for the specific molecular target. In non-selected clinical trials, molecular profiling from standard of care profiling should be collected for secondary analyses when available due to potential for prognostic impact.
Anatomic subsite of primary tumor	Anatomic location of CCA generally should not be an eligibility requirement unless in cases of molecularly-targeted therapies with an integral biomarker present nearly exclusively in a specific anatomic location (e.g. iCCA). Anatomic subsite should be considered as a stratification factor owing to potential prognostic impact (see Table 3).
Fresh or archival tumor tissue for central analysis	Archival and fresh tumor tissue samples are often scarce/insufficient for biomarker analyses in CCA. A requirement for submission of archival tissue can be a barrier to enrollment and introduces selection bias against tumors in inaccessible locations, particularly for eCCA which may be more difficult to access.
Measurable disease by RECIST 1.1 versus evaluable disease	Measurable disease is necessary for trials with objective response rate as a primary endpoint, while evaluable disease may be more appropriate for time-to-event primary endpoint trials in CCA. Requirement for measurability by RECIST 1.1 can create a selection bias against certain anatomic subsites or patterns of recurrence (e.g. locally advanced eCCA with porta hepatis soft tissue masses or CCA with peritoneal metastases may not be measurable).
Comorbidity Factors
Presence of endobiliary stents or percutaneous drains for biliary obstruction	Presence of endobiliary stents or percutaneous drains generally should not limit eligibility for most clinical trials, provided there is adequate biliary decompression based upon liver function test parameters and no evidence of active infection at enrollment. In clinical trials of agents such as cellular therapies or combination cytotoxic therapies with high risk for immunosuppression, the presence of stent(s) or biliary drains may warrant exclusion owing to higher risk for infectious complications such as cholangitis, until the safety and efficacy profile of the investigational therapy has been evaluated.
Viral hepatitis infection	Patients with active Hepatitis B virus generally should be eligible for most clinical trials if appropriately suppressed with antiviral therapy based upon viral load according to general oncology practice guidelines [27]. Patients with previously-treated Hepatitis C virus (HCV) with sustained virologic response should be eligible for most clinical trials if liver function tests are within eligibility parameters. For patients with active HCV, antiviral therapy should be considered according to institutional practice noting limited data to guide decision-making in setting of advanced cancer.

Table 3. Stratification factors in randomized trials for advanced cholangiocarcinoma

Trial	ABC-02: GEMCIS vs. GEM [5]	ABC-06: FOLFOX plus ASC vs. ASC [8]	Ramucirumab, merestinib, or placebo plus GEMCIS [28]	AMEBICA-PRODIGE 38: Modified FOLFIRINOX vs. GEMCIS [29]
Design	Phase 3	Phase 3	Phase 2	Phase 2
Line of treatment	1^st line	2^nd line	1^st line	1^st line
Sample size	410	162	309	190
Stratification factors	Anatomic site Stage (locally advanced vs. metastatic) Performance status (0 vs. 1 vs. 2) Prior therapy Center	Platinum sensitivity (yes vs. no) Serum albumin (<35 vs. ≥ 35 g/L) Stage (locally advanced vs. metastatic)	Anatomic site (iCCA, eCCA, GBC, or ampullary) Geographic region Metastatic (yes vs. no)	Center Stage Performance status (0 vs. 1) Anatomic site (iCCA vs. eCCA vs. GBC) Prior adjuvant therapy (yes vs. no)
Abbreviations: GEMCIS = gemcitabine and cisplatin; GEM = gemcitabine; FOLFOX = fluorouracil (5-FU) and oxaliplatin; ASC = active symptom control; FOLFIRINOX = folinic acid (leucovorin), fluorouracil (5-FU), irinotecan, and oxaliplatin; GBC = gallbladder carcinoma.

4.2. Leveraging single arm and basket trials and natural history cohorts

Depending on the prevalence of the target in CCA, randomized trials with selection for an integral biomarker may not be feasible for all molecularly targeted therapies. A pragmatic approach is to rely upon multi-disease basket trials for initial signal finding in biomarker-selected subgroups, with expansion to a dedicated cholangiocarcinoma cohort or development of an independent trial upon demonstration of early signals of efficacy [30]. When exceptional efficacy is observed, regulatory filing of single-arm trial data may be warranted for therapies targeting rare molecular subsets, if accompanied by adequate natural history data.

Large, multicenter patient registries confer the potential to generate natural history data and serve as non-concurrent comparator arms for single-arm trials in rare diseases such as CCA [31]. Natural history cohorts in CCA require harmonized annotation for key covariates such as anatomic subsite, comorbidity including liver disease, and driver mutation status. The development of such registries also requires collaboration and coordination between cancer centers with high volumes of CCA patients along with patient outreach and recruitment efforts, however, highlighting an area of potentially high impact for advocacy communities.

4.3. Broadening biomarker platforms used for CCA clinical trials and clinical care

Biomarker development in CCA is challenged by the heterogeneity of tumor and microenvironment in CCA with rare individual subgroups, along with a scarcity of tumor tissue available for clinical diagnostics or research. In parallel with the development of longitudinal patient registries to collect clinical outcome data, large, integrated biorepositories of clinically annotated tumor and blood samples from patients treated on clinical trials as well as those treated with standard therapies are essential to serve as a platform for biomarker discovery and validation and for consistent molecular annotation of natural history cohorts [3].

For clinical trials of targeted therapies requiring integral biomarker testing to determine eligibility, the scarcity of available tumor tissue for testing poses a challenge to accrual and risks exhausting the available sample for a single test. A broad panel including all of the established and candidate therapeutic targets in CCA (table 1) should be employed whenever possible for initial biomarker testing to avoid the risk of exhausting the patient’s available tumor sample for only a single gene or limited panel test, and to facilitate treatment with alternate targeted therapies if a different target is identified in the process. Local testing with a broad panel performed in a Clinical Laboratory Improvement Amendments (CLIA)-certified environment should be allowed for clinical trial entry to minimize duplicative testing. Further, use of a widely available local assay as the clinical trial assay (CTA) and/or companion diagnostic (CDx) could minimize the tissue requirements from individual patients and increase the sample size for CDx co-development in this tumor type with severely limited access to tumor tissue.

Noninvasive clinical biomarker validation for specific targets in CCA is a key priority owing to the scarcity of adequate tumor tissue for molecular profiling. Next-generation sequencing (NGS) of circulating cell-free tumor DNA (cfDNA) can identify exonic hotspot mutations (such as IDH1 mutations) in gastrointestinal tumors including CCA, as well as kinase domain mutations associated with resistance to FGFR2 inhibition in CCA [10,32,33]. NGS of bile cfDNA has shown concordance with tumor mutation profiles in small studies in biliary tract cancers, suggesting bile may be an alternate source of cfDNA for molecular profiling [34]. For CCA clinical trials, cfDNA assay results from a CLIA-certified context should be allowed for study entry along with provision of tumor tissue for central testing after enrollment, if available.

A key challenge to the utility of cfDNA in cholangiocarcinoma, however, is the current lack of sensitivity for detecting translocations, particularly in the FGFR2 gene which can occur across large, non-coding, intronic regions [14]. Detection of FGFR2 and other fusions occurring in intronic regions will continue to require tumor tissue NGS for detection until cfDNA assays have been validated for individual translocations.

Declaration of interest

B Keenan has grant funding from NIH T32AI007334 and the Conquer Cancer Foundation, American Society of Clinical Oncology, and research support from Partner Therapeutics.

RK Kelley has research support from Agios, Astra Zeneca, Bayer, BMS, Eli Lilly, EMD Serono, Exelixis, Merck, Partner Therapeutics, QED, Novartis, and Taiho (all to institution). She receives consulting/steering committee fees to her institution from Astra Zeneca, BMS, and Agios, and receives advisory/IDMC fees from GNE/Roche, Ipsen, and Gilead. 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.

Reviewer disclosures

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