对这款新药的报道完全相反，很多人看不懂

在一家报纸的标题里，人们急切想知道新款试验药是不是“近期阿尔茨海默症治疗的重大进展”。另一家媒体则表示该款药物“试验效果不佳”。

美国广播公司盛赞：“新款阿尔茨海默药物前景大好。”哥伦比亚广播公司则平淡地说：“阿尔茨海默新药试验结果令人失望。”

7月末，总部位于波士顿的生物医药公司百健（Biogen）和日本制药公司卫材（Eisai）宣布，研制18个月的二阶段抗体药BAN2401结果出炉，各界反馈就像哈哈镜：褒贬不一，差别仅在于立场。

位于英国伦敦的新闻机构“The Pharma Letter”总结很到位：“百健和卫材受到的影响主要因为阿尔茨海默症数据混乱。”

可能你想问，为什么会混乱？原因很多。但最主要有三个：

1. 判断早期认知衰退的手段非常杂乱。

多年来的大量阿尔茨海默症临床试验中，制药商主要对已发病患者进行测试。相比之下，百健和卫材的BAN2401主要针对早期或“先兆”阶段患者，症状通常是轻微的认知障碍。这种思路很聪明：在数十年的尝试中，没有药物可以逆转、阻止或减缓该病发展；但没准早期采取治疗效果会更好，这一策略在心脏病、高血压到癌症等方面都已证明有效。

虽然早期治疗疾病是明显进步，但还有个关键障碍：目前对阿尔茨海默症早期认知能力和精神官能变化还没有可靠的检测。

为此，卫材的团队独立开发了评判标准，叫ADCOMS，主要用于给阿尔茨海默病症状评分。根据该模型，研究期间服用BAN2401剂量最大的患者看起来状态不错：18个月里，认知能力下降比服用安慰剂的对照组少30%。在另一项更传统的测量，即ADAS-cog（或阿尔茨海默认知评估分）中，服用剂量大的患者情况更好。

但在第三项随意取名为临床痴呆评分总和（CDR-SB）的标准方面，BAN2401表现不佳，至少在统计学意义上无法证明比安慰剂更有效。尽管CDR-SB判定轻度认知障碍不太准确，但美国食品和药物管理局在判断阿尔茨海默氏症药物研究时，一直将其作为“临床标准”；虽然ADCOMS缩写看起来高级得多，问题是制定时间太短。

百健和卫材这次可能也是搬起石头砸自己的脚，7月初宣布得出“积极”结果，随后股票飙升，真正发布后外界才发现细节并不清楚。

2. 试验本身也存在问题。

先介绍下临床医学吧。其实临床医学并不是纯科学，而是更像手工艺加上一点派系。以后我会详细谈论该问题（如果有兴趣了解的话，我确实在书中写了一章讨论该话题）。现在先来看看BAN2401研究中两个典型的奇怪之处。

首先，研究小组从统计分析开始着手，产品不理想之后才改变思路。这就是为何某款追踪剂六个月前还很有效，一年后却变无效。

去年12月底，百健和卫材在研究刚好满一年时宣布，独立数据监测委员会认定该药在主要评估标准上“未达标”，即之前提到的ADCOMS标准。（类似的研究允许中期分析，以便可以出于安全或其他原因随时终止。）公司将失败归咎于统计技术不公......还表示研发仍将继续。

其次，试验可能从一开始就不平衡，意思是实验组和对照组可能匹配不均。正如一些分析师和生物技术资深观察人士指出，比起对照组，实验组内具有罹患阿尔茨海默症重大风险的人比例低得多。该决定并非公司做出，而是欧洲监管机构规定，因为监管机构担心具有遗传敏感性的病人服药测试后更可能出现脑损伤。但此举可能无意中影响了结果：由此可以判断，携带该指示基因APOe4的人可能在研究的18个月内状态下滑更严重。

看起来可能是偶然，但在进展极度缓慢的医学研究领域里，一点点突破也非常重要。

但也让我们非常困惑......

3. 我们的目标可能仍然是错的。

罹患阿尔茨海默症后，两种正常的中枢神经系统蛋白质由于某种原因在大脑皮层中大量积累，结果是神经细胞似乎窒息而死。第一种蛋白是β-淀粉样蛋白肽，如果不从大脑中清除就会大量包围神经细胞，堵塞突触并使周围组织发炎（可能是相关免疫反应的结果），最终杀死关键的脑细胞。第二种是扭曲的蛋白质纤维或名为tau的“神经纤维缠结”，在神经细胞内部聚集导致功能衰退。

β-淀粉样斑块和tau缠结是阿尔茨海默病的两个明显标志，该结论由一个多世纪前的阿洛伊斯·阿尔茨海默首次提出。但大多数试图阻止或逆转病程的治疗只针对第一种现象：神秘的淀粉样蛋白沉积。

BAN2401同样走了老路。最新试验中，靶向抗体清除大脑中的β-淀粉样蛋白斑块，无人质疑。剂量越高，清除得越干净。但正如一位专家所说，从现实世界的临床表现来看，该结果并未让人“震惊”。

BAN2401之前的几十种药物同样有针对性——而且很多情况下也能做到大量清除人类大脑中的有害蛋白质，无数的小鼠实验均已证明。但这种方法似乎并不能根治阿尔茨海默病，所以许多人开始怀疑，古老又长期被吹捧的“淀粉样蛋白假说”是否正确。

这让我想到众人皆知二阶段试验真正的“混乱”：百健和卫材希望在第三阶段找到什么全新的发现？两家公司已宣布全面推进下一个阶段，将涉及数百名志愿者，也需要更长时间。该研究将与百健手上另一项备受瞩目的抗淀粉样蛋白药物aducanumab试验同步完成，该药从Neurimmune获得授权，将与其他数十家公司共同进行。克利夫兰诊所Lou Ruvo脑健康中心的杰弗里·康明斯及其同事表示，截至今年年初，在中晚期临床试验中至少有32种淀粉样蛋白的追踪剂。

到目前为止，治疗阿尔茨海默症药物的失败率为99.6%。你没看错：基本上我们每次采用这种策略都会失败。

也许这就是为何这么多护理人员、阿尔茨海默症治疗倡导者、科学家，甚至还有一些生物技术投资者似乎对这项炒得火热的药物研究感到困惑。他们可能都想知道，为什么不试试新方法。（财富中文网）

译者：Charlie

审校：夏林

One headline wondered breathlessly if the new experimental drug was “the most promising development on Alzheimer’s in recent history.” Another said the same experimental medicine showed “little efficacy in trial.”

ABC News gushed: “New Alzheimer drug shows big promise.” CBS News said flatly: “Alzheimer’s drug results disappoint.”

When Biogen, the Boston biotech, and Eisai, a Japanese pharmaceutical company, announced the results of their 18-month Phase II clinical trial of an antibody called BAN2401 at end July, the responses were like the reflections in a funhouse mirror: wildly out of proportion, depending on where you stood.

The London–based news service, “The Pharma Letter,” summed it up well: “Biogen and Eisai suffer amid Alzheimer’s data confusion.”

Why the confusion, you ask? Lots of reasons. But here are three to start with:

1. Measures of early cognitive decline are surprisingly squishy.

In the great mass of Alzheimer’s trials that have been done over the years, drugmakers have largely tested their experimental agents on patients with full-on disease. Biogen and Eisai, by contrast, set out to evaluate BAN2401 in patients in the early or “prodromal” stage, who typically have only mild cognitive impairment. The thinking here was (and is) smart: In decades’ worth of attempts, no drug has been able to reverse, stop, or even slow this pathological terror; perhaps attacking it earlier in its progression—a strategy that has proven effective in everything from heart disease to hypertension to cancer—can yield a better result.

While attacking the disease earlier is a step forward, certainly, there’s one key snag: There aren’t great tests yet for measuring change in cognitive ability and mental function at early stages of this disease.

So the team at Eisai developed their own—called ADCOMS, for Alzheimer’s Disease Composite Score. And indeed, patients receiving the highest dosage of BAN2401 during the study seemed to do pretty well, according to this bespoke model: showing 30% less cognitive decline over the 18-month stretch than those in the placebo arm. In another, more traditional measure—ADAS-cog (or Alzheimer’s Disease Assessment Scale-cognitive subscale)—high-dose patients seemed to fare even better.

But in a third measure—the awkwardly named Clinical Dementia Rating Sum of Boxes—BAN2401 failed to demonstrate, to the point of statistical significance at least, that it was more effective than the placebo. While CDR-SB is a fairly insensitive measure for mild cognitive impairment, the FDA has long recognized it as an appropriate “clinical endpoint” for Alzheimer’s drug studies; and ADCOMS, of course—despite the far superior acronym—is the new kid on the block.

Biogen and Eisai may have also shot themselves in the foot by announcing early in July that they had “positive” results—allowing their stocks to soar—while leaving the not-so-clear details to be sorted out later.

2. This trial had its share of Mulligans.

Here’s the thing about clinical science. It mostly isn’t one—a pure science, that is. Rather, it’s a form of artisanship mixed with a little partisanship. I’ll dwell on this more another time (and, well, I did write a book chapter on this, for those inclined). But for now, consider two somewhat classic oddities of the BAN2401 study.

First, is that the research team began with one statistical analysis— then seemed to change their mind when it didn’t deliver the goods. That’s how an investigative agent could appear ineffective after a year, but curiously effective just six months later.

In late December last year, at the one-year mark of the study, Biogen and Eisai announced that an Independent Data Monitoring Committee had determined that the drug “did not meet the criteria for success” in their primary endpoint, the ADCOMS measure noted above. (Studies like this one are designed to allow such interim analyses so that they can be halted for safety or other reasons.) The companies blamed the failure on a statistical technique…and said they’d keep going anyway.

Second, is that the trial may have been off-balance from the get-go—which is to say the experimental and control groups may not have been so evenly matched. As several analysts and veteran biotech watchers have pointed out, the share of people with a major risk factor for Alzheimer’s was much smaller in the pivotal drug arm than it was in the group getting placebo. That decision was due to European regulators, not to the companies, who were concerned that patients with this particular genetic susceptibility would also be at greater risk of brain injury if they received the drug. But it also may have inadvertently skewed the results: It’s not unreasonable to think that carriers of this telltale gene (APOe4) could have declined more precipitously during the 18 months of the study, making the response of those in the experimental arm look comparatively better.

That may seem like an incidental thing, but in a medical crusade that has long been measured in inches, such asterisks matter immensely.

Which brings us to the main reason for confusion…

3. We may still be chasing the wrong damn target.

In Alzheimer’s, neurons are seemingly choked to death by two types of normal central nervous system proteins that, for some reason, begin to accumulate in large quantities in the cerebral cortex. The first are beta-amyloid peptides that, when not cleared from the brain, begin to encircle neurons en masse, clogging their synapses, inflaming surrounding tissue (perhaps as the result of an associated immune response), and ultimately kill off critical brain cells. The second are twisted protein fibers or “tangles” called tau, which aggregate within the neurons themselves and cause degeneration.

The beta-amyloid plaques and the tau tangles are the twin hallmarks of Alzheimer’s disease—a conclusion first drawn by Alois Alzheimer more than a century ago—but most efforts at stopping or reversing the disease have focused only on the former phenomenon: the mysterious amyloid deposits.

BAN2401 follows the same worn track. And in this latest trial, what no one doubts is that the targeted antibody does a bang-up job of clearing beta-amyloid plaques from the brain. The higher the dose, in fact, the more it wipes it clean. But as one expert put it, that hasn’t exactly resulted in “shock and awe” in terms of a real-world clinical effect.

Dozens of drugs before BAN2401 have likewise targeted—and, in many cases, sharply reduced—this glomming protein from the human brain, and from an untold number of mouse models too. But this approach doesn’t seem to flip the Alzheimer’s switch, causing many in the field to doubt whether the ancient and vaunted “amyloid hypothesis” is the right one.

Which brings me to the real “confusion” of this well publicized Phase II trial: What do Biogen and Eisai hope to discover in Phase III that’s different from what we see now? The two companies have announced full steam ahead for the next study phase—which will involve hundreds more volunteers and take even longer to complete. That one will be done alongside Biogen’s ongoing trial of another high-profile anti-amyloid agent, aducanumab, which the biotech has licensed from Neurimmune—which will be done alongside dozens of others. As of the start of the year, there were no less than 32 investigative agents targeting amyloid in mid- to late-stage clinical trials, according to Jeffrey Cummings and colleagues at the Cleveland Clinic Lou Ruvo Center for Brain Health.

So far the failure rate for Alzheimer’s drugs is 99.6%. That’s right: Virtually every time we’ve tried this strategy, it has failed.

Maybe that’s why so many caregivers, Alzheimer’s advocates, research scientists—and perhaps even a few biotech investors—seem so befuddled about this latest well-hyped drug study. Maybe they’re wondering why we’re not trying something else.