Editorial 

Trials Terminated Early: When Is Enough, Enough?

Alison J. Huang, Roger J. Lewis

JAMA Published Online: August 5, 2026

doi: 10.1001/jama.2026.8878

What does it mean to stop a clinical trial early? In the early days of the modern randomized trial, it meant the research had not gone according to plan.¹ Trialists were expected to adhere to the sample sizes they originally proposed based on early assumptions about underlying event rates, loss to follow-up, and variability in outcomes. When a trial was stopped prematurely, it indicated that foundational assumptions had proven incorrect, critical safety or protocol issues had emerged, or logistical challenges had grown insurmountable. Early trial termination carried an implicit presumption of failure: that the trial had been inadequately planned, poorly executed, or simply overtaken by circumstances beyond investigators’ control.

The 1960s and 1970s saw the development of systematic processes for weighing decisions about early trial termination, driven by growing recognition of trialists’ ethical obligations to participants and to the integrity of the evidence being generated. Data and safety monitoring boards (DSMBs) were tasked with evaluating accumulating unmasked trial data and making recommendations about termination based on evidence of early benefit or harm from interventions, free from conflicts of interest affecting investigators or sponsors.² Formal methods were also developed to manage the cumulative risk of a false-positive conclusion arising from repeated interim analyses of trial outcomes data, recognizing that each additional examination of data increased the probability of a spurious positive finding.^3,4 Together, these developments gave the field tools for making early stopping decisions in a principled rather than an ad hoc manner.⁵

Yet, suspicion about early trial termination persisted. Investigators feared that early termination, even when procedurally rigorous, could lead to biased estimates of intervention effects, particularly when trials were stopped for early evidence of benefit. More broadly, the trial community remained wary that any systematic approach to modifying a trial once it was underway might compromise the validity of its findings, even unintentionally. Early stopping, however justified it might seem in individual cases, carried a reputational burden that made many reluctant to embrace it as a legitimate part of trial design rather than a last resort.

Today, however, the landscape has evolved. One in 4 registered trials is discontinued before reaching its originally anticipated sample size, indicating that early termination is a relatively commonplace event.^6-8 At the same time, there is wider recognition that the assumptions about disease event rates or outcome variability on which trial sample sizes are initially based are uncertain—and often prove incorrect. Thus, thoughtful interim plans to reevaluate trial size may not only be prudent, but also responsible—a feature of well-designed trials rather than a departure from them.

This recognition may require a fundamental reframing of the underlying question. Rather than asking “When should a trial be stopped early?”—a formulation that treats early termination as an exception requiring special justification—the more appropriate question may be “What is the optimal point at which to stop a trial?”⁹ To answer this new question, trialists must balance multiple competing obligations: the statistical implications of repeated interim analyses, the responsibility to current participants to minimize exposure to ineffective or harmful interventions, the obligation to future patients to generate reliable and interpretable evidence, and the expectations of regulators and funders. The key challenge for investigators is to determine, prospectively and transparently, the conditions under which stopping a trial before reaching the maximum planned sample size represents the most scientifically and ethically defensible choice—not a departure from good trial conduct, but a fundamental expression of it.

This month in JAMA+ Trials, we examine multiple recent JAMA Network trials that were discontinued before enrolling the maximum planned number of participants. These studies illustrate the importance of early termination decisions, the value of prespecified stopping rules, and the nuances of interpreting results from terminated studies. Taken together, they indicate that termination of a trial before its original maximum planned sample size is neither inherently good nor inherently problematic—rather, its implications depend on the circumstances that prompted it, the rigor and transparency with which the decision was made, and the honesty with which its consequences for interpretation are acknowledged.

In the DOT-MS noninferiority trial evaluating the appropriateness of discontinuing first-line disease-modifying therapy for multiple sclerosis in patients with long-term stable disease,¹⁰ the DSMB recommended stopping when the trial had reached 68% of its originally planned sample because the rate of inflammatory disease activity was higher than the predefined limit in the discontinuation group. As the investigators noted in JAMA Neurology, evidence of harm appeared robust, with 8 of 45 participants in the discontinuation group vs 0 of 44 in the continuation group having recurrent inflammation. However, other prespecified analyses, including noninferiority testing and cost analyses, could not be performed due to the trial’s more limited sample size and incomplete follow-up. Although not a failure of trial conduct, this was an unavoidable consequence of responsible early stopping, which readers must weigh when interpreting the findings.

In a trial examining management of patent ductus arteriosus in preterm infants published in JAMA,¹¹ the investigators hypothesized that expectant management (no treatment unless the infant developed cardiopulmonary compromise) would result in a lower incidence of death or bronchopulmonary dysplasia (the composite primary outcome) than proactive treatment. The investigators originally planned to enroll 836 patients, with interim analyses preplanned after outcomes data became available for 25%, 50%, and 75% of infants. At the 50% interim analysis, the conditional power¹² for detecting a difference in the primary outcome was estimated at only 2%—far below the prespecified futility threshold of 15%—suggesting that the trial had little realistic prospect of detecting a meaningful difference in its primary outcome even if enrollment had continued to completion. The DSMB therefore recommended stopping on futility grounds.¹³

However, the incidence of death alone (a prespecified secondary outcome) was more than twice as high in the proactive treatment group as in the expectant management group. Coupled with evidence of higher mortality with proactive treatment from other smaller trials, this trial may persuade many clinicians to refrain from proactive treatment of patent ductus arteriosus. Thus, the trial illustrates how a trial stopped for futility on its primary end point can nonetheless yield evidence that influences clinical practice, especially when considered in the context of the totality of the evidence.¹⁴

In the HEALEY ALS platform trial evaluating the efficacy of the C5 complement–inhibitor zilucoplan among other experimental treatments for amyotrophic lateral sclerosis, interim futility analyses were prespecified after each treatment regimen accumulated 40 randomized participants. As described in JAMA Network Open,¹⁵ investigators planned to stop a treatment regimen if interim analysis indicated less than 5% probability that the treatment slowed disease progression. The zilucoplan regimen was discontinued approximately 5 months earlier than it would have been if the maximum sample had been reached.

The investigators, despite their disappointment, noted that early termination spared participants a total of 279 trial visits and 14 620 doses of investigational product.¹⁵ For a disease in which eligible participants are limited, these savings also freed participants to enroll in other investigational studies—a benefit of rigorous futility planning that extends beyond the individual trial.¹²

In contrast to the preceding examples, a trial evaluating the Janus kinase 1 (JAK 1) inhibitor filgotinib for treatment of noninfectious uveitis published in JAMA Ophthalmology was discontinued when it achieved less than 30% of its originally planned sample size for reported “business reasons” only.¹⁶ Readers speculated that early termination was prompted by the US Food and Drug Administration’s contemporaneous refusal to approve filgotinib for the alternate indication of rheumatoid arthritis, leading the drug’s manufacturer to conclude that further investigation was unlikely to be profitable.¹⁷ Given early trial data suggesting that filgotinib lowered the risk of treatment failure, some lamented that the trial was stopped early when it could have provided more insight into the effects of JAK 1 inhibitor therapy for uveitis.¹⁷

This example raises a challenging question: was discontinuation an instance in which commercial interests foreclosed potentially scientific inquiry or was it instead a responsible conservation of sponsor resources? Participants who enroll in trials do so with an expectation that their contribution will generate knowledge and a business decision that forecloses that expectation—rather than a scientific one—may sit uneasily with many. Yet one might argue that exposing participants to an experimental therapy with no realistic path to commercial availability may be ethically problematic in itself. Could continued burden or risk for participants be justified if the intervention had no prospects for being made available to future patients?

Collectively, these and other terminated trials^18-27 illustrate a spectrum of circumstances leading to termination before trials attain their maximum planned sample size, as well as the value that such terminated trials can still offer to clinicians, patients, and the research community. The difference between responsible and problematic early termination does not hinge simply on the reason given, but the degree to which stopping decisions were anticipated in the trial’s design, reported transparently, and accompanied by honest acknowledgment of their consequences for interpretation. Rather than positioning early termination as the opposite of trial success, the field should recognize that a trial stopped at the right time, for the right reasons, and reported with candor can contribute as much to medical knowledge as one that reaches its originally planned completion.