Future models will hopefully stop misleading about “timed gained” with anti-amyloid monoclonal antibodies.

Yes. The biophysical framework is better at distinguishing between normal and accelerated aging than the clinicopathologic framework. In normal aging, there is plenty of pathology 'accumulation'. Degeneration may only happen when the precipitation of monomeric peptides exceeds their replacement.

That's my hope, Elaine.

I have submitted it for publication as a viewpoint, but will be thinking of other strategies too.

Thank you, Emilia. So far, the reaction is relatively subdued. I am unsure how far this view on the open-label extension has gotten.

(5/5) Bottom line: The 40% ‘slower decline’ holds only if we compare the results to a historic cohort (link to earlier post below). But compared to the model, patients decline 40% faster than anticipated. We expected larger benefits, but they instead shrink rapidly.
bsky.app/profile/albe...

(4/9) Why does lecanemab look better? It is compared to a steeper-than-modeled slope from ADNI. Whereas the newly modeled decline is 0.05/mo ((1.5-0.6)/18), the observed decline is 0.07/mo ((1.8-0.5)/18), which means an actual 40% acceleration of decline ((0.07-0.05)/0.05)* 100).

(3/9) 2025 𝐥𝐞𝐜𝐚𝐧𝐞𝐦𝐚𝐛 𝐝𝐚𝐭𝐚: At 18 months (0.8-0.5 = 0.3) and 36 months (2.0-1.8 = 0.2), lecanemab slowed the CDR-SB decline by 37.5% vs placebo in the first period (0.3 / 0.8) \* 100). That difference narrowed to 10% in the second (0.2 / 2.0) \* 100).

(2/9) 2024 𝐬𝐢𝐦𝐮𝐥𝐚𝐭𝐢𝐨𝐧: At 18 months (0.8-0.6 = 0.2) and 36 months (2.0-1.5 = 0.5), lecanemab 𝐬𝐥𝐨𝐰𝐞𝐝 𝐭𝐡𝐞 𝐂𝐃𝐑-𝐒𝐁 𝐝𝐞𝐜𝐥𝐢𝐧𝐞 by 25% vs placebo at both timepoints (0.2 / 0.8) \* 100) & (0.5 / 2.0) \* 100).

Those are not included, making the slope of treatment look better than it actually is.

(9/9) Bottom line: The illusion of ‘increasing benefit’ driven by survivor bias and historical comparisons masks the accelerated decline on treatment. Open-label extension data are incompatible with disease modification. Clinicians, patients, and policymakers beware!

(8/9) All caveats aside, the steepening (worsening) of the slopes of decline for lecanemab and donanemab suggests that patients deteriorate more rapidly the longer they are on treatment, a pattern inconsistent with an increasing therapeutic effect.

(7/9) Add the survivor bias: By the end of the OLE, only 56% of patients remained on lecanemab, 53% on donanemab. The disproportionate influence of better-tolerating, slower-progressing participants enriching the open-label extension skews the mean CDR-SB changes.

(6/9) Comparing the extrapolated slope of the placebo arm vs the observed donanemab arm, the CDR-SB difference favoring donanemab shrinks from 0.6 at 18 months to 0.3 at 36 months. Versus the ADNI slope used for lecanemab there would be no difference. This ADNI slope is steeper.

(5/9) Comparing the extrapolated slope of the placebo arm vs the observed lecanemab arm, the CDR-SB difference favoring lecanemab does not expand but shrinks: from 0.5 at the end of the double-blind period to 0.2 at month 48 of the open-label extension.

(4/9) So why do the curves shown at AAIC look like the gap widens over time? Because instead of a concurrent placebo group, results were compared to an 𝗲𝘅𝘁𝗲𝗿𝗻𝗮𝗹 𝗵𝗶𝘀𝘁𝗼𝗿𝗶𝗰𝗮𝗹 𝗰𝗼𝗵𝗼𝗿𝘁 (ADNI)—a method riddled with bias (inclusion criteria, dropout rates, etc.).