Lecanemab After the 2026 Cochrane Review: One Patient Against a Pooled Average
A highly motivated, early-stage patient wants an anti-amyloid antibody after reading about a new pooled review that found the class's average benefit too small to matter. The disagreement is whether a discouraging population average should govern a specific, biomarker-matched individual.
Susan T., 66, still works two days a week as a tax accountant and intends to keep doing so as long as she can, in a firm she co-founded twenty-two years ago and still counts as her second family. She noticed her own decline before anyone else did — transposed numbers she used to catch instantly, a client meeting she forgot entirely last fall — and pursued the workup herself, including amyloid-PET imaging that came back positive. Genetic testing she requested independently showed she carries one copy of APOE4; her mother developed Alzheimer's disease in her seventies, a family history Susan has known about and quietly dreaded for two decades. Her MMSE is 26; she describes her own diagnosis with a directness that has clearly helped her cope: "I know what's coming. I want to try the thing that might slow it down."
She has read about lecanemab and returns to clinic today asking to start it. Her presentation is, in a specific and relevant sense, exactly what the anti-amyloid antibody trials were built to enroll: early-stage, biomarker-confirmed disease, still functionally independent. But a Cochrane review published in April 2026, pooling seventeen trials across the anti-amyloid antibody class and more than twenty thousand participants, concluded that average cognitive and functional benefit at roughly eighteen months was absent or too small to be clinically meaningful, while amyloid-related imaging abnormalities and bleeding-related harms were increased. Whether a pooled figure assembled largely from agents that never reached practice should govern a decision about one that did is precisely what the team has to work out. Her own APOE4 status raises her individual risk of exactly that harm above the trial population's average, the same population average the review just found offered, at best, a marginal benefit.
In the memory disorders clinic, reviewing the imaging together
I think she's a legitimate candidate. Clarity AD — van Dyck and colleagues, New England Journal of Medicine, 2023, 1,795 patients with early symptomatic disease — showed lecanemab significantly slowed decline on the CDR-Sum of Boxes over eighteen months compared with placebo. Donanemab's pivotal trial, TRAILBLAZER-ALZ 2, published in JAMA in 2023, showed a similar pattern using a different composite scale. Her profile — MMSE 26, biomarker-confirmed, still working — is close to the enrolled population in both.
I'm not disputing either trial's statistical significance. I'm saying the Cochrane review published this April — Nonino and colleagues, pooling seventeen trials and over twenty thousand participants across this drug class — found average benefit at that same roughly eighteen-month mark was absent or too small to be clinically important once pooled, while ARIA and bleeding-related harms increased. In Clarity AD specifically, total ARIA incidence was 21.3% on lecanemab versus 9.3% on placebo. In TRAILBLAZER-ALZ 2, three ARIA cases were fatal.
Before we let “absent or too small to be clinically important once pooled” stand as the last word on lecanemab specifically: only two of those seventeen trials studied a drug anyone can actually prescribe today. The other fifteen were bapineuzumab, solanezumab, crenezumab, gantenerumab — agents that failed and were abandoned, several of which barely cleared amyloid at all. Pooling a drug that hit its endpoint with fifteen that didn't, and reporting the average, tells you about the class as a research programme. It doesn't tell Susan what happens if she takes this one. That's the specific objection the review drew from clinicians when it came out, and I don't think it's special pleading.
She's APOE4-positive. That's not incidental — ARIA risk in these trials tracked directly with APOE4 carrier status, which means her genotype moves her risk in exactly the direction the pooled review is warning about, on a benefit the same review found was, on average, too small to clearly justify it.
Both of those are real, and I don't think either one should get quietly dropped from the conversation with Susan. What I'd add is that her APOE4 status isn't just a reason for more caution in the abstract — it's specific, actionable information. A secondary ARIA analysis of the TRAILBLAZER-ALZ trials found that adding a scan at four weeks, on top of the standard monitoring schedule, reduced the risk of symptomatic ARIA-E by 36% — though only about a sixth of donanemab-treated patients actually received that extra scan, so it's a real signal on a modest denominator rather than a settled protocol.
I'd frame this less as yes-or-no and more as: if she chooses to proceed with full understanding of her own elevated ARIA risk, she should get the most conservative monitoring protocol available, not the standard one.
Agreed: no infusion started today. Susan is referred for formal genetic counseling given her APOE4 result, and the enhanced four-week MRI monitoring protocol is adopted as the standard going forward regardless of her eventual decision on lecanemab itself.
Lecanemab is started under the enhanced monitoring protocol, with an explicit early-stopping threshold for any ARIA finding, symptomatic or not.
Care continues with standard symptomatic management and cognitive monitoring, revisited if her genotype counseling changes her view or if newer, more favorable data for her specific risk profile emerge.
Not agreed: whether the pooled Cochrane finding or the individual pivotal-trial results should carry more weight in how the team itself frames the decision for a well-informed, motivated patient. The neurologist thinks a capable adult with this profile should be allowed to weigh a modest-but-real individual trial benefit for herself; the pharmacologist thinks the pooled null finding, plus her elevated genotype-specific risk, means the team owes her a clearer recommendation against, not just balanced information.