Randomized Phase II Study of Axitinib versus Observation in Patients with Recurred or Metastatic Adenoid Cystic Carcinoma

Adenoid cystic carcinoma (ACC) is a rare neoplasm that most often arises from the salivary glands (1). ACC is characterized by relatively reduced growth compared with other carcinomas but eventually becomes a life-threatening condition (1). Even with distant metastases, a watchful wait-and-see strategy is recommended for asymptomatic patients with a slow-growing tumor, considering the low response rates and toxicities of chemotherapeutic agents. Systemic chemotherapy is only considered when a patient has a rapidly growing cancer or when there are related symptoms.

Several molecular targeted agents, which inhibit VEGFR, fibroblast growth factor receptor, platelet-derived growth factor receptor (PDGFR), or receptor tyrosine kinase (KIT), have been investigated in several phase II trials. In previous studies of multi-targeted tyrosine kinase inhibitors (TKI), such as lenvatinib, rivoceranib, regorafenib, dovitinib, axitinib, pazopanib, sorafenib, or sunitinib, the response rates and median progression-free survival (PFS) varied between 0%–47% and 5.7–17.5 months, respectively (2–12).

However, because these trials were conducted with a small number of patients in a single arm, the role of these agents is still controversial for this rare cancer. In these previous single-arm studies, a substantial proportion of patients (59%–94%) showed stable disease as their best response. As ACC has slow-growing characteristics, it is difficult to discriminate active disease control via chemotherapy from the natural stability of the disease. Therefore, randomized trials to confirm the role of TKIs in ACC are needed. To this end, in this study, we conducted the first randomized trial to affirm the efficacy of axitinib as compared with observation in patients with recurrent or metastatic ACC.

Patients that were 20 years of age or over with pathologically confirmed recurrent or metastatic ACC, which was not appropriate to curative treatment, were eligible if they had experienced disease progression within the 9 months prior to the study. Radiographic progression by the Response Evaluation Criteria in Solid Tumor (RECIST 1.1) was documented by radiologic images taken within 9 months prior to the baseline image of the study. The other eligibility criteria were an Eastern Cooperative Oncology Group (ECOG) performance status of 0–1 and adequate organ function, including bone marrow, kidneys, and liver. The exclusion criteria included patients with significant heart disease or myocardial infarction within the past 12 months, pre-existing uncontrolled hypertension of more than 160/100 mm Hg despite adequate medical therapy, deep vein thrombosis or pulmonary embolism within 6 months of study entry, history of massive pulmonary hemorrhage or hemoptysis within the past 6 months, history of uncontrolled coagulopathy, or previous history of anti-angiogenic inhibitor treatment. Prior use of cytotoxic chemotherapy, other than antiangiogenic inhibitor treatment, was allowed.

This study was an investigator-initiated, randomized, open-labeled, prospective phase II trial. Eleven Korean Cancer Study Group–affiliated hospitals participated in the study. Sixty patients were randomly assigned in a ratio of 1:1 to either the axitinib or the observation arm. The randomization was performed centrally using the permuted block randomization method. For the observation arm, a “watchful wait-and-see” approach was applied, without any cancer treatment, until disease progression was documented. Once disease progression was confirmed, patients in the observation arm were crossed over to axitinib treatment. For the axitinib arm, axitinib was administered orally (5 mg twice daily) until disease progression, unacceptable toxicity, or withdrawal of consent occurred. Each cycle consisted of 28 days. Dose modifications or delays in study drug administration were performed on the basis of the worst grade of toxicity according to the approved protocol. When grade 3 or 4 toxicities were observed, axitinib was discontinued until the toxicities were resolved to grade 1 or lower. Axitinib was then readministered with one dose-level reduction. A minus 1 reduced dose level was 3 mg twice daily, and a minus 2 reduced dose level was 2 mg twice daily. No dose reductions to less than 2 mg twice daily were permitted. When the administration of the study medication was delayed, all evaluations, including tumor evaluation, still adhered to the original schedule. However, when the administration of the study medication was either delayed by 3 weeks or more or the medication was permanently discontinued due to toxicity, the patient was withdrawn from the study. Patients were also withdrawn in cases of consent withdrawal irrespective of the reason.

Archival tumor tissue from 28 of the study patients was available for next-generation sequencing (NGS). Archival tumor tissue from the primary site was used for NGS. The FoundationOne CDx assay (13–15) was performed after informed written consent was obtained from the patients. This assay interrogates 324 genes and is optimized for DNA from formalin-fixed, paraffin-embedded samples. It is able to detect base substitutions, indels, copy-number alterations, rearrangements, microsatellite instability (MSI), and tumor mutational burden. Gene-level copy number changes were calculated using segmented log-ratio values (Circular Binary Segmentation) of tumor and normal samples.

The primary endpoint was the 6-month progression-free survival (PFS) rate, as determined by the investigators in accordance with RECIST 1.1 (16). PFS was defined as the time from randomization to radiographic progressive disease or death, and data were censored at the most recent disease assessment. The secondary endpoints included response rate, overall survival (OS), PFS, response duration, and toxicities. All randomized patients were included for intention-to-treat analysis. Efficacy analysis was performed on a per-protocol basis. Non-evaluable patients were excluded from the per-protocol analysis.

The response evaluations were conducted via RECIST 1.1 and performed every 8 weeks until cancer progression. Tumor assessment was conducted every 8 weeks for both groups. If disease progression was suspected due to symptomatic aggravation, an earlier image assessment was performed per each investigator's decision. Adverse events (AE) were monitored and recorded according to CTCAE version 4.03 during the treatment phase and for 6 weeks after the final dose of axitinib.

We expected a 20% 6-month PFS rate in the observation arm, with an expected HR of 0.496 and a one-sided type I error of 7.5% at 80% power. We planned for an accrual time of 30 months and a follow-up period of 12 months after the last accrual. Under the considerations given above, we estimated that a total of 47 events were needed for statistical significance. To observe these events, 24 patients in each arm would be required. This sample size was based on the assumption that patient survival would follow an exponential distribution and that no patients would be lost to follow-up. Including a 5% drop-out rate, a minimum of 26 patients would be needed from each arm, giving a total of 52 required patients. However, since it would be somewhat unlikely to expect 47 events out of just 52 patients, we decided that 30 patients in each arm (60 patients in total) would be more appropriate. Therefore, the target total number of patients was 60.

The PFS of axitinib in the observation arm was defined as the date of the first dose of axitinib after the crossover to the date of documented disease progression or death. OS was defined as the time from the date of randomization to the date of death. The probability of survival was estimated using the Kaplan–Meier Method. The comparison of survival rates between the two arms was made using log-rank tests. All analyses were performed with SPSS for Windows 21.0 (IBM Corporation). Data were analyzed until April 2020.

The trial was conducted in accordance with the Declaration of Helsinki, Good Clinical Practice guidelines, and national policy on bioethics and human biologic specimens. This trial was registered with ClinicalTrials.gov (NCT02859012). The study protocol was approved by the Institutional Review Board of each participating institution. Written informed consent was obtained from all patients before participation.

Between December 2016 and October 2017, 63 patients were screened, and 60 of the screened patients were enrolled in this study. All enrolled patients showed evidence of disease progression within 9 months before study participation. Patients were randomly allocated into either the observation or axitinib arm (30 patients in each arm; Fig. 1). The baseline characteristics of patients are shown in Table 1. The median age was 56 (range, 26–77), and 46.7% of patients were men. The proportion of patients with distant metastasis was similar in both groups. In the observation arm, 46.7% of patients had not received any prior systemic treatment, while 26.7% of patients in the axitinib arm had not received any prior systemic treatment. All patients in the observation arm were followed for disease status and crossed over into the axitinib arm if disease progression was confirmed. Among the 30 patients in the observation arm, 26 patients crossed over to axitinib after disease progression was documented. Four patients did not crossover, one patient died before progression could be documented, one was lost to follow-up, and two withdrew their consent for this study.

Response evaluation was conducted in the evaluable 54 patients (27 patients in each arm), and the efficacy results are presented in Fig. 2. The 6-month PFS rate was 73.0% [95% confidence interval (CI), 52.0%–86.0%] in the axitinib arm and 23.0% (95% CI, 9.0%–41.0%) in the observation arm, with a median follow-up of 31.9 months and 28.4 months, respectively (Fig. 2A). The median PFS was 10.8 months (95% CI, 7.1–13.6 months) in the axitinib arm and 2.8 months (95% CI, 1.7–4.2 months) in the observation arm (HR, 0.25; 95% CI, 0.14–0.48; P < 0.001; Fig. 2A). Median OS was not determined in the axitinib arm, but was 27.2 months (95% CI, 20.2–32.8 months) in the observation arm (HR = 0.60; 95% CI, 0.26–1.38 months; P = 0.226; Fig. 2B).

The majority of patients on axitinib experienced tumor shrinkage (17 of 27 patients, 63.0%), and there was no partial response (PR) or progressive disease (PD) in this group (Fig. 2C). The disease control rate (DCR) was 100.0% (95% CI, 87.2%–100.0%) in the axitinib arm versus 51.9% (95% CI, 31.9%–71.3%) in the observation arm.

After crossover to axitinib, 3 patients responded to axitinib, and the ORR of axitinib in the observation arm was 11.5% (95% CI, 2.5%–30.2%). Two out of 3 responders were chemotherapy-naïve patients. The DCR was 92.3% (95% CI, 74.9%–99.1%). For 2 patients, (7.7%) disease progressed even after crossover to axitinib treatment (Fig. 2D). The median PFS of axitinib after crossover was 14.5 months (95% CI, 10.7–20.3 months; Table 2, per-protocol population; evaluable patients).

A total of 60 patients (30 in each group) were included in the safety analysis. A summary of the common AEs is presented in Table 3 (safety population). The most frequently reported adverse event for the axitinib arm was grade 1 or 2 stomatitis (43.3%), with anorexia (33.3%) and proteinuria (33.3%) the next-most frequently reported AEs. The most frequent grade 3 AE was hypertension (23.3%). No grade 4 or 5 toxicities were recorded. Eight patients (6 in the axitinib arm, 2 after crossover to axitinib in the observation arm) required at least one dose reduction. Six patients in the axitinib arm and 4 patients who crossed over to axitinib in the observation arm interrupted treatment due to toxicities, which included grade 3 atrial fibrillation, hyponatremia, QT prolongation, hyperbilirubinemia, osteonecrosis of jaw, anorexia, and spinal fracture. The dose intensity was 76.2% in the axitinib arm and 76.6% in patients who crossed over to axitinib in the observation arm.

NGS using FoundationOne CDx assay was performed on the pre-axitinib archival tissue samples of 28 patients (Fig. 3). The mutation burden for each patient ranged from 4 to 23 mutations per megabase. The most frequently identified mutations were in ARID1B (8/28, 28.6%). Mutations in KDM6A and PARP4 were detected in 6 patients. NOTCH1, MLL2, MLL3, and SF3B1 mutations were detected in 5 patients. FGFR4 mutations were detected in 4 patients, and MYB fusions were detected in 3 patients. The mutation burden did not correlate with response, PFS, or OS. Furthermore, MYB fusion did not affect response, PFS, or OS. Mutations in FANCM were a positive predictive factor for PFS and mutations in PCTH2 were a negative predictive factor for PFS. Patients with SFB1, ARID1A, or BCOR mutations exhibited poor overall survival compared with others.

In this randomized study, axitinib demonstrated substantial PFS benefit to patients with recurrent or metastatic ACC as compared with observation alone. In addition, the clinical efficacy of axitinib was shown in patients who crossed over to axitinib from the observation arm after documented disease progression. Acceptable and manageable toxicity profiles also conferred an advantage for patients with ACC who were treated with axitinib.

Although there have been several clinical trials for various treatments for ACC, prior studies have lacked comparator arms. ACC has a broad spectrum of disease courses, from indolent status to fast growing; therefore, the tumor characteristics of participants in clinical trials can affect the results. To minimize the heterogeneity of disease, many previous studies have limited the study cohort to patients with confirmed disease progression within a 6- to 12-month period (4, 6, 7, 17, 18). However, despite this kind of inclusion criteria, no reliable outcome can be guaranteed because there is no control group. Our approach, using a two-arm randomized study design with an observation arm as a control can return more reliable evidence of efficacy.

In two previous single-arm phase II trials of axitinib, the response rates were 8% and 9.1%, respectively, and the median PFS were 5.5 months and 5.7 months (7, 12). Although these two trials did not meet their primary endpoints, the sample sizes were relatively small. False negatives occur in small-cohort phase II trials. Therefore, in this study, we aimed to confirm the role of axitinib with a sufficient sample size. Compared with the previous studies, we found the response rate in the axitinib arm was low, but the median PFS was longer at 10.8 months. In addition, after crossover to axitinib, the response rate was higher (11.5%), which reflected the results from the previous single-arm axitinib study. In the previous single-arm axitinib study, 18.2% of patients had previously been treated with angiogenesis agents. However, in our study, we excluded patients who had a previous history of antiangiogenic inhibitor treatment. This discrepancy between the proportion of patients with a history of previous exposure to antiangiogenetic inhibitors may account for the efficacy difference between the studies.

In the single-arm phase II study of the multikinase VEGFR inhibitor, lenvatinib (2), the response rate was 15.6%, and the median PFS was 17.5 months. However, 62.5% of patients had at least one grade 3 or 4 toxicity, and 18 patients (54.5%) discontinued treatment due to AEs. In this study, there was no grade 4 toxicity at all. In addition, only 6 patients in the axitinib arm (20.0%) discontinued treatment due to AEs. Therefore, axitinib has fewer toxicities compared with lenvatinib, which is a benefit for sufficiently sustained treatment and positive outcomes.

No difference in OS was observed between the two groups, which might be attributed to the extensive crossover to axitinib in the placebo arm due to disease progression. It is difficult to conclude whether early use of axitinib is better than delayed use after progression; however, axitinib has been shown to induce a definite delay in progression and may thus be effective against symptomatic deterioration due to progression.

Progress in genomic testing and molecular biology has yielded several reports of the genomic abnormalities associated with ACC (19–23). According to the large-scale genomic analysis of 1,045 recurrent or metastatic ACC samples, recurrent or metastatic ACCs were enriched for mutations in NOTCH and chromatic remodeling associated genes (e.g., KDM6A, KMT2C/MLL3, or ARID1B). In addition, patients with either a NOTCH1 mutation or a KDM6A mutation showed poor prognosis in the previous study (23). Our genomic analysis findings are consistent with this large-scale genomic analysis; the most frequently detected alterations involved ARID1B, KDM6A, PARP4, NOTCH1, and MLL3. However, the prognostic impact of particular gene alterations differed from previous reports. In this study, mutations in SFB1, ARID1A, and BCOR were negative prognostic factors.

There have been continuing efforts to identify predictive biomarkers of TKI to harmonize results from previous trials with genomic data. In previous reports, NOTCH1 or KDR (VEGFR2) mutations or the amplification of 4q12, RET, or FGFR1 were suggested to have the potential to act as predictive biomarkers of TKI (2, 7). However, mutations in FANCM and PCTH2 showed potential predictability for axitinib efficacy in this study. However, a lack of evidence remains for treatment approaches based on genomic data, primarily due to inadequate sample sizes. The genomic analysis in this study was also limited by the small number of patients. Therefore, further, more comprehensive research with larger numbers of patients is required to investigate and validate any predictive genomic markers for axitinib.

This trial had several limitations. First, this study was not conducted as a double-blind placebo control study; therefore, a potential risk of investigator bias may exist. However, physicians could recognize the patient's allocation by the AEs. Given that a placebo has no AEs and anti-angiogenic agents do, even if this trial was conducted as a double-blind placebo control study, it would not have been able to achieve a perfectly blinded effect. Second, the response evaluation and the decision to crossover to axitinib were based on each investigator's assessment, not by a central review. Each investigator-directed decision was solely for the safety and benefit of their patients in the observation arm when disease progression was documented. Third, most of the tumor tissue was not metastatic but from the primary site. Because of the significant change in the biological profile between primary and recurrent ACC, this would be a potential limitation for interpreting the NGS data. Fourth, we did not capture quality-of-life data or detailed information regarding the pathologic subtype.

Despite these limitations, our study is the first randomized trial for ACC. Our study addresses the long-standing problem that single-arm ACC studies may not adequately capture the benefits of therapy given the low overall major response rates to antiangiogenic agents.

In conclusion, axitinib significantly increased the 6-month PFS rate as compared with observation in recurrent or metastatic ACC. This first randomized trial confirms that axitinib can actively suppress tumor growth with tolerable toxicities in progressive ACC.

K.-W. Lee reports grants from AstraZeneca, Ono Pharmaceutical, Merck Sharp & Dohme, Merck KGaA, Pfizer, BeiGene, Astellas Pharma, ALX Oncology, Zymeworks, Macrogenics, Five Prime Therapeutics, Oncologie, Pharmacyclics, MedPacto, Green Cross Corp, ABLBIO, Y-BIOLOGICS, Genexine, Daiichi Sankyo, and Taiho Pharmaceutical (to institution), as well as personal fees from ISU ABXIS, Bayer, Daiichi Sankyo, and Bristol Myers Squibb (consultation) outside the submitted work. S.-B. Kim reports grants from Novartis and Sanofi-Aventis, as well as non-financial support from Dongkook outside the submitted work. S.-B. Kim is also an advisory board consultant for Novartis, AstraZeneca, Lilly, Dae Hwa Pharmaceutical Co. Ltd, ISU Abxis, and Daiichi-Sankyo, and reports stocks in Genopeaks and NeogeneTC. B. Keam reports grants from MSD, AstraZeneca, and Taiho Pharmaceutical, as well as personal fees from AstraZeneca, MSD, Handok, CBS Bio, and personal Genexine outside the submitted work. No disclosures were reported by the other authors.

E.J. Kang: Conceptualization, resources, methodology, writing–original draft, project administration. M.-J. Ahn: Resources, methodology, writing–review and editing. C.-Y. Ock: Resources, visualization, writing–review and editing. K.-W. Lee: Resources, project administration, writing–review and editing. J.H. Kwon: Resources, writing–review and editing. Y. Yang: Resources, project administration, writing–review and editing. Y.H. Choi: Resources, writing–review and editing. M.K. Kim: Resources, project administration, writing–review and editing. J.H. Ji: Resources, project administration, writing–review and editing. T. Yun: Resources, project administration, writing–review and editing. B.-H. Nam: Data curation, software, formal analysis. S.-B. Kim: Resources, supervision, project administration, writing–review and editing. B. Keam: Conceptualization, resources, supervision, methodology, writing–review and editing.

We thank the participating patients and their families, all study coinvestigators, and research coordinators. We appreciate Korean Cancer Study Group for approval (KCSG HN 16–08) and supporting this study. We also thank HERINGS for its support with data management and statistical analyses. Axitinib was provided by Pfizer Inc., and genomic studies were supported by Roche Pharmaceuticals. This work is partially funded by the Adenoid Cystic Carcinoma Research Foundation. Also, this study was partially supported by a grant from the National R&D Program for Cancer Control, Ministry of Health and Welfare, Republic of Korea (1720150). This study is an investigator-sponsored trial, and the study was conducted independently of the funder. The funder had no role in study design, data collection, data analysis, or data interpretation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.