The study was a randomised, double-blind, placebo-controlled, multicentre, phase 3 trial, involving 57 hospital sites in 17 countries (Austria, Belgium, Canada, China, Finland, France, Germany, Israel, Italy, Japan, Netherlands, Poland, Singapore, South Korea, Spain, UK, and USA).
The study protocol was approved by the institutional review board of each participating institution and complied with the Declaration of Helsinki, current Good Clinical Practice guidelines, and local laws and regulations. An independent data monitoring committee, comprising three oncologists and a statistician, ensured the overall integrity of the trial and safety of participants. All participants provided informed consent before enrolment.
Eligibility included histologically confirmed, metastatic and/or unresectable GIST, with failure of at least: (1) prior imatinib (due to either disease progression or intolerance) and (2) prior sunitinib (due solely to progression to decrease heterogeneity, since the definition of intolerance is more variable with this agent). Patients could have received other systemic therapies, including investigational agents, except any VEGFR inhibitors other than sunitinib. Additional inclusion criteria included: at least one measurable lesion on computed tomography or magnetic resonance imaging; resolution of all toxic effects of prior therapy to grade 1 or less; adequate haematological, hepatic, cardiac, and renal function; and an Eastern Cooperative Oncology Group (ECOG) performance status of 0 or 1.
Enrolled patients were randomised in a 2:1 ratio to receive either regorafenib 160 mg orally once daily or matching placebo, for the first 3 weeks of each 4-week cycle. All patients also received best supportive care (defined as any method to preserve the comfort and dignity of the patient, excluding disease-specific antineoplastic therapy such as TKI therapy other than study drug, chemotherapy, radiation therapy or surgical intervention). Blinded study drug administration was continued until disease progression, occurrence of unacceptable toxicity, or patient withdrawal from the study.
In the event of centrally assessed tumour progression, treatment assignment could be unblinded. Patients originally assigned to the placebo arm were offered the option to cross over to receive open-label regorafenib, and patients originally assigned to the regorafenib arm could continue to receive open-label regorafenib, both at the discretion of the investigator. Throughout both the blinded and the open-label phases of the trial, the dose of study drug could be delayed or reduced according to a prespecified schedule in the case of unacceptable toxicities (see Online Supplemental Tables 2–5
Tumour assessments were made at baseline, then every 4 weeks for the first 3 months, every 6 weeks for the next 3 months, and subsequently every 8 weeks until the end of study drug administration. Intervening tumour assessments could be made more frequently if clinically indicated. In addition to central review, an investigator assessment was also made at each evaluation. During the open-label period, only investigator assessments were made.
Safety and tolerability were assessed by analysis of adverse events, physical examinations, vital signs, ECOG performance status, and laboratory assessments, on days 1 and 15 of each treatment cycle for the first 6 cycles. Cardiac function was assessed with 12-lead electrocardiogram at screening, day 1 of the first two treatment cycles (and subsequent cycles at the discretion of the investigator), and at treatment end. Severity of adverse events was rated by investigators according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE) version 4·0.
The primary endpoint was PFS per modified Response Evaluation Criteria In Solid Tumors (RECIST) 1·1, assessed by central radiology reviewer blinded to patient data. The prospectively defined RECIST modifications, which were unique to this study and developed to apply specifically to GIST, included the following criteria: (a) no lymph nodes were chosen as target lesions; enlarged lymph nodes were followed as non-target lesions; (b) no bone lesions were chosen as target lesions; (c) positron emission tomography was not acceptable for radiological evaluation. A progressively growing new tumour nodule within a pre-existing tumour mass had to meet the following criteria to be considered as unequivocal evidence of progression on our modification to RECIST 1·1: (a) the lesion was at least 2 cm in size and definitively a new active GIST lesion (e.g. enhancing with contrast or other criteria to rule out artefact); or (b) the lesion had to be expanding on at least two sequential imaging studies. The blinded central radiology review was performed according to a prospectively agreed central imaging charter and conducted by an external imaging contract research organisation. Two readers reviewed the images. Adjudication was used when only one reader assessed a progression or when the date of progression was discordant between the two independent readers.
Secondary endpoints included overall survival (OS), time to progression (TTP), objective response rate (ORR), and disease control rate (DCR), defined as rate of complete response or partial response (PR) plus stable disease (SD) lasting for at least 12 weeks. Exploratory endpoints (not reported here) included health-related quality of life, pharmacokinetics, and biomarker evaluation.
With 199 patients randomised, assuming a target treatment effect of 100% improvement in PFS, a randomisation ratio of 2:1 (regorafenib to placebo), a one-sided alpha of 0·01, and a power of 0·94, 144 events were needed for the final PFS analysis. A preplanned interim analysis of OS was performed at the time of the final PFS analysis.
Randomisation and masking
Patients were randomly assigned on a 2:1 basis to regorafenib or placebo using a computer-generated randomisation list prepared by the study sponsor (preallocated block design). Investigators received the randomisation number for each participant through an interactive voice response system (IVRS), which was also used to manage study drug supply.
Randomisation was stratified by (1) treatment line (failure of prior imatinib and sunitinib [“true third-line”] versus failure of prior imatinib, sunitinib, and other GIST therapies) and (2) geographical region (Asia versus rest of world).
Randomisation was blinded so that neither the patient, nor the investigator, nor the sponsor knew which agent was being administered. To maintain blinding, study medication was labelled with a unique drug pack number pre-printed on each bottle, which was assigned to the patient through the IVRS. Unblinding for individual patients could occur via the IVRS for emergencies; serious adverse events did not necessarily precipitate immediate unblinding.
Statistical analyses were performed using Statistical Analysis System software version 9·1 or higher (SAS Institute Inc, Cary, NC, USA).
The null hypothesis that both treatment arms would have the same PFS distribution was tested against the alternative hypothesis that the distribution of PFS times in the regorafenib arm would differ from that of the control arm.
PFS and OS estimates were calculated using the Kaplan–Meier method. Hazard ratios (HRs) and 95% confidence intervals (CIs) were derived from a Cox proportional hazard model with stratified log-rank test. ORR and DCR were analysed using the Cochran–Mantel–Haenszel test.
Role of funding source
The study sponsor (Bayer HealthCare) provided regorafenib and matching placebo, and collaborated with the principal investigator, G. D. Demetri, and an international steering committee of academic investigators on protocol design, data collection and interpretation, and preparation of this report. All logistical study operations were managed by the sponsor. Data were collected by the sponsor and analysed by the principal investigator, steering committee, and sponsor. All authors had full access to all data and vouch for the accuracy and completeness of the data presentation and analysis. The authors had final responsibility to submit for publication.