The JASMINE trial was designed as a multicenter, randomized Phase III registration trial, registered as jRCT2031210384 and conducted under the Japanese Good Clinical Practice framework. Its purpose was to evaluate the safety and efficacy of an exclusive human milk diet, a nutritional strategy in which very low birth weight infants receive human milk-based fortification rather than products derived from bovine milk. The population under study, infants born weighing very little, is among the most fragile in medicine, and nutritional decisions made in the first weeks of life can influence growth, feeding tolerance, and vulnerability to serious gastrointestinal disease. Because the intervention involved distinct nutritional products requiring different preparation and administration procedures, the investigators state that blinding of the treating clinicians was not practically feasible, a constraint that shaped the entire evidentiary architecture of the study.

Recognizing that an open-label design invites the possibility of observer bias, the trial incorporated a series of structural safeguards. Randomization was performed centrally using a permuted-block scheme stratified by study site and birth weight, with allocation concealment maintained until the moment of assignment. Clinical management across the participating neonatal intensive care units followed predefined protocols, and the major outcomes, including body weight, time to full enteral feeding, and adverse events, were prospectively defined before the trial began. Although blinded endpoint adjudication was not incorporated into the design, the trial operated under GCP-compliant monitoring that included on-site source data verification, a process in which monitors check that recorded data match the original clinical records. Mizuno’s response is explicit that the open-label design and the absence of blinded endpoint adjudication are genuine limitations of the trial and should be considered when interpreting outcomes that may be influenced by clinical management decisions.

The sharpest questions in the correspondence concerned the involvement of Prolacta Bioscience, the company that funded the trial and whose product was under evaluation. Mizuno’s response draws a distinction that he argues is essential for readers: the fact of sponsor funding is separate from the question of the sponsor’s role in the analysis and interpretation of the results. Industry sponsorship, he notes, is a standard model for pivotal Phase III trials intended to support regulatory approval, because such trials demand substantial resources for multicenter conduct, data management, pharmacovigilance, quality assurance, statistical analysis, and the preparation of regulatory submissions. As disclosed in the published article, Prolacta funded the study and contributed to its statistical design, but the statistical analyses of the trial data were performed by a third-party contract research organization, Innovative Analytics, working to a prespecified statistical analysis plan rather than by Prolacta personnel.

The statistical workflow described in the response is unusually detailed for a journal correspondence. The protocol and the statistical analysis plan were finalized before database lock and before any unblinded efficacy analysis, and the plan prespecified the primary endpoint of weight gain velocity, the non-inferiority margin, the sample size calculation, the intent-to-treat and per-protocol analysis sets, and a testing hierarchy in which claims of superiority could be tested only after non-inferiority had been established. Data management and monitoring were conducted through the contract research organization in accordance with GCP, including source data verification at participating sites. The analyses themselves were executed in SAS version 9.4 by Innovative Analytics. Sponsor personnel did provide data checking and statistical support during manuscript development, but, according to the response, Dr. Lee did not determine the primary endpoint, perform the primary or secondary analyses, alter the prespecified plan after examining the results, or select alternative analyses for publication; his involvement occurred after the trial analyses had been completed, and scientific interpretation and the decision to submit remained with the named authors.

Two admissions in the response stand out for their candor. First, the statistical analyses were not performed with the analysts masked to treatment groups, a limitation that the prespecification of methods, endpoints, analysis populations, and testing hierarchy was intended to mitigate but could not eliminate. Second, and stated with unusual directness, there was no separate replication of the primary statistical analyses by an external academic statistical group. Mizuno writes that the team wishes to be explicit about this point. In partial compensation, the trial was subject to a form of external scrutiny that few academic studies undergo: as a Phase III registration trial under Japanese GCP, JASMINE was inspected by the Pharmaceuticals and Medical Devices Agency, and the trial data, clinical study report, and statistical evidence were evaluated by the agency as part of the regulatory review supporting marketing authorization in Japan. Mizuno acknowledges that regulatory oversight and independent academic reanalysis are distinct things, but argues that the inspection and review provided meaningful external checks on the conduct, data integrity, and reporting of the trial.

The response also addresses a subtle disclosure question raised by the correspondents: the competing-interest statement in the published article referred to the individual competing interests of the named authors, while the sponsor’s financial and operational roles were reported separately in the Funding and Acknowledgments sections. Those separate disclosures covered the sponsor’s funding of the trial, Prolacta’s contribution to the statistical design, and the involvement of sponsor personnel during manuscript development. Mizuno agrees that these roles should be considered together when readers assess the independence of the evidence, and he frames the entire response as an opportunity to make the respective roles of the sponsor, the contract research organization, and the academic investigators more explicit. The episode illustrates a recurring tension in biomedical publishing: formal compliance with disclosure conventions can still leave readers uncertain about who did what, and correspondence of this kind often does the work of clarifying the division of labor that the original paper left implicit.

The second front of the dispute concerns a subsequent cost-utility analysis of the exclusive human milk diet in Japan, authored by Igarashi, Reyes, and Mizuno. The correspondents correctly noted that JASMINE observed only a single case of necrotizing enterocolitis, a devastating inflammatory bowel disease of prematurity, making it statistically inappropriate to derive a robust treatment-effect estimate for that condition from the trial alone. The health economic model therefore relied on external published evidence: the relative risks for the modeled complications, including the necrotizing enterocolitis relative risk used in the base-case analysis, were drawn from a published modeling study by Scholz and colleagues, with assumptions and data sources reported in the publication. Mizuno emphasizes that the necrotizing enterocolitis estimate in the model was an external modeling input, not an efficacy estimate derived from JASMINE, and that the economic analysis should therefore not be interpreted as demonstrating a treatment effect for that condition from the trial. The authors of the economic analysis also disclosed their financial relationships, with two authors reporting payments from Prolacta Biosciences to participate in the independent research and one reporting no disclosures.

The correspondents further observed that the necrotizing enterocolitis relative risk used in the base-case model differed from the estimate reported in the 2019 Cochrane review of human milk fortification. Mizuno attributes this difference to the use of different evidence sources rather than any trial-derived claim, and points to the uncertainty analysis built into the economic model: the robustness of the economic conclusions was tested through deterministic and probabilistic sensitivity analyses, including 10,000 Monte Carlo simulations, and through a scenario analysis using relative risk estimates from the largest meta-analysis of exclusive human milk diets available at the time, a 2024 meta-analysis published in Nutrients. Sensitivity analyses of this kind are the standard toolkit of health economics, allowing modelers to show whether a cost-effectiveness conclusion survives when uncertain inputs are varied across plausible ranges, and their presence here is offered as evidence that the economic findings were not hostage to any single assumption about necrotizing enterocolitis risk.

The broader significance of this exchange lies in what it reveals about the anatomy of trust in industry-funded clinical research. The primary objective of JASMINE was to evaluate the efficacy and safety of an exclusive human milk diet in Japanese very low birth weight infants within a regulatory framework, and the trial was randomized, conducted under GCP, monitored with source data verification, analyzed to a prespecified plan by a third-party organization, inspected and reviewed by the national regulator, and subsequently peer reviewed. Yet Mizuno concedes the open-label limitations, the unblinded analysts, and the absence of independent academic reanalysis, and argues that transparency about the respective roles of sponsor, contractor, and investigators is what allows readers to judge the evidence for themselves. Whether that standard of transparency satisfies critics will shape not only the reception of this trial but the ongoing scientific discussion about optimal nutritional strategies for vulnerable preterm infants, a debate in which every assumption, disclosure, and statistical choice is scrutinized by clinicians, regulators, and the families whose children stand to benefit or be harmed by the answer.

Subject of Research: Methodological and conflict-of-interest defenses of the JASMINE Phase III trial of exclusive human milk diets in very low birth weight infants

Article Title: Response to “Trial design, conflict of interest, and evidence independence in the JASMINE trial and its economic evaluations”

Article References: Mizuno, K. (2026). Response to “Trial design, conflict of interest, and evidence independence in the JASMINE trial and its economic evaluations”. Journal of Perinatology. https://doi.org/10.1038/s41372-026-02929-x

Image Credits: AI Generated

DOI: 10.1038/s41372-026-02929-x

Keywords: JASMINE trial, exclusive human milk diet, very low birth weight infants, necrotizing enterocolitis, conflict of interest, open-label trial, Good Clinical Practice, PMDA, cost-utility analysis, Prolacta Bioscience, neonatal nutrition, statistical analysis plan