Statistical Method Inflates Alzheimer’s Drug Trial Success 29x

Summary: A new research letter warns that a recently promoted statistical method used to support a class of Alzheimer’s drugs can dramatically overstate clinical effects. The study evaluated “quantile aggregation,” a technique that groups patients by measured outcomes, averages those results, and then examines trends across those groups. Computational simulations show this approach can hide patient-level variability and inflate the apparent connection between amyloid reduction and cognitive improvement by as much as 29-fold.

Originally applied in a high-profile reanalysis of Eli Lilly’s donanemab trial data, quantile aggregation was put under scrutiny by investigators at the Brown University School of Public Health. Their analysis, published in JAMA Neurology, explores how the method operates when applied to cognitive measures and amyloid burden — the protein deposits commonly associated with Alzheimer’s disease — and highlights several significant limitations.

Key Facts

  • The method: Quantile aggregation reorganizes trial participants by post-treatment biomarker levels rather than by their original randomized groups, then averages outcomes within those quantiles and looks for trends across them.
  • Large inflation of effect: In simulations reflecting recent trial conditions, this technique exaggerated the strength of an amyloid–cognition relationship by up to 29 times its true size.
  • Masked variability: Averaging outcomes across broad patient groupings smooths over individual differences in cognitive change, producing a misleadingly consistent signal.
  • Breaks randomization: By pooling treated and placebo participants into the same aggregated blocks, quantile aggregation removes a trial’s randomized structure and cannot reliably establish causation between amyloid clearance and cognitive benefit.
  • Resurrecting failed trials: When the researchers applied the method to data from the 2014–2023 solanezumab trial — which failed to slow cognitive decline — the analysis nonetheless produced a spurious strong link between lower amyloid and improved cognition.
  • Independent scrutiny: The study’s senior author, Sarah Ackley, emphasizes that independently conducted academic audits of statistical approaches are crucial to evaluating the evidence behind major therapeutic claims.

Source: Brown University

The JAMA Neurology research letter focuses on the technical properties and real-world implications of quantile aggregation. The technique was used in a reanalysis of randomized trial data for donanemab, and proponents have argued it reveals a clear link between amyloid removal and cognitive outcomes. The Brown-led team set out to test whether that claim holds up under rigorous simulation and historical data checks.

This shows a brain made of pills.
Advanced biostatistical modeling demonstrates that grouping and averaging heterogeneous trial cohorts through quantile aggregation completely masks individual patient variability, artificially amplifying weak therapeutic correlations up to 29-fold. Credit: Neuroscience News

Sarah Ackley, an assistant professor of epidemiology at Brown and head of the Computational Epidemiology Lab, says many researchers still view amyloid clearance as a plausible mechanism to slow Alzheimer’s-related decline. But the team’s results show that quantile aggregation can transform weak, noisy relationships into apparently strong and clinically meaningful effects purely through its aggregation steps.

In carefully controlled simulations modeled on recent Alzheimer’s trials, the authors found that even when the true link between amyloid reduction and cognitive benefit was small or absent, quantile aggregation frequently produced large, statistically convincing signals. The main drivers are the loss of individual-level variability and the abandonment of randomized comparisons; both make the method prone to producing false-positive associations.

To demonstrate how misleading the approach can be, the researchers reanalyzed the solanezumab trial — a study that showed no clinical benefit and no reliable amyloid clearance. Feeding those trial data into the quantile aggregation framework produced an apparently robust relationship between reduced amyloid and improved cognition, despite no substantive evidence in the original randomized trial to support that conclusion.

Ackley stresses that these findings do not resolve the broader scientific question of whether amyloid removal can benefit patients. Rather, the work highlights the need for more rigorous statistical approaches and transparent data sharing. As anti-amyloid treatments become more widely used and financially supported by public programs, robust and reproducible evaluation methods are essential to guide clinical and policy decisions.

“Our analysis was intentionally straightforward,” Ackley said. “It demonstrates the value of independent academic review: when researchers work outside commercial incentives, they can rigorously examine methods that influence how high-stakes drugs are interpreted.”

Key Questions Answered:

Q: If a drug clears amyloid plaques, does that automatically mean it treats Alzheimer’s?

A: Not necessarily. Many scientists hypothesize that lowering amyloid may slow memory loss, and it is a central target for several new drugs. However, cognitive measures are noisy and individual responses vary widely. A weak or inconsistent biological effect can be turned into a misleadingly strong clinical claim by inappropriate analytic methods like quantile aggregation.

Q: How does quantile aggregation create a false impression of success?

A: Think of averaging entire classroom test scores instead of looking at each student. Exceptional or poor individual outcomes are smoothed away, producing a deceptively clean average. Quantile aggregation groups patients by a biomarker, averages their results, and fits trends across those averages, which can erase the real variability that matters for causal inference.

Q: Why did a failed drug look effective when analyzed this way?

A: To test the method’s robustness, the team applied quantile aggregation to solanezumab trial data — a study that did not show clinical or biomarker success. The aggregation procedure nonetheless produced a strong apparent association between lower amyloid and better cognition, demonstrating how the method can distort null results into apparently positive findings.

Editorial Notes:

  • This article was edited by a Neuroscience News editor.
  • The underlying journal paper was reviewed in full.
  • Additional context was added by staff reporting.

About this Alzheimer’s disease and neuropharmacology research news

Author: Juan Siliezar
Source: Brown University
Contact: Juan Siliezar – Brown University
Image: Image credit: Neuroscience News

Original Research: Open access. “Methodological Considerations for Quantile Aggregation in Alzheimer Disease Trials” by Michael D. Flanders, Michelle Caunca, Renaud La Joie, Lon S. Schneider, and Sarah F. Ackley. JAMA Neurology. DOI: 10.1001/jamaneurol.2026.1240


Abstract

Methodological Considerations for Quantile Aggregation in Alzheimer Disease Trials

Is amyloid reduction an appropriate surrogate outcome for clinical benefit in anti-amyloid trials? This question is central to interpreting current results and designing future studies. Cognitive measures are inherently noisy, which has motivated alternative analytic strategies. Quantile-aggregation methods regroup participants by post-treatment amyloid burden rather than by randomized assignment and have produced clear amyloid–cognition relationships in data from TRAILBLAZER-ALZ 2. The statistical properties and limitations of this approach for interpreting trial data have not been fully characterized; this research letter addresses those methodological considerations.