Lecanemab vs donanemab efficacy meta-analysis

Four Primary Endpoints: Effect Sizes, Confidence Intervals, and Statistical Models

Across all four primary outcome measures, lecanemab and donanemab produced statistically significant improvements versus placebo in patients with early Alzheimer's disease, with p-values reaching below 0.0001 in every case. The magnitude and consistency of those effects, however, varied substantially by endpoint — a pattern that carries direct implications for how anti-amyloid therapies are evaluated and approved.

The Alzheimer's Disease Composite Score (ADCOMS) showed the smallest absolute effect size — HR = −0.05 (95% CI: −0.07 to −0.03, p < 0.00001) — but this reflects the composite scale's compressed range rather than a lack of clinical signal. The CDR-SB, which captures memory, cognitive function, and daily living abilities, returned an effect size of HR = −0.49 (95% CI: −0.67 to −0.30, p < 0.00001). Both endpoints were analysed using a fixed-effects model, consistent with the low heterogeneity observed across contributing trials.

The ADAS-Cog 14, a 14-item cognitive subscale widely used in regulatory submissions to the FDA and the EMA, produced an effect size of SMD = −1.06 (95% CI: −1.54 to −0.57, p < 0.0001). This represents a clinically meaningful reduction in cognitive impairment, though moderate heterogeneity (I² = 47%, p = 0.15) was present, suggesting some variability in how individual trials captured this endpoint. A fixed-effects model was nonetheless applied.

The most dramatic effect size emerged for amyloid burden on PET: HR = −72.99 SUVr (95% CI: −88.58 to −57.41, p < 0.00001). This figure reflects the substantial cerebral amyloid deposition clearance achieved by both agents. Critically, this endpoint was analysed using a random-effects model — the only endpoint in the meta-analysis to require one — due to extreme heterogeneity (I² = 95%, p < 0.00001).

What Heterogeneity Statistics Reveal About Trial Consistency in Anti-Amyloid Therapy

The heterogeneity profile across the four endpoints tells a coherent story about where anti-amyloid therapies produce predictable versus variable effects. ADCOMS and CDR-SB both returned I² = 0% — the lowest possible heterogeneity — with heterogeneity p-values of 0.76 and 0.55 respectively. This near-perfect consistency across trials justifies the fixed-effects model applied to both and suggests that the functional and cognitive benefits measured by these scales are relatively uniform regardless of the specific trial protocol or patient population.

ADAS-Cog 14 introduced moderate heterogeneity at I² = 47% (p = 0.15). This level of variability is common in cognitive endpoint meta-analyses and likely reflects differences in patient selection, baseline cognitive severity, and how individual trials administered the 14-item scale. Importantly, the heterogeneity p-value of 0.15 does not reach the conventional threshold for statistical significance, which partly justifies the continued use of a fixed-effects model despite the elevated I².

The extreme heterogeneity for amyloid PET (I² = 95%, p < 0.00001) is the most analytically significant finding in the heterogeneity data. It indicates that the degree of amyloid clearance varies enormously across trials — almost certainly driven by differences in baseline amyloid burden, dosing regimens, treatment duration, and the specific PET tracers and SUVr thresholds used. This variability is precisely why a random-effects model was required: the 95% I² means that 95% of the observed variance in amyloid PET results reflects genuine between-study differences rather than sampling error.

"The amyloid PET endpoint returned I² = 95% — meaning 95% of observed variance reflects genuine between-study differences — yet the pooled effect of −72.99 SUVr was still statistically significant at p < 0.00001, confirming robust amyloid clearance despite substantial cross-trial variability."

The Biomarker–Cognition Dissociation: Three Mechanistic Explanations

The dissociation between the amyloid PET effect size (HR = −72.99 SUVr) and the ADAS-Cog 14 cognitive improvement (SMD = −1.06) is one of the most clinically important findings in this meta-analysis, and it is not unique to lecanemab and donanemab — the same phenomenon was reported in the TRAILBLAZER-ALZ 4 trial. The meta-analysis identifies three distinct mechanistic pathways through which this dissociation arises.

1. Multi-step translation from amyloid clearance to clinical benefit

Removing amyloid-beta (Aβ) plaques from the brain is not the final step in restoring cognitive function — it is the first. According to the meta-analysis, Aβ clearance requires multiple downstream processes, including synaptic remodeling and neuroinflammatory relief, before clinical improvements become measurable. Each of these steps introduces delay and biological variability. The timeline over which synaptic connections are restored following plaque removal may extend well beyond the duration of the clinical trials in which cognitive outcomes were assessed, meaning that the full cognitive benefit of amyloid clearance may not yet have been captured.

2. Ongoing tau pathology offsetting amyloid clearance benefits

Alzheimer's disease is defined by two hallmark pathologies: amyloid-beta plaques and tau neurofibrillary tangles. Anti-amyloid monoclonal antibodies such as lecanemab and donanemab target only the first of these. The meta-analysis notes that the continuous progression of tau pathology may offset the potential benefits of Aβ clearance, even when amyloid is successfully removed. This is consistent with the understanding — supported by research published through the NIH and leading neuroscience journals — that tau burden is more directly correlated with neuronal loss and cognitive decline than amyloid burden alone. Patients who enter trials with advanced tau pathology may experience limited cognitive benefit from amyloid clearance precisely because the tau-driven neurodegeneration continues unabated.

3. Insensitivity of current assessment scales in early AD

The third mechanistic explanation is methodological rather than biological: current cognitive assessment tools, including the ADAS-Cog 14, may not be sensitive enough to capture the subtle cognitive changes that characterise the earliest stages of Alzheimer's disease. In early AD — the population targeted by lecanemab and donanemab — patients by definition have relatively preserved cognitive function at baseline. The floor effects and ceiling effects inherent in scales designed for moderate-to-severe AD may compress the measurable cognitive signal even when genuine improvement is occurring. This would systematically underestimate cognitive benefit relative to the biomarker response, producing the dissociation observed in the meta-analysis.

Implications for Drug Development, Regulatory Strategy, and the Future of AD Clinical Assessment

The findings of this meta-analysis carry several concrete implications for how anti-amyloid therapies are developed, assessed, and approved. Lecanemab and donanemab have demonstrated that anti-Aβ monoclonal antibodies can achieve statistically significant improvements in cognitive function and daily living activities in early-stage Alzheimer's disease patients, and can effectively reduce cerebral amyloid deposition. That dual demonstration — biomarker and functional — is what enabled regulatory progress for both agents.

However, the observed dissociation between substantial biomarker response and more moderate clinical cognitive improvement points to the complex pathophysiology of Alzheimer's disease and the potential limitations of current assessment tools. For R&D teams and patent strategists tracking the anti-amyloid space, this has several downstream consequences.

First, the high heterogeneity in amyloid PET outcomes (I² = 95%) signals that amyloid clearance magnitude is highly sensitive to trial design variables — baseline amyloid burden thresholds, PET tracer selection, SUVr cut-offs, and dosing protocols. Patent claims and clinical development programmes that specify these parameters precisely will be better positioned to produce reproducible and regulatorily acceptable biomarker data.

Second, the moderate heterogeneity in ADAS-Cog 14 (I² = 47%) compared to zero heterogeneity in ADCOMS and CDR-SB suggests that composite and functional scales may produce more consistent evidence across trials than item-level cognitive tests. According to the WHO's global dementia action plan framework, standardised outcome measurement is a core priority for dementia research — a priority that the heterogeneity data in this meta-analysis underscores empirically.

Third, the mechanistic explanation that current scales are insufficiently sensitive for early AD creates a clear innovation opportunity: next-generation cognitive assessment tools, digital biomarkers, and fluid biomarker combinations (such as plasma p-tau217) that can detect subtle changes earlier in the disease course. Monitoring the patent landscape in this area — alongside the clinical trial pipeline — is increasingly essential for any organisation competing in the Alzheimer's therapy space.

The meta-analysis ultimately confirms that lecanemab and donanemab represent a meaningful advance in early Alzheimer's disease treatment — all four primary endpoints reached statistical significance, and the consistency of the functional outcome data (I² = 0% for ADCOMS and CDR-SB) is particularly compelling. The biomarker–cognition dissociation is not evidence of therapeutic failure; it is evidence of the multi-step, multi-pathology complexity of a disease that remains one of the most active areas of pharmaceutical R&D globally, as documented by WIPO's annual technology trend reports on neurodegenerative disease innovation.