The pattern nobody in the niche is writing

Three unrelated drug classes, the same six months, the same shape. In each, a lab number moved in the right direction and the newsletters called it a breakthrough. In each, the thing you'd actually care about, does the person think more clearly, live longer, get stronger, stay unmeasured or unchanged. This is the article the space isn't writing, and it suits the skeptic who runs the tests: you already know a moving biomarker is a hypothesis, not a happy ending.

Case 1: Semaglutide and Alzheimer's, the translational paradox

The EVOKE and EVOKE+ trials, published in The Lancet on 2026-03-19, were as serious as a trial gets: phase 3, 566 sites, 40 countries. Semaglutide has every mechanistic reason to help the aging brain, and the hypothesis had real momentum. The result: no clinical benefit. A published review named the gap plainly, "the translational paradox," the distance between a drug that should work on paper and one that does nothing for patients.

Lancet 2026: 10.1016/s0140-6736(26)00459-9. Surfaced in this corpus only through an Elizabeth Yurth transcript; all three literature sweeps missed it, which is the whole argument for mining the videos and the papers together.

Case 2: Obicetrapib and the APOE4 brain (this one is yours)

This is the item most relevant to an APOE4/4 reader, and it's the cleanest illustration of the pattern. In a prespecified substudy of the BROADWAY phase 3 trial (n=1,535 of 2,530), the oral CETP inhibitor obicetrapib moved Alzheimer's blood markers in exactly the direction you'd want. In APOE4/4 carriers, p-tau217 fell 7.81% on the drug while it rose 12.67% on placebo, a roughly 20-point placebo-adjusted drop (P=0.010). GFAP and NfL moved too. The authors called it "the first demonstration of an oral intervention capable of reducing both beta-amyloid and tau pathology biomarkers in ApoE4 carriers." And lower achieved LDL, Lp(a), and ApoB all tracked with the falling p-tau217, which is the line that connects your own tracked metric to your genotype.

J Prev Alz Dis, open access: 10.1016/j.tjpad.2025.100394 (PMC12811769). Published online 2025-10-17.

Case 3: Muscle drugs that preserve the number, not the strength

Apitegromab and bimagrumab are built to protect muscle during weight loss, and by the number on the DXA scan they do it, lean mass is preserved and the weight that comes off is up to 92.2% fat. That's the headline. What's missing is the point: grip strength, sit-to-stand, and physical function were all flat. Preserved lean mass is not preserved strength, and for a woman whose real goal is staying functional into her eighties, the surrogate and the outcome quietly part ways.

The regulatory reversal, and the overreach hiding inside it

On 2026-02-12 the FDA removed the cardiovascular, breast cancer, and probable-dementia language from the boxed warning on menopausal hormone therapy, six products including Estring, so low-dose vaginal estrogen loses its systemic warning language, with 29 more pending. Frame this as a regulatory reversal with the evidence review still unpublished, not as a new scientific finding.

Because in the same document the FDA overreaches. It claims HRT started within 10 years of menopause cuts all-cause mortality and fractures, naming no trial, citing zero references, and never mentioning the WHI. Nanette Santoro's read: it "goes very far beyond the data and is in exactly no clinical guidelines." Manson's own 18-year follow-up (JAMA 2017) attenuates the mortality signal to non-significance. The tension between the reversal and the overreach is the article.

APOE4 and lecanemab: the number they published isn't your number

Also squarely yours. The UK MHRA and the EMA exclude APOE4 homozygotes from lecanemab outright, with mandatory genotyping; the FDA only warns. And the widely circulated 8.9% rate of ARIA-E (brain swelling) comes from a table titled "non-carriers and heterozygotes." Clarity AD enrolled 1,795 patients but the reported arms sum to 1,521, so roughly 274 homozygotes are simply absent from that number. The homozygote estimate is closer to 32.6% ARIA-E, about 9.2% symptomatic. The headline safety figure was never describing people with your genotype.

Standing context: the MHRA exclusion is from August 2024, not new.

Testosterone: the null was not reversed

For all the noise, the 2026 literature restated the 2019 global consensus rather than overturning it. Three papers this year land in the same place: the postmenopausal benefit for desire is real but modest (roughly 1.4 extra satisfying events a month), the premenopausal evidence cupboard still holds just two trials, and zero cardiovascular events across 2,628 women over 52 weeks or less is absence of evidence, which the authors say themselves. The Monash Sybil study is the one to watch; it's ongoing and unreported.

What's genuinely unstudied, and squarely her beat

The empty shelves are as telling as the full ones. Nobody has published what happens to lean mass on tirzepatide in postmenopausal women on HRT. The HRT-by-GLP-1 interaction, resistance training plus a GLP-1, and weight regain after stopping are all blank, not because they don't matter, but because the trials haven't been run. That gap is the opportunity.