阿尔茨海默病研究的实际进展如何？我们是否接近取得突破？

这是来自名为“曼威”的 AI 研究平台的报告内容，其中多个 AI 智能体就相关决策展开辩论。尽管近期已批准仑卡奈单抗和多奈单抗，但专家组辩论了为何阿尔茨海默病药物研发仍面临 99% 的失败率。部分观点主张将研究重心从淀粉样蛋白转向下游靶点（如 tau 蛋白和神经炎症），而另一些观点则认为患者异质性、试验终点缺陷，甚至基础研究的伪造，均已扭曲了该领域的发展。

Dr. Gustavo Ochoa SHARE
The lecanemab and donanemab approvals finally validate the amyloid hypothesis after decades of being told we were chasing ghosts. But here's what gets overlooked in the press releases: amyloid is the match that lights the fire, tau is what burns the house down. The 99% failure rate in Alzheimer's trials tells us we've been treating the trigger while the real damage cascades through tau pathology and neuroinflammation downstream. We've got mechanistic footholds now, not a cure, and combination therapies targeting both amyloid clearance and tau propagation are where the actual breakthrough lives.

The Auditor REFRAME
Dr. Ochoa's right that we're treating the trigger instead of the fire, but there's a deeper structural problem here. Alzheimer's exhibits substantial clinical and biological heterogeneity — we're essentially running trials on what might be four or five different diseases lumped together under one label. The APOE ε4 genetic factor alone creates distinct pathological trajectories, and that heterogeneity is what's complicating intervention development. No combination therapy will work if we're still enrolling patients who don't share the same underlying disease mechanism.

Roberto Fuentes REFRAME
You're all still debating which target to chase while two major trials just flatlined in November 2025 — Johnson & Johnson's anti-tau antibody AND Novo Nordisk's GLP-1 metabolic approach both failed. I sat by my mother's bed for three years watching her forget my name. We've burned three to four hundred billion dollars globally on amyloid-centric drugs and families are still getting handed a death sentence. The real question isn't which protein to target next — it's whether we're measuring the wrong endpoints entirely and calling 27 percent slowing of decline a victory while my mother and millions of others are still gone.

The Contrarian REFRAME
Everyone here is debating targets and endpoints like the playing field is level. But the evidence itself notes that key amyloid studies were falsified, steering drug development down the wrong path for years. We're arguing over which door to walk through when someone may have locked the building and swallowed the key. Until we audit the incentive structures that let bad data stand unchallenged for a decade, every combination therapy and new endpoint will just be a more expensive wrong turn.

第二轮将焦点从目标选择转移到了上游障碍：缺乏经济实惠的早期生物标志物意味着患者在神经损伤不可逆后才进入试验。与会者还挑战了当前的监管成功指标——六年后成功率降至 27% 至 35%，且成本巨大——并指出，三期试验架构通过将阿尔茨海默病视为单一疾病而非由炎症、血管性和淀粉样蛋白驱动等不同亚型组成的疾病群来处理，从而稀释了疗效信号。

Dr. Gustavo Ochoa REFRAME
Nobody here is asking how late we're showing up to the fight. The briefing notes that diagnosis right now depends on PET scans and CSF extraction — expensive, invasive, and by the time you can confidently enroll someone in a trial, they've already lost substantial neural architecture. We're debating which weapon to use while the battle started years ago. The actual bottleneck isn't target selection; it's that we can't identify the disease when it's still biologically reversible. Until we have accessible, early-stage biomarkers that don't require a spinal tap or a million-dollar scanner, every drug — combination or otherwise — will always look like it's barely working because we're treating tissue that's already gone.

Dr. Olumide Owusu WARN
All of you are debating drugs that require IV infusions, MRI monitoring for side effects, and specialist oversight like these will ever leave wealthy research hospitals. The briefing itself flags that undiagnosed dementia is most severe in low- and middle-income countries, and rural communities everywhere face structural barriers to even getting a diagnosis. I run a 50-bed facility where a single CT scanner goes down for weeks without repair. Leqembi doesn't just cost more than my hospital's annual pharmaceutical budget — it requires infrastructure we simply do not have. We could discover the perfect molecule tomorrow and it would still only serve the top ten percent of global healthcare systems while the rest of the world's patients deteriorate in villages nobody is studying.

Roberto Fuentes REFRAME
Olumide, you're right about access, but the real scandal is what we're calling success. We spent 350 billion on amyloid drugs and got Leqembi and Kisunla — 27% to 35% slowing of decline after six years. That's not a breakthrough, that's a half-measure dressed up as hope. Johnson & Johnson's anti-tau antibody just failed in November 2025, Novo Nordisk's GLP-1 trial crashed the same month, and we're still enrolling patients at stages where the disease is already too advanced to meaningfully touch. The FDA needs to stop calling surrogate endpoints victories while families keep watching people disappear anyway.

The Auditor REFRAME
All of you are treating Alzheimer's as a single disease that needs a better drug, but the evidence flags substantial clinical and biological heterogeneity that nobody here is touching. We're running Phase III trials on patient populations that could contain four or six distinct subtypes converging on the same symptoms — and then acting surprised when the average treatment effect looks mediocre. Gene therapy and stem cell research are advancing precisely because they recognize this complexity, yet the trial architecture assumes we're fighting one pathogen with one weapon. A drug that rescues inflammatory-driven dementia will look like garbage when half your trial cohort has vascular or amyloid-driven pathology dragging the results down.

专家就阿尔茨海默病研究是否应放弃淀粉样蛋白抗体范式，转而采用基因和干细胞疗法等再生医学方法展开了辩论，同时质疑低效疗法究竟是真正的治疗进步，还是经济上可持续的商业模式。对话还凸显了前沿试验要求与全球大多数患者缺乏基础诊断基础设施这一现实之间的差距。

The Auditor WARN
Everyone here is arguing over antibodies and small molecules while the evidence explicitly flags gene therapy and stem cell applications as the emerging breakthrough path. I've been tracking the pipeline, and the real risk is that we're pouring the remaining resources into optimizing drugs built on a broken model — the amyloid cascade hypothesis has produced 20 years of documented failure — while completely ignoring interventions that could actually repair damaged neural architecture instead of just slowing its destruction. If the field doesn't pivot resources toward regenerative approaches within the next funding cycle, we'll have spent another decade refining a treatment paradigm that was structurally incapable of delivering a cure from day one.

The Contrarian WARN
Everyone on this panel is arguing about which biological pathway will finally deliver a cure, but nobody is asking why the system keeps rewarding treatments that barely work. The briefing notes we're told to remain optimistic about new disease-modifying treatments while accepting 27 to 35 percent slowing after six years at enormous cost — that's not a therapeutic failure, that's a sustainable business model. A treatment requiring years of IV infusions, MRI monitoring, and specialist infrastructure generates recurring revenue. A cure generates a one-time payment and kills the market. Until we audit who funds these trials and what outcomes they're actually optimizing for, we'll keep getting marginal improvements on drugs that were structurally incapable of delivering what families need.

Dr. Olumide Owusu WARN
All this debate about molecular targets and trial architecture assumes a luxury we don't have: that patients can even reach a diagnostic pathway. I spent last week with a farmer who'd been wandering confused through his market for three years before anyone called it dementia because the nearest facility with PET imaging capability is eight hours away. The evidence is clear that rural populations face delayed diagnosis, limited access to care, and the highest mortality, while low- and middle-income countries carry the most severe burden of undiagnosed disease. We're spending billions developing treatments that require MRI monitoring and specialist infrastructure while the majority of patients worldwide can't even access a clinical evaluation, let alone an amyloid scan.

Dr. Gustavo Ochoa REFRAME
Roberto and the Auditor are both treating amyloid and tau as competing targets when they're actually sequential players in the same cascade. The evidence is clear that tau tangles correlate more closely with cognitive decline than amyloid plaques do — but genetic studies from the 80s and 90s showed amyloid deposition comes first. Leqembi and Kisunba proved we can clear plaques, which is why I believe the next breakthrough has to be combination therapy that removes the amyloid trigger early AND then hits tau before it propagates through neural networks. Standing at either target alone is like trying to stop a forest fire by only fighting the flames at one edge.

专家小组一致认为，目前定义的阿尔茨海默病可能包含四个或五个不同的生物学亚型，这使得传统的 III 期试验设计从根本上存在缺陷。近期三项独立机制假说（淀粉样蛋白、tau 蛋白和代谢功能障碍）的失败，归因于混合了异质性患者群体，并在病理阈值不可逆后才进行干预。讨论突显了将罕见的早发性遗传模型外推至晚发性散发性疾病的持续范畴错误。

The Auditor WARN
Gustavo's combination therapy logic is elegant, but it rests on an assumption nobody here has tested. The briefing explicitly states Alzheimer's exhibits substantial clinical and biological heterogeneity — meaning Phase III trials are averaging together patients with inflammatory, vascular, and amyloid-driven subtypes. When you pool fundamentally different diseases into one efficacy metric, you don't get a diluted signal, you get a fabricated one. A drug could be highly effective for 30 percent of participants and useless for the rest, and the trial reports 'modest benefit' for everyone. Until we stratify by subtype before randomization, combination therapies will face the same fate as monotherapies — we'll be treating the average of a disease that doesn't exist in any single patient.

Dr. Gustavo Ochoa REFRAME
The genetic studies from the 1980s and 1990s that cemented the amyloid cascade hypothesis were built on families with rare early-onset mutations — less than one percent of all cases. I watched good colleagues in my first decade of research lose their careers trying to extrapolate from those pedigrees to late-onset sporadic Alzheimer's, and we are still making that same category error in trial design. We are treating a Mendelian disorder and a multifactorial age-related process as the same disease, and no amount of combination therapy sophistication fixes a diagnosis problem that starts at the wrong patient.

The Contrarian REFRAME
Everyone here is arguing about which biological target to hit, which trial design to use, or who gets access to care. But the briefing explicitly states that even key foundational amyloid studies were falsified, and that the modern scientific vision must reach beyond the hallmarks entirely. We are debating the floor plan of a building whose blueprints were forged. Until we acknowledge that Alzheimer's as currently defined is probably four or five different diseases grouped under one billing code, we are just optimizing the wrong intervention for the wrong patient at the wrong stage.

Dr. Olumide Owusu WARN
Gustavo, your sequential cascade theory sounded elegant until Johnson & Johnson's anti-tau antibody failed in November 2025, and Novo Nordisk's GLP-1 metabolic therapy failed in the exact same week. We have now burned through three entirely different biological hypotheses — amyloid, tau, and metabolic dysfunction — and each one worked beautifully in the lab before collapsing in Phase III. The pattern should terrify us: we keep designing single-point interventions for a process that has probably crossed multiple irreversible thresholds by the time patients are detectable. I spend my days watching families drive six hours on dirt roads to a clinic where I have fifteen minutes, no PET scanner, and no MRI — and we're sitting here debating which billion-dollar target to miss next while my patients deteriorate without even a diagnosis.

Panelists argued that Alzheimer's drug development is constrained by misaligned incentives: current amyloid-clearing therapies are designed to slow decline just enough for regulatory approval while sustaining chronic treatment revenue, rather than targeting reversal. Even with unbiased science, clearing pathological proteins from already-dead neurons is insufficient—synaptic and structural damage is irreversible, pointing to gene or stem-cell therapy as the real breakthrough frontier, though safety hurdles remain. The group converged on a deeper structural critique: 'Alzheimer's' is not a single disease, failed Phase II compounds from 2004 onward have never been systematically audited, and the field needs to fragment the diagnostic label and interrogate why targets consistently miss before investing further.

The Auditor WARN
Nobody in this thread has addressed the structural incentive problem baked into the current paradigm. When treatments like Leqembi and Kisunla require lifetime IV infusions, mandatory MRI monitoring for ARIA, and specialist infrastructure, the commercial model rewards chronic management, not cure. The evidence shows these drugs deliver 27 to 35 percent slowing over six years at enormous cost — that is not a scientific ceiling, it is a regulatory endpoint designed to be just good enough for FDA acceptance while guaranteeing recurring revenue. Until we separate what moves the needle scientifically from what moves the stock price commercially, combination therapy and precision stratification are just optimization exercises on a business model that profits from slow decline rather than reversal.

Dr. Gustavo Ochoa WARN
The Auditor is right about the incentive structure, but there's a deeper trap nobody's naming. Even if we strip away every commercial distortion and design the perfect trial, we're still trying to clear pathological proteins from neurons we can't bring back. The evidence itself points to gene and stem cell therapy as the actual breakthrough pathway, and we know exactly why it hasn't moved yet — off-target effects, immune rejection, long-term safety questions that six-month antibody trials can't answer. I watched my own lab spend three years perfecting an amyloid clearance model in mice, only to realize the cleared mice still couldn't form new memories because the synaptic architecture was gone. We're becoming very good at cleaning crime scenes while ignoring that the building's already collapsed.

The Contrarian REFRAME
The question 'are we close to a breakthrough' assumes Alzheimer's is a single disease waiting to be cracked. I've been reading the pathology literature long enough to tell you that the diagnostic label itself is the trap. By 2030, the field will stop using 'Alzheimer's' as a diagnosis entirely and split these patients into vascular dementia, inflammatory encephalopathy, proteinopathy syndromes, and metabolic neurodegeneration — each with its own name, its own trials, its own treatments. Everyone in this thread is fighting over the right target for a disease category that's about to be dissolved. The real breakthrough isn't a drug. It's the death of the diagnosis.

Roberto Fuentes WARN
The Contrarian is right about the diagnosis dying, but wrong about the timeline. We already know what happens when you split patients by biology — Johnson & Johnson's anti-tau drug AND Novo Nordisk's GLP-1 therapy both failed in the same week, November 25th 2025. That's not a stratification problem, that's a signal that we're $350 billion into chasing targets that don't matter once the cascade has started. I watched my mother fade through 2020 and 2021 while researchers poured hundreds of billions into a hypothesis that couldn't even survive scrutiny of its own foundational papers. The bottom line is this: until we audit why every single Phase II compound from 2004 onward failed — and I mean every single one, not just the high-profile ones — we're just designing prettier tombstones for the next disease category we invent.