INFORM2 NivEnt: The first trial of the INFORM2 biomarker driven phase I/II trial series: the combination of nivolumab and entinostat in children and adolescents with refractory high-risk malignancies

INFORM2 NivEnt is an exploratory non-randomized, open-label, multinational and multicenter seamless phase I/II trial of nivolumab and entinostat in children and adolescents with relapsed / refractory or progressive high-risk solid tumors and CNS tumors. A schematic overview of the trial is provided in Fig. 1. Since the age spectrum comprises a broad range of physical development of the study patients, phase I is divided in 2 age cohorts: 12–21 years and 6–11 years (in the absence of a pediatric liquid formulation for entinostat, no patients younger than 6 years can be enrolled). Although one could argue that for adolescents no dose escalation is necessary since there is no biological reason to expect another RP2D [34], it was decided to include a dose escalation for adolescents since the available adult data is still immature. The separate dose escalations for children 6–11-year-old and adolescents 12–21-year-old was included to fulfill regulatory requirements. The two age cohorts will run independently and follow a 3 + 3 design with two dose levels [35]. With only two dose levels, a model-based phase I design such as the Continual Reassessment Method was considered but would not offer any advantage to the traditional 3 + 3 design. RP2D is the dose at which up to 1 of 6 patients experiences dose limiting toxicity (DLT). Assuming more similarity with adult data, the trial will start with the older age cohort. The younger age cohort starts if either of the following events occur: 0 of 3 patients of the older age cohort experience a DLT in the DLT observation period OR a maximum of 1 of 6 patients of the older age cohort experience a DLT in the DLT observation period. Patients entering the trial on RP2D can seamlessly enter phase II. In order to maximize the chance of detecting early signs of activity of the combination treatment, phase II consists of a biomarker defined four group (A - D) phase II basket trial. If patients harbor more than one biomarker fitting group A, B or C, they will be allocated according the following priority: A: highest, B: intermediate and C: lowest priority. The aim is to identify the biomarker cohorts in which the combination treatment shows promising activity, i.e. a response rate exceeding the historical response rate of p₀ = 12% which was observed in a comparable pediatric patient population treated in phase I/II trials [3]. A response rate of p₁ = 30% would be considered as the target response rate to be detected with high probability.

A Bayesian adaptive design will be used with the objective of stopping cohorts early which are showing no evidence of activity (stopping for futility) [36‐38]. Interim analyses will be based on the Bayesian posterior probability distribution of the best response rate, r. Futility will be declared if the posterior probability that r is ≥ p₁ is smaller than c_Futility. Activity will be declared if the posterior probability that r is ≥ p₀ is larger or equal to c_Activity. The Bayesian model will be based on the Beta-binomial conjugate family. For futility, the prior is chosen as an optimistic Beta(p₁, 1-p₁), and for activity the choice is accordingly a sceptical Beta(p₀, 1-p₀). Every cohort will be evaluated separately. An interim analysis is planned after 10 patients. If recruitment allows, further interim analyses are planned every 10 patients. Based on the evaluation of the operating characteristics of the trial design, c_Futility was selected as 0.01 and c_Activity. as 0.90 [37]. The resulting decision rules for futility and efficacy in terms of number of observed responders among enrolled patients are shown in Fig. 2. The probability of stopping for futility at first interim when the treatment is not active is at least 27.8% whereas it is only 2.8% for an active treatment with true response rate of 30%.

Three scenarios are possible following interim analysis:

Recruitment and treatment of patients will be performed in 5 Society of Pediatric Oncology and Hematology (GPOH) academic phase I/II centers in Germany and additional ITCC (http://​www.​itcc-consortium.​org/​) academic centers in Stockholm (Sweden), Utrecht (The Netherlands), Paris (France) and Vienna (Austria) together with the academic centers in Sydney, Melbourne and Perth (Australia). An up to date full list of study sites can be obtained on ClinicalTrials.​gov, NCT03838042. The protocol (available on request) was or will be approved by all respective competent authorities and ethics committees at all participating institutions, and written informed consent will be requested from all patients and/or legal representatives.

Key inclusion criteria include children and adolescents with refractory/relapsed/progressive high-risk solid and CNS tumors with no standard of care treatment available, age ≥ 6 to ≤21 years. Determination of biomarkers for patient stratification into group A - D is not part of this trial and will be performed in laboratories complying with DIN EN ISO/IEC 17025 or similar. For several countries, biomarkers are routinely determined according the INFORM registry molecular pipeline [13], which has been technically validated. Equivalently valid molecular pipelines are accepted. The following methods are routinely applied as described previously [13]: whole exome sequencing, low coverage whole genome sequencing, RNA sequencing, and DNA methylation array [39]. To try to ensure that molecular profiling matches the current tumor episode as closely as possible (i.e. to minimize the time in which tumor molecular characteristics may have evolved), enrollment within 12 weeks (this also includes the time for the molecular analyses and allows patient transfer to trial sites, wash out etc.) after biopsy/puncture/resection is requested (see Fig. 1). A complete overview of the inclusion- and exclusion criteria is outlined in the Additional file 1.

In this trial, nivolumab will be administered using the approved dose for adults, which is also the pediatric RP2D (3 mg/kg every 2 weeks) [40]. An ongoing phase I trial with entinostat monotherapy in children identified a RP2D of 4 mg/m2 weekly [41]. The recommended adult dose regimen is 5 mg weekly – 10 mg every 2 weeks (which equals 2.9 mg/m2 weekly – 5.8 mg/m2 every 2 weeks). The combination of entinostat, nivolumab ± ipilimumab was considered safe and tolerable in adults [42]. In the phase I part of the trial, entinostat has 2 dose levels: 2 mg/m2 and 4 mg/m2 (the latter reflecting the single agent pediatric RP2D). The dose escalation will be performed separately in both age groups.

After one so called priming week (to “prime” the immune system with entinostat before applying checkpoint inhibition) with 1 dose of entinostat on day 1, cycles of 4 weeks each will start with entinostat (orally) on day 1, 8, 15 and 22 and intravenous administration of nivolumab on day 1 and 15. Thereafter, every cycle follows the same schedule. There is no pause between cycles. All groups A – D will receive the same treatment.

The toxicity profiles of nivolumab and entinostat show a large overlap. Therefore, AEs will always be managed first by delaying nivolumab and entinostat administration together in combination with symptomatic treatment:

An overview about diagnostic and therapeutic measures, timing of disease assessment, and study visits is displayed in Additional file 4. In brief, routine visits are planned weekly during the first 2 cycles and every 2 weeks thereafter. Response assessments are scheduled every 8 weeks.

Phase I (dose finding) will recruit patients regardless of their biomarker status. Dose finding in the older age cohort will recruit 3–12 patients (if the younger age cohort is not allowed to open due to too much DLTs in the first 3 patients of the older age cohort, the total number of patients would be 3). If dose finding is initiated in the younger age cohort, it will also require 3–12 patients. This will result in 3–24 patients in total. Patients entering the trial on RP2D in the phase I can seamlessly enter phase II. Patients will be allocated to the four separate cohorts A – D. The sample size in cohorts A, B and C will be restricted by availability of the patients with the biomarkers during the 3-year study recruitment period. The upper limit of patients planned to be recruited to each of the cohorts A, B and C is 20–25. Since the actual patient number is not known exactly, expected type 1 error rate and power are reported averaged over the range of patient numbers to be expected in each cohort. Given the choice of the criteria c_Futility and c_Activity in the Bayesian adaptive design and assuming that patient numbers between 20 and 25 are equally likely, the expected type 1 error rate is 6.6% and the expected power is 76%. No second interim analysis is planned for cohorts A, B and C. In cohort D recruitment is expected to be faster and the sample size can be determined on the basis of power and 1-sided alpha-level, using the same criteria for futility and efficacy as in cohorts A, B and C. The smallest sample size such that alpha is ≤5% and power ≥ 80% is n = 29, for type 1 error is 5.0% and power is 80.3%. The upper limit of patients planned to be recruited to cohort D is therefore 29. A second interim analysis is planned after 20 patients for this cohort. Analytical calculations of the operating characteristics were derived [38] and power calculations were made with the R package BDP2 Version 0.1.3 (from CRAN (https://​CRAN.​R-project.​org/​package=​BDP2)) which is the core of the web tool BDP2 workflow (http://​biostatistics.​dkfz.​de/​BDP2/​).