BGPT: Paper Review: Lecanemab in clinical practice: real-world outcomes in early Alzheimer’s disease

Explore by Goal

Quick Explanation Copied

Critical read (real-world lecanemab at TLVMC)

In a single-center real-world cohort, this paper reports feasibility and safety metrics (e.g., ARIA incidence) for lecanemab delivered in routine practice, alongside MMSE change analyses with an age-stratified signal at 6 months. However, the study is observational (no control arm), has limited follow-up at 12 months, and uses MMSE as the primary cognitive endpoint, which collectively limit causal inference about clinical benefit.

Key safety signals reported:

ARIA occurred in 18.6% of treated patients, mostly asymptomatic, with 1 symptomatic hospitalization leading to discontinuation.
Infusion-related reactions occurred in 22.1%, described as mild and transient.

Causal caution:

Because there is no untreated control group and significant attrition by follow-up time, observed MMSE changes cannot be cleanly attributed to lecanemab rather than normal progression, selection effects, or regression-to-the-mean.

Long Explanation

Paper Review (real-world outcomes): “Lecanemab in clinical practice: real-world outcomes in early Alzheimer’s disease”

doi: 10.1186/s13195-025-01763-1 (Alzheimer’s Research & Therapy; received Mar 2025; accepted May 2025).

Study aim: Evaluate demographic characteristics, treatment response, and safety of early-stage AD patients treated with lecanemab in routine care at Tel Aviv Medical Center (TLVMC).

1) What the paper actually did (methods snapshot)

Setting & timeline: lecanemab administration began Nov 2023; screening/evaluation included patients assessed July 2023–Jan 2025.
Eligibility approach: biomarker-confirmed early AD via structured workflow including neurological evaluation, MRI, LP and/or amyloid PET, genetic testing (APOE), and MDT consensus.
Cognitive outcome: MMSE at baseline, 6 months, and 12 months.
Safety monitoring: scheduled MRI surveillance for ARIA (ARIA-E/ARIA-H) and infusion reaction observation as described in methods/results.
Analysis model: linear mixed-effects model for MMSE over time (ITT), with age subgrouping (≤74 vs ≥75) aligned to CLARITY-AD subgroup logic as described by authors.

2) Cohort flow & throughput (screen → treat → follow-up)

Interpretation constraints (skeptical):

Drop-off between treated and assessed-at-timepoints is substantial, particularly at 12 months, which can bias observed cognitive trajectories (attrition bias).

3) Clinical efficacy signal (MMSE change): what is statistically reported

6-month MMSE change (ITT): age-stratified significance

The paper reports significant MMSE decline over 6 months in ITT with mixed-model statistics (overall: F(1,45.13)=7.41, p=.009).
In subgroup analysis at 6 months: younger (n=31) shows F(1,24.67)=8.06, p=0.009; older (n=22) is non-significant (F(1,19.25)=0.67, p=0.424).
At 12 months (ITT n=31), the paper reports no significant MMSE change (F(1,17.18)=2.49, p=0.133).

Skeptical interpretation

Because there is no concurrent placebo/untreated control, “decline vs baseline within the treated group” does not separate treatment effects from expected natural history, practice/measurement variability, regression-to-the-mean, and selection of participants who successfully reach follow-up.

Also, MMSE may be insensitive to some AD-relevant cognitive domains and language-based syndromes, while the paper itself discusses diagnostic complexities where MMSE alone may be inadequate (e.g., language variants).

4) Safety: ARIA incidence and symptomatic severity

ARIA occurred in treated patients (TLVMC)

ARIA occurred in 16/86 (18.6%); most were asymptomatic; 1 patient had symptomatic moderate ARIA-E with ARIA-H and required hospitalization and IV steroids, then discontinued therapy.
The paper reports that an ARIA covariate (mixed model) did not significantly influence MMSE change (ARIA main effect p=0.264; interaction p=0.433).

Context: how to compare ARIA across studies

Direct ARIA-rate comparisons across cohorts are vulnerable to differences in: MRI protocols and reading practices, baseline microhemorrhage burden, exclusion criteria (e.g., APOE ε4 homozygotes), duration of MRI monitoring, and how “at-risk” cohorts are defined. The TLVMC paper explicitly excluded ApoE ε4 homozygotes based on safety profile considerations referenced from CLARITY-AD and later regulatory/adoption guidance.

This exclusion strategy is consistent with broader appropriate-use guidance focusing on ARIA risk mitigation.

5) Discontinuation & infusion reactions

Infusion-related reactions

Treatment discontinuation

The paper reports treatment discontinuation in 19.8% of patients (with reasons including ARIA, financial barriers, comorbidities, personal preference).
ARIA-related discontinuation is described further (e.g., one symptomatic ARIA leading to discontinuation; additional discontinuations after ARIA imaging findings).

6) Biomarker–cognition links (what they found, and what remains unknown)

No baseline CSF biomarker correlations with MMSE change (as reported)

They report no significant correlations between baseline MMSE (baseline, 6 months, or 12 months) and baseline CSF t-tau, p-tau181, or Aβ42.

Critical caveat: the paper does not report effect sizes or confidence intervals for these correlations in the provided text, and sample size at later timepoints is small, so null findings may be due to limited statistical power or unmeasured confounding.

7) Real-world “feasibility” claim: what is supported vs what is assumed

Supported by reported operational metrics

Authors describe building an infrastructure (MDT workflow, dedicated coordinator, on-site diagnostic capabilities) and report that 86 patients initiated treatment from 169 screened, implying workable throughput in their tertiary setting.

Assumptions / external validity concerns

Single-center real-world evidence may not transfer to other health systems with different MRI availability, MRI-reading workflows, infusion capacity, and payer coverage. The authors themselves emphasize infrastructure and MRI monitoring as essential.
Cost/insurance barriers are explicitly present: the treatment was not covered by Israel’s public health insurance and had to be paid out-of-pocket or via private insurance; this can create “healthy user”-like selection effects and distort the observed safety/efficacy mix.
No datasets were generated/analyzed for public sharing. While methodologically normal for some retrospective clinical workflows, this limits independent verification of statistical details and downstream re-analyses.

8) What the paper matches vs what it can’t answer (counterpoints)

Matches the trial literature directionally (with caution)

Authors frame their age-stratified signal at 6 months as “consistent with clinical trial data.” The paper references CLARITY-AD lecanemab results and subgroup rationale for age.

But the present paper’s causal inference is weaker because it does not include a placebo/control group, and MMSE is not the same as CLARITY-AD’s CDR-SB primary endpoint.

What remains unknown

Longer-term efficacy and safety durability beyond 12 months are not assessed in this interim reporting window; authors call for continued longitudinal follow-up.
Individual-level treatment response heterogeneity is not fully characterized; e.g., no multivariate model predicting who improves vs worsens is reported in the provided text (and no significant biomarker correlations were found).

9) Suggested “next paper” questions (what would falsify/clarify)

Include a contemporaneous control (e.g., matched untreated biomarker-confirmed early AD) to estimate absolute effect on cognitive trajectories relative to natural history. (This directly targets the paper’s “no control group” limitation.)
Report effect sizes + CIs for biomarker correlations and ARIA–MMSE relationships, not only p-values, to assess magnitude and uncertainty.
Standardize cognitive battery beyond MMSE or at least contextualize MMSE sensitivity, given authors’ own discussion about MMSE not capturing all relevant syndromic profiles.

10) Author reviews (go deeper)

Open BGPT author-review pages for every full author listed in the manuscript header.

Note: Competing interests are stated as “no competing interests” in this manuscript text provided.

Feedback:

Updated: July 09, 2026