Sample size matters in Alzheimer’s research because underpowered studies—those with too few participants—can fail to detect whether a drug actually works, leading families to believe a treatment is ineffective when it merely wasn’t tested on enough people to prove it. This happens more often than most people realize: a trial with 50 participants per arm might miss a real 25% reduction in cognitive decline, while a properly powered study of 800 participants per arm would catch it. Right now, as of 2026, there are 192 active clinical trials testing 158 novel Alzheimer’s agents across more than 4,500 sites globally with over 50,000 participants enrolled, and the difference between which of these trials produces reliable results often comes down to a single question: did the researchers enroll enough people? When a news headline announces that a new Alzheimer’s drug “showed promise in a trial,” the first question to ask is how many people were in that trial.
A study with 100 participants tells you something very different than one with 2,000. The smaller study might have found something interesting, but it cannot reliably predict whether that finding will hold up when millions of people take the drug. This distinction matters to anyone trying to understand what Alzheimer’s research actually says—not just what it claims.
Table of Contents
- What Does Sample Size Mean in Alzheimer’s Clinical Trials?
- The Statistics Behind Trial Design and Sample Size Limitations
- Real-World Examples—How Large Are Actual Alzheimer’s Trials?
- Different Trial Types, Different Sample Size Needs
- The Attrition Problem—Why Sample Size Must Account for Dropouts
- How Technology Is Reducing Sample Size Requirements
- Why Caregivers Should Understand Sample Size When Evaluating Alzheimer’s News
What Does Sample Size Mean in Alzheimer’s Clinical Trials?
Sample size is the number of participants enrolled in a research study. In Alzheimer’s trials, researchers must decide upfront how many people to recruit based on statistical calculations that determine whether the study will have enough power to detect a real treatment effect. Power—usually set at 80 to 90 percent in modern trials—is the probability that the study will find a difference between treatment and placebo if that difference actually exists. A trial with 90% power has a 90% chance of detecting a true effect and a 10% chance of missing it by random bad luck.
The challenge is that Alzheimer’s disease progresses slowly and unpredictably. Cognitive decline measured by standard tests like the ADAS-Cog (Alzheimer’s Disease Assessment Scale-Cognition) varies widely from person to person. one person might lose 3 points per year on this test, while another loses 8. Because of this variation, researchers need large sample sizes to reliably separate the signal—does the drug work?—from the noise of natural variation. If you test a drug on 30 people, you cannot tell whether a small observed difference is real improvement or just the luck of which people were randomly assigned to each group.
The Statistics Behind Trial Design and Sample Size Limitations
To properly power an Alzheimer’s trial, researchers use formulas that account for four things: the minimum effect size they want to detect (e.g., slowing cognitive decline by 25%), the expected variation within the population (the standard deviation, typically 6 to 6.5 points on the ADAS-Cog), the desired statistical power (usually 90% in recent trials), and the acceptable error rate (typically 5%, written as p < 0.05). Plugging these into a power calculator might yield a requirement of 800 participants per arm—1,600 total including placebo control. This creates a real tension: sample size directly drives trial cost, timeline, and feasibility. A 2026 Alzheimer's trial costs between $50 million and $200 million or more.
That expense means researchers often face pressure to cut corners on sample size, enrolling fewer participants than statistically ideal. A smaller trial is cheaper and faster, but it runs a critical risk: if the true effect size is smaller than the researchers predicted, the underpowered trial will miss it. This happened implicitly in some earlier Alzheimer’s drug trials where promising preliminary results in small studies failed to replicate in larger confirmatory trials. Families reading headlines about “breakthrough” results in a 100-person study should know that this does not yet tell them whether the drug works for the general population.
Real-World Examples—How Large Are Actual Alzheimer’s Trials?
Lecanemab (marketed as Leqembi), one of the first amyloid-targeting monoclonal antibodies approved for early Alzheimer’s disease, was tested in the Clarity AD trial with 1,795 total participants—898 receiving active drug and 897 receiving placebo. This was a Phase 3 trial (the final stage before potential approval), and this sample size was large enough to detect a 35% slowing of cognitive decline over 18 months. The trial showed a 27% slowing in the treatment group, a real effect that crossed the statistical threshold of significance.
Donanemab, another similar drug, was evaluated in TRAILBLAZER-ALZ 2 with approximately 1,800 participants randomized to treatment or placebo, with stratification by tau pathology status (whether participants had elevated tau in the brain, a marker of neurodegeneration). These large sample sizes—in the range of 1,800 to 2,000 per arm for Phase 3 trials—have become standard for disease-modifying drugs targeting biomarkers. Smaller trials of 200 to 400 participants per arm are more common for early-stage Phase 2 work, where researchers are still figuring out the right dose and looking for preliminary efficacy signals, not confirming that a drug works.
Different Trial Types, Different Sample Size Needs
Prevention trials—studies testing drugs in cognitively unimpaired people to prevent Alzheimer’s from ever developing—require drastically larger sample sizes, typically 3,000 to 5,000 or more participants with follow-up lasting 4 to 7 years. This is because people without cognitive symptoms have much slower decline to measure. You cannot detect a treatment effect over several years if most participants never decline at all, so you need a much larger group. Early-stage trials (Phase 1 and 2) might enroll only 50 to 200 participants because they are designed to answer safety and dose questions, not efficacy.
These smaller trials are appropriate for their purpose, but they provide no proof that a drug works. Biomarker-based trials—where the primary outcome is amyloid or tau reduction in the brain measured by PET scan or cerebrospinal fluid, not cognitive scores—can achieve adequate power with smaller sample sizes because biomarkers change more predictably than cognitive scores. A study might need only 150 to 200 participants per arm to detect a 25% reduction in amyloid pathology over 4 years with 80% power. This is a major reason why 27% of current Alzheimer’s trials now use biomarker endpoints: they are more sensitive, require fewer participants, and can give answers faster. However, a biomarker improvement does not automatically mean cognitive benefit—a limitation researchers and regulators are still working through.
The Attrition Problem—Why Sample Size Must Account for Dropouts
Researchers always plan for attrition. People move, develop other medical problems, cannot tolerate side effects, or simply decide to withdraw from long studies. Most Alzheimer’s trials assume 30 to 40% of participants will drop out before the study ends. This means if a calculation shows you need 1,600 enrolled participants to power a trial, you actually need to recruit approximately 2,000 to 2,500 because roughly one in three will not make it to the end. If researchers underestimate dropout, they end up with fewer than the planned sample size actually completing the trial, and their power evaporates.
This is not a theoretical problem. Prevention trials lasting 6 to 7 years routinely lose 40% or more of participants. Trials in people with dementia face even higher attrition because participants are cognitively impaired and may lack the capacity to continue consenting. A study that enrolls 3,000 people for a 5-year prevention trial needs to anticipate that perhaps 1,800 will finish. Conversely, a trial planning for 30% dropout but experiencing only 5% dropout ends up with more power than calculated—this is not ideal either, because the reported p-value (the statistical significance) is based on the planned sample size, and exceeding it introduces statistical issues the authors must address.
How Technology Is Reducing Sample Size Requirements
Recent advances are beginning to reduce how many participants trials need. AI-driven patient selection—using machine learning to identify people most likely to benefit from treatment or progress fastest—can reduce required sample sizes by as much as 22% while maintaining 90% power. A trial that would have needed 1,000 participants per arm might need only 780 if AI algorithms identify the subgroup most likely to respond. This approach is becoming more common in 2025 and 2026 trials, especially those targeting specific biomarker profiles.
Multi-visit analysis is another shift. Traditionally, trials analyzed only the final visit (say, the 18-month endpoint). Newer statistical methods use all post-baseline visits—every 3-month or 6-month assessment—to estimate treatment effect while controlling Type I error (false positives). This yields greater statistical power from the same sample size because there is more information per person. The FDA now endorses amyloid and tau biomarkers as primary outcomes in pathology-positive populations, a change that reflects recognition that smaller, more precisely targeted trials can replace massive unselected populations.
Why Caregivers Should Understand Sample Size When Evaluating Alzheimer’s News
When a news story reports a breakthrough in Alzheimer’s treatment, your first critical filter should be the sample size and type of trial. A press release describing results from a 50-person Phase 1 trial conducted at one medical center is not equivalent to FDA approval of a drug after a 1,800-person Phase 3 trial across multiple countries. The Phase 1 result may be real and interesting, but it is preliminary.
If media coverage does not mention sample size, statistical power, or whether the trial was randomized and controlled, you are getting an incomplete picture. As of 2026, there are 8 Phase 3 trials expected to reach completion, and 29 Phase 2 trials finishing—these completions should generate clearer answers than Phase 1 or Phase 2 data, but only if you know which stage each announcement refers to. Asking “how many people were in this study?” before deciding whether to discuss a drug with a doctor is a reasonable and scientifically sound question.
- —
