Olaparib in Treatment-Refractory IDH1- and IDH2-mutant Cholangiocarcinoma: Safety and Antitumor Activity from the Phase 2 NCI 10129 Trial

Abstract

Introduction:

Neomorphic Isocitrate dehydrogenase (IDH) mutations lead to accumulation of 2-hydroxyglutarate (2-HG), an oncometabolite implicated in tumor progression via inhibitory effects on alpha-ketoglutarate (α-KG). Moreover, mutant IDH-dependent accumulation of 2-HG results in homologous recombination deficiency (HRD), preclinically rendering tumors sensitive to Poly (ADP-Ribose) polymerase (PARP) inhibitors. Here, we report the results of the cholangiocarcinoma (CCA) arm of the NCI 10129 olaparib in IDHmt solid tumors basket trial.

Methods:

We evaluated olaparib 300 mg twice daily in an open-label phase 2 clinical trial for treatment refractory IDHmt solid tumors. Patients in the IDHmt CCA arm enrolled in 2 cohorts 1) IDH inhibitor (IDHi) pre-treated and 2) IDHi untreated with a primary endpoint of overall response rate (ORR).

Results:

NCI 1029 enrolled 30 patients with IDHmt CCA with no objective responses seen and recruitment was closed early. Median PFS was 2.4 months 95% CI (1.9 – 6.5) and median OS was 12.9 months 95% CI (6.3 – not reached). Eight patients (27%) had clinical benefit (CB) with progression-free survival ≥ 6 months. Patients with CB had lower baseline 2-HG levels compared to those without CB (1.4 μmol/L vs 5.9 μmol/L, p = 0.01).

Conclusion:

Olaparib does not have sufficient single agent activity to warrant further development in IDHmut CCA. However, a subgroup of patients demonstrated CB and exploratory analysis reveal this subgroup to be enriched for lower baseline 2-HG levels. Future clinical trials leveraging the HRD properties of IDH mutations are warranted with enhanced patient selection and novel combination therapies.

Precis:

Olaparib does not reveal radiographic response for patients with IDH1- and IDH2-mutant Cholangiocarcinoma. However, prolonged disease control was seen in a subset of patients, which correlated with lower tumor 2-HG levels and prior treatment with IDH inhibitor.

Main Text

Introduction

Cholangiocarcinoma (CCA) is a highly aggressive malignancy of the bile ducts associated with poor outcomes and rising incidence. The majority of patients with CCA are diagnosed with metastatic or unresectable disease and systemic therapy with gemcitabine, cisplatin and durvalumab is the current standard of care. Treatment with these agents is associated with a median survival of approximately 1 year.¹ In the second line setting there is no consensus on standard treatment. The recent molecular characterizations of intrahepatic CCA reveal this to be a heterogenous disease with multiple molecular subtypes. For example, mutations in isocitrate dehydrogenase (IDH) are identified in approximately 13% of the intrahepatic subtype and are less commonly seen in the extrahepatic subtype.^2–5 Furthermore, the IDH1 inhibitor (IDHi), ivosidenib has shown modest improvements in treatment refractory IDH-1 mutant CCA with median progression-free survival (PFS) 2.7 months with ivosidenib vs 1.4 months with placebo (HR 0.37 [95% CI 0.25–0.54]; p<0.0001) and overall survival (OS) to 10.8 months with ivosidenib vs 9.7 months with placebo (HR 0.69 [95% CI 0.44–1.10]; p=0.060).⁶

Neomorphic mutations in IDH 1/2 result in the accumulation of the oncometabolite 2-hydroxyglutarate (2-HG), which has been implicated in tumor progression via inhibitory effects on alpha ketoglutarate dependent dioxygenases including histone demethylases.^7–10 Moreover, the mutant IDH-dependent accumulation of 2HG results in homologous recombination deficiency and renders these cells exquisitely sensitive to Poly (ADP-Ribose) polymerase (PARP) inhibitors in pre-clinical models.¹¹ We have demonstrated IDH1-dependent PARP inhibitor sensitivity in a range of clinically relevant models, including tumor xenografts in vivo, as a way to exploit 2HG-mediated homologous recombination deficiency. We therefore designed an open-label phase 2 study evaluating the PARP inhibitor olaparib in treatment refractory tumors with IDH mutations.

The National Cancer Institute (NCI) Protocol 10129 (ClinicalTrials.gov identifier: NCT03212274) was a national trial aimed to take advantage of mutant IDH 1/2-mediated homologous recombination deficiency with the PARP inhibitor olaparib that evaluated patients in 3 cohorts: 1) IDH 1/2-mutated glioma 2) IDH 1/2-mutated CCA and 3) IDH 1/2-mutated all-comers (non-glioma or CCA). Here we report the results of NCI 10129 from the CCA arm, which evaluated the activity of olaparib in patients with IDH 1/2 mutated CCA.

Methods

Study design and participants

The NCI 10129 study was a 3 arm, open label phase 2 clinical trial performed through the NCI Experimental Therapeutics Clinical Trials Network (ETCTN). The study was sponsored by the Cancer Therapy Evaluation Program (CTEP) with the Yale Cancer Center as the lead academic organization. Eligible patients had mutations in IDH 1 or 2 and were assigned to a study arm based on histology: 1) IDH-1/2 mutant glioma 2) IDH-1/2 mutant CCA or 3) IDH-1/2 mutations in any solid tumor excluding glioma or CCA. All eligible patients were aged 18 years or older, had measurable disease according to the Response Evaluation Criteria in Solid Tumors (RECIST version 1.1) for extracranial tumors, had an Eastern Cooperative Oncology Group (ECOG) performance score of ≤2, had an estimated life expectancy of 4 months or more in the judgement of the investigator, and had adequate organ function (as assessed by renal, hepatic, hematological, and coagulation parameters). Patients were excluded if they had previously received olaparib or had not progressed on standard therapies. The complete study protocol, including a list of all eligibility criteria is included in the supplementary data. The NCI Central Institutional Review Board approved the study, which adhered to Good Clinical Practice guidelines. All patients provided written, informed consent as a condition of study participation.

Procedures

All treated patients received olaparib 300 mg twice daily continuously in 28-day cycles. Pre-defined dose modifications were allowed to manage clinically significant toxicities related to individual agents as specified in the protocol (supplementary data). Participants received study treatments until radiographic or symptomatic progression, unacceptable toxicity, death, or withdrawal from the study. Radiographic tumor assessments were performed at baseline and every 8 weeks thereafter, from enrollment until radiographic disease progression.

Oncometabolites were assessed by measuring D- and L-2HG concentrations in patients at multiple time points: 1) pre-treatment 2) day 1 of each cycle, and at end of treatment using a validated liquid chromatography tandem mass spectrometry (LC-MS/MS) method, as reported by us previously.¹² The linear calibration curve range in aqueous solution was 0.0002–5 μM for D- and L-2HG. Intra- and inter-day precision and accuracy were within the generally accepted criteria for bioanalytical method validation (<15%).

Molecular profiling was assessed by whole exome sequencing (WES) using Agilent Sureselect V6 + UTR probeset and Illumina NovaSeq 6000 at the Frederick National Laboratory for Cancer Research. For WES, reads were mapped to the human hg19 reference genome using the Burrows-Wheeler Alignment tool. BAM files were processed using GATK best practice workflow. GATK MuTect2 were used to call somatic variants. Copy number data was inferred from WES data through use of the Sequenza algorithm. Homologous recombination deficiency (HRD) scores were calculated by scarHRD based on Sequenza output.¹³ Mutational signatures were identified using COSMIC version 2 database.

Circulating tumor DNA (ctDNA) was assessed from whole blood that was processed to plasma for storage at multiple time points: 1) pre-treatment 2) day 1 of each cycle, and at end of treatment. Cell free DNA (cfDNA) was extracted from plasma.¹⁴ Somatic mutations from tumor in cfDNA were quantified by error-suppressed deep sequencing, which queries thousands of potential point mutations and insertions/deletions within 43 mutation-prone regions for 24 cancer-associated genes.^14–16 Targeted deep sequence was performed in 75 base-pair, paired end mode on Illumina HiSeq2500.

Outcomes

The primary endpoint for each cohort was the objective response rate. The overall response is defined a complete or partial response as determined by investigator assessment using RECIST (version 1.1) for extracranial solid tumors. Secondary endpoints included progression-free survival, overall survival, duration of response, and safety and tolerability of olaparib monotherapy in the study population.

Statistical analysis

Historical control data suggests that the response rate is 10% or less for the refractory tumors enrolled on this study, including the CCA cohort. The CCA cohort classified patients into 2 sub-cohorts: 1) IDH inhibitor naïve and 2) IDH inhibitor exposed. Furthermore, the study included an early stopping rule if at least 2 of the first 15 patients in each sub-cohort did not have an objective response by RECIST (version 1.1). If neither sub-cohort was terminated at interim, each sub-cohort would accrue to a target of 30 patients (60 total for the CCA cohort). The overall response was estimated with a 90% creditable interval in each cohort and overall survival (OS) and progression-free survival (PFS) curves were generated using the Kaplan Meier method. Univariate and multivariate logistic regression models were created to identify variable associated with prolonged PFS and OS.

Role of the funding source

The clinical trial was funded by the NCI through grant #: 1UM1CA186689. Additional funding was provided from the Rising Tide Foundation (CCR-18–700), and the Catherine Ivy Foundation. The olaparib was supplied from AstraZeneca through a cooperative research and development agreement (CRADA) and clinical trials agreement with the NCI CTEP.

Results

Between January 30, 2019 and May 10, 2022, 30 patients were enrolled in the IDHmt CCA arms of NCI 10129 (Supplementary Fig. 1). Baseline demographics and disease characteristics are summarized in Table 1. All 30 patients had received prior platinum therapy. One patient in the IDHi pre-treated cohort was erroneously enrolled without prior IDHi exposure. At the time of the February 13, 2023 data cutoff, all 30 patients had discontinued olaparib and no patients remained on therapy.

Table 1.

Baseline Characteristics of the Intention to Treat Population

Characteristic	NCI 10129 CCA Arm (N = 30)
Age
Median	66
Range	47 – 76
Sex – no. (%)
Male	10 (33)
Female	20 (67)
Race – no. (%)
White	25 (83)
African American	1 (3)
Asian	2 (7)
Unknown	2 (7)
ECOG performance status – no. (%)
0	15 (50)
1	13 (43)
2	2 (7)
Location of primary tumor – no. (%)
Intrahepatic CCA	29 (97)
Extrahepatic CA	1 (3)
IDH Mutation – no. (%)
IDH 1	24 (80)
IDH 2	6 (20)
Prior liver directed therapy – no. (%)	11 (37)
Prior surgical resection of primary tumor – no. (%)	4 (13)
Number of prior systemic therapies – no. (%)
1	12 (40)
2	5 (17)
≥3	13 (43)

There were no objective responses seen by RECIST v1.1 in either the IDHi pre-treated or naïve groups (Figure 1). Consequently, both treatment arms were discontinued according to the prespecified futility analysis. For all enrolled CCA patients, the median PFS was 2.4 months (95% CI, 1.87 – 6.5) and median OS was 12.9 months (95% CI, 6.3 – not reached) (Figure 2A–B). Eight (27%) patients had clinical benefit (CB) with ≥ 6-month PFS, with 4 patients in each cohort alive without progression at 6 months. For the 8 patients with CB, 7 had IDH1 mutated tumors and 1 had an IDH2 mutated tumor. Median PFS in the IDH inhibitor naïve cohort was 2.07 months (95% CI, 1.67 – 11.47) and 3.77 (95% CI, 1.87 – not reached) months in the IDHi pre-treated cohort, P = 0.88) (Figure 2C). The median OS in the IDHi naïve cohort was 7.73 months (95% CI, 3.73 – not reached) and 15.3 months (95% CI, 6.30 – not reached) in the IDHi pre-treated cohort, P = 0.04 (Figure 2D).

Figure 1. Radiographic Response.

Waterfall plot for evaluable patients, representing radiographic best response of tumor change from baseline according to RECIST v1.1.

Figure 2. Survival Curves.

Progression-free survival (2A, top left panel) and overall survival (2B, top right panel) curves for intention to treat population. Progression-free survival (2C, bottom left panel) and overall survival (2D, bottom right panel) stratified by IDH inhibitor-pretreated (blue) and IDH inhibitor treatment naïve (gold).

The oncometabolites D-2-HG, L-2-HG, and total 2-HG were measured prior to treatment, on-treatment at cycle 2 day 1, and at progression. A non-statistically significant increase in median baseline total 2-HG levels was observed on-treatment, P = 0.6, and at progression, P = 0.4 (Figure. 3A). Results for L-2-HG and D-2-HG are provided in Supplementary Fig 2A–B. Patients with CB had statistically significant lower median baseline total 2-HG, P = 0.01, compared to patients with no clinical benefit (NCB) (Figure. 3B). Furthermore, an increase in median % change of total 2-HG for each patient from baseline, on-treatment, and progression increased for patients with CB, P = 0.02, and NCB, P = 0.17 (Figure. 3C). Exploratory analysis for ctDNA for IDH variant allele frequency in each treatment cycle was available for 9 patients (2 CB and 7 NCB) and revealed stability for patients with CB (Supplementary Fig 3).

Figure 3. Oncometabolite Measurements.

A) Median Total-2-HG levels at baseline, cycle 2 day 1, and at disease progression. B) Total-2-HG levels at baseline for patients with clinical benefit vs no clinical benefit. C) Median percent change of Total-2-HG levels at baseline and at cycle 2 day 1 for patients with clinical benefit and patients without clinical benefit.

Tumor WES and RNAseq for at least one time point was available for 23 patients and is reported by descriptive statistics (Supplementary Fig 4). No damaging alterations were identified in genes known to confer HRD. No association of HRD assessments with CB vs NCB were identified.

All treatment emergent adverse events (TEAE) with ≥ 10% incidence are reported in table 2. Nine patients had at least 1 dose hold for a TEAE attributed to olaparib (6 anemia, 1 AST/ALT elevation, 1 thrombocytopenia, and 1 combined anemia and thrombocytopenia). Seven patients underwent olaparib dose reductions (3 for ≥ grade 3 anemia, 2 for creatinine clearance changes, 1 patient decision, 1 ≥ grade 3 ALT/AST elevation). No patients discontinued study treatment for TEAE attributed to olaparib.

Table 2.

Treatment emergent adverse events with ≥ 10% incidence

Event	Any Grade	Grade ≥ 3
Any event – no. (%)	30 (100)	12 (40)
Any serious event – no. (%)1	13 (43)	12 (40)
Most common events – no. (%)
Fatigue	12 (40)	2 (7)
Nausea	9 (30)	0
Vomiting	4 (13)	0
Diarrhea	4 (13)	0
Anorexia	3 (10)	0
Constipation	3 (10)	0
Dysguesia	3 (10)	0
Elevated serum creatinine	3 (10)	0
Hematologic toxicities – no. (%)
Anemia	15 (50)	7 (23)
Thrombocytopenia	7 (23)	2 (7)
Decreased lymphocytes	5 (17)	1 (3)
Decreased white blood cells	5 (17)	0

Discussion

The treatment approach for metastatic CCA changed significantly during the course of this study, with the approval of immune checkpoint inhibitors, fibroblast growth factor receptor (FGFR) inhibitors for tumors with FGFR alterations, and ivosidenib for CCA tumors with IDH1 mutations.^6,17–19 These important advances reveal the value of precision medicine in this disease; however, survival outcomes for unresectable/metastatic cholangiocarcinoma remain quite poor and timely development of new therapeutic strategies is an urgent priority in the field. The NCI 10129 study is the first clinical trial to evaluate a PARP inhibitor for oncometabolite-mediated HRD in IDH mutant CCA.

In this open-label, phase 2 clinical trial we did not observe any objective responses by RECIST criteria, and the study closed early for futility per the pre-specified early stopping rule. Olaparib was well tolerated in a heavily pre-treated patient population with 60% of patients having received at least 2 prior systemic therapies in the unresectable/metastatic setting. For this late line patient population, the PFS and OS compare quite favorably to the historical controls. Furthermore, the WES results do reveal enrichment for evidence of HRD in the patient population, which supports the pre-clinical findings that supported the study that IDH mutations are associated with HRD. The randomized ClarIDHy clinical trial with ivosidenib versus placebo is the largest clinical trial and best survival benchmark to date for patients with IDH mutated cholangiocarcinoma, in the second line setting. In ClarIDHy the radiographic response rate was modest at 4% and median PFS and OS in the investigational arm were 2.7 months and 10.3 months, respectively.^6,20 Furthermore, in the placebo arm of ClarIDHy, 0 out of 76 patients had a PFS ≥ 6 months. In contrast, we report 8/30 (27%) patients with ≥ 6-month PFS in NCI 10129. Moreover, the OS in NCI 10129 compares favorably to the ivosidenib arm of ClarIDHy, despite our trial having a more heavily pre-treated patient population. Notably, we did not identify any difference in survival by number of prior treatments (Supplementary Fig. 5A–B) or mutations in IDH1 vs IDH2 (Supplementary Fig. 5C–D). However, the comparison of subgroups is limited due to small sample size and potential for confounders. For example, the survival for IDHi pre-treated vs naïve (Fig. 2C–D) is potentially impacted by the fact that many patients in IDHi pre-treated cohort may have been healthier as they likely received prior IDHi in the context of a clinical trial. Despite these limitations, the median OS of 12.9 months in the intent to treat population is encouraging for patients with treatment refractory CCA.

The oncometabolite findings also reveal patients with CB to have lower 2-HG levels than patients with NCB, which could be explained by enhanced tumor sensitivity to 2-HG mediated HRD or perhaps by a lower tumor burden for these patients and more indolent disease which could bias the survival interpretation. However, patients with CB had lower levels of baseline total 2-HG than NCB regardless of receipt of prior IDHi (Supplementary Fig. 6A–B), numer of prior treatments (Supplementary Fig. 6C–D), or mutation in IDH1 vs IDH2 (Supplementary Fig. 6E–F). Thus, while the study did close early for futility, the survival and oncometabolite analysis may suggest additional novel strategies targeting DNA repair are warranted in IDHmt cholangiocarcinoma.

The efficacy findings from this clinical trial do not justify further development of single agent PARP inhibitors for late line patients with IDHmt cholangiocarcinoma. However, given the survival analyses, future strategies should be considered. First, more potent synergistic DNA damage combinations should be considered such as PARP inhibitor and ATR inhibitor combinations, which are being evaluated in the ongoing NCI 10222 (NCT03878095) cholangiocarcinoma arm. Second, a more uniform patient population with a prespecified number of prior lines of therapy may better inform the true potential survival benefits. Third, maintenance PARP inhibitors have been successful in other tumor types, and maintenance PARP inhibitor after 6 months of gemcitabine + cisplatin + immune checkpoint inhibitor could be considered for future study designs. Furthermore, our genomic analyses revealed median HRD sum score was 33 (range 7–83) and 12/24 (50%) of patients had HRD mutational signature 3 identified in their tumors. These clinical observations add to the pre-clinical data suggesting that tumors with neomorphic IDH mutations have HRD and may benefit from novel strategies targeting DNA repair. The potential of combining ivosidenib with a PARP inhibitor seems unlikely to be successful given the therapeutic effects of ivosidenib in reducing 2-HG, making these tumors less likely to be HRD.

Our study is limited primarily by the sample size as enrollment was not continued beyond our pre-specified early stopping rule. While response rate was our primary endpoint, given the difficulty for radiographic response in later line cholangiocarcinoma, future studies should focus on survival endpoints. Furthermore, the lack of randomization prevents direct comparisons and makes it challenging to adequately interpret the survival findings.

In conclusion, the NCI10129 clinical trial evaluated the feasibility and safety of olaparib in late line IDHmt cholangiocarcinoma, but failed to demonstrate sufficient efficacy measured by response rate to continue enrollment beyond the early stopping rules. However, the PFS and OS endpoints compare favorably with historical controls, and further studies with alternative endpoints in a more homogenous patient population may be warranted. Furthermore, measurement of the oncometabolite 2-HG may help identify patient subgroups more likely to benefit from these therapies. Thus, based on the survival results we believe that future strategies targeting DNA repair in IDHmt cholangiocarcinoma is still warranted with novel and more potent DNA damaging therapies.

Data availability statement:

Data may be made available based on a reasonable request to the corresponding author.