What Is an OLE Clinical Trial and How Does It Work?

An open-label extension, or OLE, is a follow-up study that lets participants in a completed clinical trial keep receiving the drug being tested, sometimes for years after the original trial ends. Everyone in the OLE knows exactly what they are taking, including those who were on placebo during the blinded phase, and the primary goal is usually to track long-term safety rather than to prove the drug works. OLE trials have become one of the most common tools in drug development for answering a question that short, tightly controlled studies cannot: what happens when someone stays on a treatment for a long time?

Why OLE Studies Exist

Most randomized controlled trials last months, occasionally a year or two. That is enough to show whether a drug has an effect but often not enough to reveal rare side effects or understand how the benefit holds up over time. OLE studies are designed to fill that gap by accumulating more patient-years of exposure under reasonably monitored conditions. The stated objective of most OLEs is to collect long-term safety and tolerability data.¹ When the primary purpose is genuinely to deepen the safety profile of a new medicine, researchers have argued these studies play a legitimate and useful role in drug development.²

There is also a practical incentive. Patients who responded well during the blinded trial often want to keep taking the drug, particularly when alternatives are limited. And drug developers want to maintain the relationship with those patients, both to gather data and to build toward regulatory approval. Growing interest in real-world evidence for regulatory decisions has increased the importance of understanding how these uncontrolled extensions can be used alongside external comparisons.³

How the Transition From Blinded to Open-Label Works

When a randomized, placebo-controlled trial ends, the study is “unblinded” and participants learn whether they were on the active drug or placebo. Those who choose to enter the extension are switched to the active treatment. For participants already on the drug, the dose typically stays the same initially. For those crossing over from placebo, the transition often involves a careful ramp-up period.

A trial of zonisamide in children with epilepsy illustrates the mechanics well. After the blinded phase, children who had been on placebo entered a transition period of two to eleven weeks, during which they were started on a low dose that was gradually increased to the target. The subsequent open-label period lasted roughly a year, and doctors could adjust dosing based on how each child responded.⁴ This flexibility is a defining feature of OLEs: once the controlled phase is over, clinicians often have more latitude to tweak doses or manage side effects as they would in routine practice.

The Selection Bias Problem

Not everyone from the parent trial enrolls in the extension, and who opts in is not random. Participants who tolerated the drug well and experienced a benefit are far more likely to continue. Those who had bad side effects or saw no improvement tend to drop out before the extension even starts. This means the OLE population is already filtered for people who do well on the drug, which can make results look more favorable than they would be for a general patient population.

Research has shown this effect can be substantial. In one analysis that modeled how patient selection bias inflates outcomes, an OLE study reporting a responder rate of 43% was re-estimated at just 28% after accounting for the filtering of patients who enrolled.⁵ That is a 15-percentage-point gap between what the extension appeared to show and what a less biased analysis suggested. Any time you see impressive-looking efficacy numbers from an OLE, this built-in selection effect is worth keeping in mind.

Retention during the extension itself adds another layer to this problem. In two OLE studies of dalfampridine for walking impairment in multiple sclerosis, 269 patients entered one extension and only 154 completed it over up to five years; in the other, 214 entered and 146 finished over about three years.⁶ When roughly 30 to 40 percent of participants drop out, the people who remain are an increasingly select group, and outcomes measured only among completers will overestimate the drug’s real-world performance.

How Knowing What You Are Taking Changes Things

In a blinded trial, neither you nor your doctor knows whether you are on the drug or placebo, which helps keep expectations from coloring the results. Once the study goes open-label, everyone knows. That awareness can shift how patients perceive their symptoms and how clinicians rate outcomes. Patient expectations, the type of treatment, and interactions between doctors and patients have all been identified by stakeholders as contributors to open-label bias, particularly in patient-reported outcomes such as pain, fatigue, and quality of life.⁷

A related phenomenon shows up with biosimilars, the near-copies of biologic drugs. When patients are switched from a brand-name biologic to a biosimilar in an open-label setting, they report more side effects and discontinue more often than patients switched under blinded conditions. One systematic review found that median discontinuation rates for any reason were about 14% in open-label biosimilar studies compared with roughly 7% in double-blinded ones; discontinuation specifically for adverse events was about 6% versus 3%.⁸ The drugs were essentially the same, but knowing about the switch made patients more likely to experience and report problems. This “nocebo effect” is a vivid demonstration of how unblinding alters outcomes, and it applies directionally to OLE studies as well, where patients may feel better partly because they know they are now receiving the real drug.

The Delayed-Start Design

Researchers have developed a clever way to extract more information from the extension phase. In a delayed-start analysis, you compare two groups at a point well into the OLE: the “early-start” group, who received the active drug from the beginning of the parent trial, and the “delayed-start” group, who were on placebo initially and only began the drug when the extension started. If the early-start group still has a measurable advantage even after the delayed-start group has been on the drug for months, it suggests the benefit is not just symptomatic relief but something more durable. It may point toward disease modification, meaning the drug is actually slowing the underlying condition rather than just masking symptoms.

This approach has been used in several high-profile programs. In Friedreich’s ataxia, the MOXIe extension study compared scores on a neurological rating scale at 72 weeks in the open-label period, looking at whether patients who had started omaveloxolone earlier maintained an advantage over those who started later.⁹ In Alzheimer’s disease, the Clarity AD extension of lecanemab found that the separation between early-start and delayed-start groups was maintained between 18 and 36 months, and patients treated from the beginning continued to show benefits across clinical and quality-of-life measures.¹⁰

The logic is compelling, but it is not bulletproof. The delayed-start design still suffers from the same selection and dropout biases as any OLE. If sicker patients in the delayed-start group disproportionately dropped out, the surviving comparison could be misleading. Still, it is one of the better tools available for teasing apart symptomatic effects from disease-modifying ones in settings where running another multi-year randomized trial would be impractical or unethical.

What Happens When Placebo Patients Cross Over

One of the most interesting moments in an OLE is watching what happens to the former placebo group. If the drug genuinely works, people crossing over should start improving. The speed and magnitude of their improvement, relative to where the original treatment group was at the same point, can be informative.

In a lupus study, patients who had been on placebo during the blinded phase of a ustekinumab trial crossed over to the drug during the extension. By week 112, roughly 79% of these crossover patients achieved a standard clinical response, and 92% had a meaningful improvement on a disease activity score.¹¹ The group that had been on ustekinumab all along showed numerically similar or slightly higher response rates. These convergence patterns are reassuring, though they carry the usual OLE caveats about selection and expectation bias.

In rare neurodegenerative diseases, crossover data can be especially revealing. A study of deferiprone in pantothenate kinase-associated neurodegeneration found that patients who switched from placebo to the drug during the 18-month extension appeared to slow their rate of disease progression, while those who had been on deferiprone all along maintained a stable trajectory.¹² These are small patient populations where large randomized trials are nearly impossible to run, which makes the extension data, imperfect as it is, an essential part of the evidence base.

Missing Data and the Statistics of Dropout

When a third or more of participants leave before the extension ends, the data they would have contributed does not simply disappear from the analysis; its absence distorts what remains. The people who drop out are rarely a random sample. They may be the ones getting worse, the ones experiencing side effects, or the ones who moved and lost touch with the study site. Ignoring those gaps, which is what happens if you just analyze the available data, tends to paint a rosier picture than reality.

Statisticians have developed methods to handle this, and the choice of method can meaningfully change the results. One case study of a long-term open-label trial re-examined treatment satisfaction data using several different approaches for missing information. The original analysis had simply ignored missing values. The re-analysis used techniques that made different assumptions about why data was missing, from “missing for random reasons” to “missing because the patient was doing poorly.” Throughout these sensitivity analyses, the effect remained significant, providing more robust evidence than the original naive approach.¹³

Research on imputation methods for extension trials has also found that when the outcome of interest is binary, like “responded or did not respond,” it is better to impute the underlying continuous measurement and then classify the result, rather than directly imputing the yes-or-no outcome.¹⁴ These technical choices happen behind the scenes, but they have real consequences for whether an extension study’s results are trustworthy. When reading OLE data, it is worth checking whether the investigators tested their conclusions against different assumptions about why patients dropped out.

Tracking Biomarkers Over Time

Not all OLE endpoints are subjective. In diseases where biological markers can be measured, extensions offer a chance to watch what happens to those markers over long periods, including during any gap between the parent trial and the start of the extension.

The phase 2 lecanemab study in early Alzheimer’s disease provides a detailed example. During the treatment gap before the extension began, blood-based markers like the amyloid ratio and a tau-related protein began drifting back toward their pre-treatment levels, while amyloid levels measured on brain scans changed more slowly. Once the extension started and treatment resumed, dose-dependent reductions in brain amyloid reappeared, along with improvements in the blood markers.¹⁵ The fact that these objective markers responded in a predictable way, deteriorating during the gap and improving again with treatment, strengthens the case that the drug’s effect is real rather than a product of expectation or selection bias. Biomarker data is one of the most persuasive kinds of evidence an OLE can provide precisely because it is harder for knowing you are on the drug to influence a PET scan.

OLE Evidence in Neurodegenerative Disease

Rare and progressive neurological conditions have become a major setting for OLE studies, partly because the diseases worsen slowly and partly because running long placebo-controlled trials feels ethically fraught when there are few or no alternatives.

In ALS caused by SOD1 gene mutations, the open-label extension of a tofersen trial showed that patients who started treatment earlier had meaningfully better outcomes than those whose treatment was delayed by six months. Over 12 months, the early-start group showed a difference of about 3.5 points on a functional rating scale, roughly 9 percentage points better lung function, and early survival data suggesting a lower risk of death or need for ventilation.¹⁶ These numbers helped build the case for regulatory approval even though the original blinded trial had not met its primary endpoint on the functional scale. The extension data, combined with biomarker reductions, carried significant weight.

This pattern, where a blinded trial produces mixed or inconclusive results but the extension data fills in a more optimistic picture, is both the promise and the peril of OLE studies. The data is genuinely informative, but it is also generated under conditions that favor the drug. Regulators have to weigh both sides carefully.

Patient-Reported Outcomes in Long-Term Extensions

Beyond clinical measures and biomarkers, OLEs increasingly track how patients feel in their daily lives. Migraine studies have been a productive area for this kind of data. In a rimegepant trial, patient-reported quality-of-life measures and disability scores continued to improve through 64 weeks of open-label treatment, reaching levels considered clinically meaningful.¹⁷ In immune thrombocytopenia, an extension of the ADVANCE IV trial tracked fatigue, physical functioning, and disease-specific symptoms at regular intervals over the first year.¹⁸

Patient-reported outcomes are especially susceptible to open-label bias, since how you rate your own pain or fatigue is inevitably shaped by whether you believe you are on an effective treatment. But they also capture something that objective measures miss: whether a drug makes a noticeable difference in how someone lives. The tension between these two realities is one reason why OLE data on quality of life is often reported but rarely treated as definitive.

Challenges in Pediatric Populations

Running clinical trials in children adds layers of complexity, and OLEs are no exception. In pediatric multiple sclerosis, for instance, there are ethical concerns about exposing children to placebo given how active the disease can be, and recruitment has been a persistent challenge. Researchers in the field have questioned whether running a phase 3 trial for every drug approved in adults is even feasible for the pediatric MS population, and have suggested that study designs should be reconsidered based on whether the primary goal is proving efficacy or establishing safety.¹⁹ In this context, open-label extensions of adult trials, or smaller pediatric studies with long open-label follow-up, may be the most realistic way to generate safety data for younger patients.

The Cost and Design Debate

Traditional OLE studies use clinical trial infrastructure, study sites, regular visits, monitoring, and data collection, which makes them expensive to run over multiple years. As real-world data sources have become more accessible, some researchers have proposed hybrid designs that use electronic health records, pharmacy data, or direct-to-patient technology to follow patients after a trial ends, reducing the burden and cost while still capturing important long-term outcomes.²⁰

There is also a more fundamental criticism of OLEs: that the money spent on them could be better used running properly controlled long-term safety studies. Because OLEs lack a concurrent control group, they cannot cleanly distinguish drug effects from the natural course of the disease or from placebo-like improvements. Critics have argued that a well-controlled safety study, while potentially costing somewhat more than an OLE, would produce far more useful information and better protect both study participants and future patients.²¹ In practice, though, asking patients to accept re-randomization to placebo after they have already been on an active drug is a tough sell, both ethically and logistically.

Ethical Dimensions of Continued Access

For patients with serious or life-threatening conditions, the OLE is often their only path to continued access to a drug that may not yet be approved or available outside the trial. This creates an ethical tension: the extension is structured as a research study, but for many participants it functions as treatment. If the drug eventually fails to win approval, or if the sponsor decides to stop the study, patients who depended on it can be left without options.

Questions also arise about equity. Not everyone from the parent trial is offered the extension, and geographic, financial, or logistical barriers can prevent some participants from continuing. The population that enters an OLE tends to skew toward those with better access to the study site, more stable health insurance, and fewer competing life demands. This filters not just the science but also who benefits from continued access to experimental therapies. These issues have been raised as concerns in bioethics discussions about the obligations sponsors have to participants after the original trial concludes.

There is no consensus on how far those obligations extend, but the tension is real: OLE studies sit at the intersection of research and clinical care, and the rules governing each do not always align neatly. For patients, the practical reality is that an OLE may be the best available option, even though its design is imperfect and its future uncertain.