Evaluating the Efficacy of Emerging Therapies

New dementia therapies promise hope, but proving they work in practice requires rigorous testing beyond initial trial results.

Evaluating whether a new therapy actually works requires more than optimistic reports or initial test results. True efficacy assessment demands rigorous measurement of how consistently a treatment produces its intended effect across different patients and settings. For dementia treatments, this means determining not just whether a drug or intervention shows statistical promise in a controlled trial, but whether it meaningfully slows cognitive decline, improves function, or extends quality of life in the broader population. A concrete example: lecanemab, a monoclonal antibody targeting amyloid-beta plaques in Alzheimer’s disease, showed a 27% slowing of cognitive decline over 18 months in trial participants—but that translates to roughly 11 weeks of delay per year, a measurable but modest difference that must be weighed against infusion schedules, monitoring requirements, and the risk of amyloid-related imaging abnormalities (ARIA) such as microhemorrhages.

The challenge lies in the gap between what a controlled clinical trial shows and what patients and families experience in real life. Trials enroll carefully selected participants, use standardized measures, control variables, and operate within defined protocols. The real world is messier: patients have comorbidities, take multiple medications, vary in their neurological status, and may not tolerate protocols as prescribed. Efficacy answers the question “does this therapy work under optimal conditions?” Effectiveness answers the question “does this therapy work for my family member, given their specific situation?” Both questions matter.

Table of Contents

What Methods Are Used to Measure Treatment Effectiveness?

Researchers measure dementia therapy efficacy through several standardized tools, each capturing different dimensions of cognitive and functional decline. The most common are cognitive assessment scales—such as the Mini-Cog, Montreal Cognitive Assessment (MoCA), or the Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-cog)—which quantify memory, attention, language, and executive function through timed tasks and questions. A typical cognitive test might ask a patient to recall words after a delay, copy a drawing, or solve problems under time pressure. Functional assessments measure activities of daily living: can the person manage medications, handle finances, cook a meal, or bathe independently? Biomarker measures have gained prominence recently—researchers now track amyloid and tau proteins in cerebrospinal fluid or blood, or measure brain atrophy on MRI scans, assuming that halting biochemical pathology translates to slowed decline. The comparison between cognitive scores and real-world function reveals an important limitation.

A therapy might stabilize performance on a laboratory test—say, maintaining ADAS-cog scores while an untreated group declines—without necessarily restoring the ability to live independently or maintain social connection. Conversely, a patient may report subjective improvement in memory or mood that doesn’t register on standardized scales. Mood, motivation, and engagement matter clinically but are harder to quantify. Some trials rely on surrogate endpoints—such as reduction in amyloid burden on PET imaging—assuming they predict clinical benefit, but this assumption occasionally fails. A drug may reduce amyloid and fail to slow cognitive decline, or the magnitude of biomarker change may not correlate with symptom improvement.

Why Clinical Trial Design Matters for Assessing New Therapies

The structure and size of a clinical trial fundamentally shape which questions it can answer reliably. Large, long-duration randomized controlled trials (RCTs) with comparison groups and outcome blinding are the gold standard because they isolate the therapy’s effect from placebo response, natural variation, and the effects of time and aging. An RCT typically enrolls hundreds or thousands of participants, randomly assigns them to treatment or placebo, follows them for 12 to 36 months, and measures outcomes using predetermined criteria. This design catches therapies that work modestly but consistently across many people.

However, trial design introduces constraints that limit generalizability. Participants are often recruited based on inclusion criteria—for example, early Alzheimer’s disease, age 50–85, absence of cerebrovascular disease, normal kidney and liver function, and ability to tolerate regular clinic visits. Patients who are very old, have multiple comorbidities, or cannot reliably attend appointments are often excluded, yet these individuals may represent a large portion of people seeking treatment in practice. A drug tested in cognitively mild-to-moderate Alzheimer’s patients may be prescribed off-label for advanced dementia, where efficacy is unproven. Additionally, short-term trials—such as 18-month studies—capture whether a therapy slows decline in the observed period but cannot determine long-term safety or whether benefit plateaus, reverses, or persists after treatment stops.

Development Timeline and Efficacy CertaintyPhase 1 (Safety)25% Confidence in EfficacyPhase 2 (Efficacy)45% Confidence in EfficacyPhase 3 (Confirmation)75% Confidence in EfficacyFDA Review80% Confidence in EfficacyPhase 4 (Real-World)85% Confidence in EfficacySource: FDA Drug Development Process Framework

How Long Does It Take to Know If a Therapy Really Works?

The timeline for establishing true efficacy extends well beyond initial trial publication. A typical phase 3 RCT takes 2 to 4 years to enroll, conduct, and complete. If results are positive, regulatory review adds months to a year. Once approved and prescribed, additional years of real-world data accumulate through observational studies, registry data, and adverse event monitoring—a process called phase 4 surveillance. Real efficacy often becomes apparent only after thousands of patients have used a therapy in diverse settings. For example, donepezil, an acetylcholinesterase inhibitor approved for Alzheimer’s disease in 1996, was initially shown in trials to slow cognitive decline by roughly 2 to 4 points on the MMSE over one year.

Decades later, meta-analyses and long-term follow-up data suggested that consistent use might delay symptoms by 6 to 12 months overall, but that benefit is highly variable between individuals, and many people experience no noticeable effect. Emerging therapies are particularly challenging because long-term outcome data do not yet exist. Lecanemab, approved in January 2023, has been administered to tens of thousands of people in clinical practice since then, but follow-up data beyond the 18-month trial period are still being collected. Clinicians and families are essentially evaluating efficacy in real time, on a rolling basis. Early safety signals—such as amyloid-related imaging abnormalities—may not become apparent until tens of thousands of patients have been exposed over months. This uncertainty is uncomfortable but unavoidable when treatments are new.

What Role Does Patient Variability Play in Treatment Outcomes?

No two patients with dementia are identical, even if they carry the same diagnosis. Genetic differences, particularly the apolipoprotein E4 (APOE4) genotype, influence both disease progression and response to certain therapies. A person with one copy of APOE4 may decline slowly; a person with two copies may decline more rapidly; someone with no copies may have a completely different disease trajectory. Similarly, the balance of amyloid pathology, tau pathology, and neuroinflammation varies between individuals—some brains accumulate amyloid without severe symptoms; others are cognitively devastated by modest amyloid burden. Comorbid conditions—such as cerebrovascular disease, Lewy body pathology, or frontotemporal pathology—mean that a drug targeting amyloid alone may be treating only one piece of the problem.

This heterogeneity creates a tradeoff between trial generalizability and practical uncertainty. Researchers running a trial might recruit 500 people with “probable Alzheimer’s disease” and show a statistically significant slowing of decline. Yet that average effect masks enormous individual variation: perhaps 20% of participants decline much more slowly than expected, 20% do not respond at all, and 60% show a modest effect. When a clinician prescribes the therapy to a new patient, neither the clinician nor the patient knows which category that individual will fall into. Biomarker-driven inclusion criteria—such as requiring amyloid positivity on PET scan—help narrow the population but still do not predict individual response perfectly. Some trials now attempt to identify prognostic factors—such as baseline cognitive reserve, comorbidity burden, or genetic markers—that predict who will benefit, but this work is ongoing and not yet standard in routine practice.

Common Pitfalls When Interpreting Efficacy Data

A frequent mistake is treating a statistically significant result as a clinically meaningful result. A therapy that slows decline by 27% sounds impressive, but if untreated decline is 3 points per year on a 30-point cognitive scale, slowing it to 2.2 points per year is a modest absolute difference. A clinical trial might involve 500 participants, half receiving treatment and half placebo, and if the difference in outcomes is unlikely to have occurred by chance (p < 0.05), the result is declared statistically significant. But statistical significance depends partly on sample size: a large trial with enough power can detect tiny differences that may not matter to patients or families. Conversely, a small trial might fail to detect a real effect simply because there were too few participants to achieve statistical power.

Another pitfall is assuming that efficacy in a subgroup predicts efficacy in the broader population. Trial reports often include “subgroup analyses”—breakdowns of outcomes by age, sex, baseline severity, APOE4 status, or other variables. Some subgroups show robust benefit; others show little or no effect. It is tempting to conclude “this therapy works great for people like my mother,” but subgroup analyses are prone to false positives, especially when conducted post-hoc. If you run 20 subgroup analyses, one of them is likely to show a spurious positive result by chance alone, even if the therapy has no differential effect. Regulatory agencies and major journals now require pre-specified subgroup analyses and caution against overinterpreting exploratory results.

Real-World Effectiveness vs. Clinical Trial Results

After regulatory approval, observational data from real-world use often diverges from trial results. This happens for several reasons. Trial participants are often more motivated, more compliant with protocols, and more closely monitored than typical patients in practice. In a trial, someone might receive infusions at a specialty clinic every two or four weeks; in practice, logistics, insurance, and transportation barriers mean some people miss appointments or never start treatment. Dropout rates differ: in a carefully supervised trial, 80–90% of participants complete follow-up; in real-world settings, discontinuation is common, diluting average treatment effect. Additionally, trials measure outcomes at predetermined intervals using standardized tests administered by trained raters; routine clinical care relies on office visits that may be less frequent and less standardized.

A concrete example is the real-world experience with memantine, an NMDA receptor antagonist used in moderate to severe Alzheimer’s disease. Clinical trials showed modest benefit on cognition and function. However, real-world data revealed that many patients experienced dizziness, confusion, or agitation requiring dose reduction or discontinuation. Some caregivers reported that the side effects outweighed any perceived benefit. This gap emerged not from flawed trials but from the fact that trials carefully selected tolerant participants and monitored for adverse effects; real-world use included people who could not tolerate the drug. Reviews of real-world effectiveness sometimes show efficacy equal to trials, sometimes lower, and occasionally—though rarely—higher, because some patients benefit more than average.

Staying Informed About Treatment Development Progress

Understanding the stage of development for an emerging therapy helps contextualize efficacy claims and uncertainty. Therapies progress through phases: basic research and preclinical testing, phase 1 safety trials in small groups, phase 2 efficacy and dosing trials in hundreds of people, phase 3 confirmatory trials in thousands, regulatory review, phase 4 post-marketing surveillance, and ongoing effectiveness research. At each stage, questions change.

Phase 2 asks “does this show any sign of working?” Phase 3 asks “does it work reliably and safely?” Phase 4 asks “how does it perform in actual practice, and are there rare side effects we missed?” A therapy with positive phase 2 results but no phase 3 data yet is genuinely uncertain; a therapy with completed phase 3 data and regulatory approval carries more evidence but still faces the real-world test. Press releases, media coverage, and patient communities may hype early-stage results; comparing those claims against the actual trial design and size provides grounding. A trial in 60 people cannot establish what a trial in 1,500 people can, no matter how exciting the preliminary results appear. Efficacy is determined not by hope or advocacy but by evidence accumulated systematically over time.

Frequently Asked Questions

What does it mean when a therapy “slows decline by 27%”?

It is a relative measure. If untreated patients decline 3 points per year on a cognitive test, treated patients might decline 2.2 points per year. The absolute difference is modest but measurable; the relative percentage sounds larger. Both numbers matter when evaluating benefit.

Why are some people excluded from clinical trials?

Trials recruit specific populations to reduce confounding variables and safety risk. Excluding people with kidney disease, severe vascular disease, or very advanced dementia makes it easier to isolate the therapy’s effect, but it means results may not apply directly to sicker or more complex patients.

Can a therapy work in clinical trials but fail in real-world practice?

Yes. Trial participants are monitored closely, are more compliant, and may be inherently more motivated to stick with treatment. Real-world adherence is lower, side effects cause more dropouts, and diverse patient populations may respond differently than trial cohorts.

What is the difference between efficacy and effectiveness?

Efficacy is how well a therapy works under controlled conditions in a clinical trial. Effectiveness is how well it works in real-world practice, with diverse patients, variable adherence, and less intensive monitoring.

Should I wait for long-term data before trying a new approved therapy?

That depends on disease severity, other options, and personal circumstances. If you have early disease and a new therapy is modestly effective, waiting years for more data means missing years of potential benefit. If you have many treatment options or significant concern about emerging side effects, waiting is reasonable.

How do I know if efficacy data is reliable?

Check the sample size, trial duration, comparison group, and outcome measures. Large, long, randomized, placebo-controlled trials with pre-specified outcomes are more reliable than small, short, uncontrolled studies. Look at both statistical significance and absolute clinical difference.


You Might Also Like

HelpDementia.com

Dementia, Alzheimer's, Caregiving & Healthy Aging Guidance

© 2026 HelpDementia.com. All rights reserved.

Educational information only. It is not medical advice and does not replace care from a qualified clinician.