Claims of increasing long-term benefit from lecanemab and donanemab are being called into question by a reanalysis of published data from the pivotal trials that led to the antiamyloids’ approval.

Investigators found that cognitive decline was steeper during the drugs’ open-label extension (OLE) phases than during the preceding double-blind treatment periods and that the OLE decline rates were similar to those seen with placebo during the controlled trials. 

The investigators noted that the steeper decline occurred despite roughly 40%-45% of participants dropping out before the extension phases ended — attrition that should have enriched the remaining cohort with those experiencing the slowest rate of cognitive decline.

“At best, this is the same slope of decline as placebo, and I say ‘at best’ because we’re using the best possible slope we can have. All the people that didn’t do well were already out,” the study’s senior author, Alberto Espay, MD, MSc, professor and Research Endowed Chair of the James J. and Joan A. Gardner Family Center for Parkinson’s Disease and Movement Disorders at the University of Cincinnati in Cincinnati, told Medscape Medical News. 

The study was published online on August 10 in BMJ Neurology Open.

No Placebo Comparison

The original 18-month, double-blind, placebo-controlled phase 3 trials of lecanemab (Clarity AD) and donanemab (TRAILBLAZER-ALZ 2) showed modest but statistically significant differences on the Clinical Dementia Rating-Sum of Boxes (CDR-SB), corresponding to a roughly 25%-35% relative slowing of decline compared with placebo. However, the absolute between-group differences were small but statistically significant.

Both trials subsequently launched OLEs in which all participants received active treatment. The extension publications from Eisai and Eli Lilly reported progressively greater benefit over time, but those claims were based on comparisons with untreated historical cohorts from registries such as the Alzheimer’s Disease Neuroimaging Initiative, not with a contemporaneous placebo group.

In contrast, the reanalysis compared rates of decline within each trial across the double-blind and open-label phases. Because individual-level participant data were unavailable, the investigators used two approaches. They simulated participant-level trajectories calibrated to published trial data and direct comparisons of observed annualized decline rates. Both approaches produced consistent results.

Patients who complete 3 or 4 years of treatment are not representative of the full trial population, said Bruno Imbimbo, PhD, of Chiesi Farmaceutici in Parma, Italy, co-author of the study and who previously managed an OLE for a gamma secretase modulator in Alzheimer’s disease.

“The sub-cohort that completes the extension has a more benign natural history,” he told Medscape Medical News. “They look better at baseline.”

The Unblinding Effect

In the simulation-based analysis, among participants who received lecanemab from baseline, the annualized rate of CDR-SB decline increased from -0.78 points per year during the double-blind phase to -1.30 points per year during the OLE (P < .001). A similar pattern was seen with donanemab, with annualized decline increasing from -1.07 to -1.73 points per year (P < .001).

“In a disease-modifying paradigm, the slope should flatten over time. If there is no effect, it stays the same,” Espay said. “What we see here is the slope steepens.”

In both trials, participants who received a placebo during the double-blind phase and then switched to an active drug showed no meaningful change in decline rate during the OLE phase. This pattern suggests the benefit observed during the controlled phase was partly a placebo effect that faded once blinding ended, the investigators note.

In combined analyses, the OLE decline rate did not differ significantly from the double-blind placebo rate for either drug.

A secondary analysis showed that lecanemab placebo participants declined significantly more slowly than matched untreated historical controls (P < .001), suggesting a placebo effect that may have inflated the original treatment contrast. In the donanemab trial, the decline in the placebo arm did not differ significantly from untreated control groups.

Three independent analytical approaches — including a direct comparison that did not rely on simulated data — all produced consistent results, said the study’s lead statistician Alok Dwivedi, PhD, MSc, professor and director of the Center for Integrated Biostatistics and Epidemiology at the University of Missouri School of Medicine in Columbia, Missouri.

“Even under the most conservative assumptions, without simulating data at all, you still observe the same finding,” he said.

Tammy McGuire, a spokesperson for Eli Lilly, told Medscape Medical News the TRAILBLAZER-ALZ 2 extension period remained blinded and that steeper decline during later follow-up is expected because participants are further along in Alzheimer’s disease progression. However, the reanalysis describes both extensions as open-label; the TRAILBLAZER-ALZ 2 extension paper itself uses the term “long-term extension.” 

Espay said the distinction was largely terminological, as all participants received active treatment after the randomized phase, and the conclusions do not depend on whether the extension is labeled as blinded or open-label.

Reanalysis Methodology Questioned

Reached for comment on the reanalysis, Sarah Ackley, PhD, assistant professor of epidemiology at the Brown University School of Public Health in Providence, Rhode Island, who was not involved in the research, raised a methodological concern. The simulation, she noted, didn’t actually detect the acceleration in cognitive decline, since that pattern was already visible in the published data without any modeling. 

Ackley also pointed to errors in the simulation code, including a mismatch between the stated statistical methods and the analysis the authors implemented, and noted that the investigators relied on only one reconstruction of patient-level data — even though many different datasets could be compatible with the same published averages.

“I think this is an interesting analysis that raises an important issue,” Ackley said. “I just can’t get behind the specific analysis.”

However, she added that her concerns do not alter the underlying observation. The published aggregate data do show a steeper decline after 18 months in early-start groups, and that pattern warrants closer examination. Claims of increasing long-term benefit with these drugs would require a longer randomized trial rather than comparisons with historical cohorts, she added.

Dwivedi acknowledged Ackley’s critique as partly valid but said the goal was to quantify absolute slopes rather than chase statistical significance. As for her concern that the researchers relied on only one reconstruction, he said the analysis was actually run with random slopes and produced the same directional result — though the published code used a simpler model that more closely matched the reported data.

The simulation was not intended to independently discover the acceleration, Espay said, but rather to quantify a pattern already visible in the published data. The investigators are examining whether a formal clarification or correction to the simulation code is warranted, he said, but a direct comparison of published slopes — which does not depend on reconstructed patient-level data — showed the same directional result.

“The code concerns should be addressed, but without access to the participant-level trial data, neither claims of increasing long-term benefit nor concerns about accelerating decline can be evaluated as definitively as they should be,” Espay said.

Medscape Medical News contacted several investigators involved in the pivotal trials for comment on the reanalysis.They included: Christopher van Dyck, MD, of Yale School of Medicine in New Haven, Connecticut, and the first author on the Clarity AD trial and its OLE; Reisa Sperling, MD, of Harvard Medical School in Boston and a co-author on the extension study; Paul Aisen, MD, of the Keck School of Medicine of USC in Los Angeles and a co-author on the Clarity AD trial; and Liana Apostolova, MD, MSc, of Indiana University School of Medicine in Indianapolis and a co-author on TRAILBLAZER-ALZ 2. All declined to comment.

Stephen Salloway, MD, MS, of Brown University’s Warren Alpert Medical School and a co-author on TRAILBLAZER-ALZ 2, did not respond to requests for comment at press time. Eisai likewise did not respond.

The study received no specific funding. Espay reported receiving grant support from the National Institutes of Health (NIH) and the Michael J. Fox Foundation; consulting relationships with Mitsubishi Tanabe Pharma America, Amneal Pharmaceuticals, Acorda Therapeutics, AbbVie, Bial, Supernus Pharmaceuticals, NeuroDiagnostics, Inc., Intrance Medical Systems, Merz, Praxis Precision Medicines, Citrus Health, and Herantis Pharma; and publishing royalties from Lippincott Williams & Wilkins, Cambridge University Press, and Springer. He is co-inventor of a patent that formed the basis for Regain Therapeutics, which he co-founded; he also reported having no financial relationship with the company and owns no stock in any pharmaceutical company with which he has a scientific advisory relationship. Imbimbo is an employee of Chiesi Farmaceutici and is listed among the inventors of Chiesi patents for anti-Alzheimer drugs; Chiesi has not been involved in anti-Alzheimer drug development since 2014. Ackley reported receiving research support from the NIH/National Institute of Aging and Sanofi (funds to the institution) and serves as an uncompensated board member of the Center for the Future of Knowledge. Dwivedi reported having no relevant financial relationships.