The approval of eflornithine (brand name Iwilfin) for pediatric high-risk neuroblastoma offers an interesting example of how the FDA evaluates single-arm trials supported by external control arms (ECAs). External controls are increasingly common in oncology and rare disease drug development, but they remain a source of careful regulatory scrutiny. In this case, the FDA’s assessment shows how comparability, study design, and interpretability are weighed when historical patient data served as the primary comparator to a novel therapy.
Eflornithine is an FDA-approved oral medication (Dec, 13th 2023) used to reduce the risk of relapse in pediatric and adult patients with high-risk neuroblastoma. Eflornithine was studied in Study 3(b), a single-arm trial designed to assess the product’s ability to reduce relapse after standard multimodal therapy, including immunotherapy. To contextualize efficacy in the absence of a concurrent randomized control, the sponsor proposed an externally controlled trial (ECT), using patient-level data from the National Cancer Institute (NCI)/Children’s Oncology Group trial ANBL0032.
ANBL0032 had originally been a multi-center, randomized, open-label trial comparing standard therapy with immunotherapy plus cis-retinoic acid (cis-RA) in newly diagnosed patients with high-risk neuroblastoma. The historical trial had shown a significant benefit for immunotherapy in both event-free survival (EFS) and overall survival (OS), ultimately leading to its adoption as standard of care and the approval of dinutuximab in 2015.
After the careful evaluation of the design and comparability of the ECA, the Agency agreed with the overall structure of the ECT while highlighting some limitations.
As FDA reviewers noted: “ FDA agrees that the historical control rates for Strata 1 and 2 were pre-specified as described and known prior to ECT initiation. For Stratum 1, the K-M estimate for EFS at 24 months was 0.85 (95% CI: 0.76, 0.90). For Stratum 2, the K-M estimate for EFS at 24 months was 0.46 (95% CI: 0.29, 0.61). However, FDA considers the comparisons of EFS and OS results with benchmark historical control rates to be exploratory only. In general, FDA considers the results of time-to-event endpoints in single arm studies to be uninterpretable, due to the lack of a comparator which would allow for attribution of the treatment effect to the investigational agent of interest.
It is notable that the results of Study 3(b) were published in 2018 (Sholler et al), and thus known prior to the design of the externally controlled trial. Therefore, the design of the ECT could not be prespecified, which FDA considers to be a limitation. However, FDA provided recommendations on the development of the statistical analysis plan which may reduce the potential for bias introduced by the sponsor’s knowledge of Study 3(b) and ANBL0032 results.”
The FDA noted a key limitation of the ECA: the results of Study 3(b) had already been published in 2018 (Sholler et al), meaning that the historical outcomes were known prior to the design of the ECT. This meant the ECT could not be fully pre-specified in a blinded manner. The Agency acknowledged this limitation but also indicated that careful planning of the statistical analysis could help mitigate potential bias introduced by prior knowledge of Study 3(b) and ANBL0032 results.
This assessment illustrates a central regulatory principle: in the absence of a contemporaneous comparator, time-to-event outcomes from a single-arm study are difficult to interpret, as it is challenging to determine whether observed effects reflect the investigational therapy or the natural course of the disease. FDA recognized that the external control derived from Study ANBL0032 provided a benchmark for contextualizing Study 3(b) results, while emphasizing that comparisons of EFS and OS with historical rates remain exploratory.
The FDA review further detailed the selection of patient populations for both the investigational and ECAs. Of 140 patients enrolled in Study 3(b) who received eflornithine, 87 had received prior immunotherapy on ANBL0032 and were included in the investigational analysis. An additional 5 patients treated per protocol but not previously enrolled in ANBL0032 were also included, resulting in 92 patients for comparative analysis. For the ECA, 1,328 patients received upfront therapy including immunotherapy, of whom 516 were included in the matched analysis after exclusion due to missing covariate data.
A key regulatory concern was comparability of the populations. FDA reviewers carefully evaluated whether the ECA could serve as a meaningful comparator. Their assessment highlighted both strengths and limitations: “Given that patients who enrolled Study 3(b) elected to enroll on an additional clinical trial, whereas patients on the control arm did not, it is possible that the groups differed in ways that may be associated with differences in outcome, such as social determinants of health (SDOH), performance status, or other related unmeasured factors. Since literature suggests that exposure to household poverty has an impact on patient outcomes (i.e., EFS and OS) (Bona, 2021) FDA considered analyses to assess the potential impact of imbalances in SDOH between the investigational and external control arm on the treatment effect. Data regarding cytogenetics were limited in ANBL0032. All the chromosomal alterations/genetic mutations recorded in patients in the DFMO arm (i.e., 1p deletion, 11q deletion, 17q gain, ALK mutations) are associated with worse prognosis; however, these data were not available for patients on the control arm. Sensitivity analyses performed by FDA assessed the potential impact of these unmeasured confounding factors on the estimated treatment effect in the ECT.
FDA considers the populations of matched patients in the external control arm and the investigational arm to be adequately comparable to permit an inferential analysis while acknowledging the need to evaluate the impact of unmeasured confounders using statistical analyses, where feasible.”
This evaluation reflected the balance regulators must consider: while the external control provided a framework for interpretation, residual confounding, especially from unmeasured variables such as geography, SDOH, cytogenetic features, and performance status, remained a concern. For example, some chromosomal alterations and ALK mutations, which have prognostic significance, were recorded in the investigational arm but not available for control patients. Additionally, 14% of control patients were enrolled outside the U.S., raising potential differences in care delivery and outcomes. FDA reviewers requested sensitivity analyses to assess whether these factors materially affected the estimated treatment effect.
Another consideration was the completeness of response assessments. While tumor imaging was largely complete, approximately 10% of control patients lacked imaging to verify response, and 25% had missing bone marrow evaluations at the end of immunotherapy. FDA noted that the INRC 1993 criteria do not require bone marrow evaluation for overall response determination, and all patients in the ECA with missing marrow assessments were documented as complete or very good partial responders. Such details illustrate the meticulous scrutiny the Agency applies to ensure that endpoint definitions are comparable across datasets.
The FDA’s review illustrates how ECAs are interpreted in context rather than in isolation. While acknowledging that time-to-event endpoints in single-arm studies are inherently limited, the Agency concluded that the matched external control population allowed for meaningful inferential analyses, particularly when sensitivity analyses were applied to assess the impact of unmeasured confounding. The review emphasizes a critical regulatory principle: the acceptability of external controls depends on data reliability, transparency in patient selection, and rigorous evaluation of comparability.
In summary, the evaluation of eflornithine’s ECA demonstrates the FDA’s careful balancing of scientific rigor and practical considerations in rare disease drug development. It highlights how regulators weigh the strengths of historical data against residual uncertainty, use methodological guidance to mitigate bias, and require sensitivity analyses to probe potential confounding. For developers, this case illustrates that external controls can be informative for regulatory decision-making, but only when the Agency can verify comparability and appropriately account for limitations.
Reference:
NDA/BLA Multi-disciplinary Review and Evaluation - Application number: 215500Orig1s000
https://www.accessdata.fda.gov/drugsatfda_docs/nda/2024/215500Orig1s000MultidisciplineR.pdf
