Being cognitively sophisticated and also long-lived has its dark side. The distinctly human style of navigating existence, our facility for reason, memory, empathy, and imagination, depends on a vast and still unfathomed network of cells in the brain. If too many of these cells die, we become unmoored. Memory falters, personality changes, language slips, the grasp on complexity vanishes. Humans are not uniquely susceptible to diseases of neurodegeneration, but I’d wager we’re unique in our fear of them and in the magnitude of the attendant suffering.

Dementia, roughly meaning “out of mind,” the major consequence of profound neuronal loss, comes in many and overlapping forms. The most prevalent by current definitions is Alzheimer’s disease, estimated by the World Health Organization to represent about two thirds of all cases of dementia. There are, broadly, two types of Alzheimer’s: the rare, early-onset, familial form (about 1 percent of cases) and the late-onset, sporadic form (the remaining 99 percent). Dementia can also develop in more than half of people with Parkinson’s disease and manifests in other, rarer conditions as well.

It’s now almost a commonplace that the prevalence of dementia is expected to go up mightily over the next decades and affect a projected 152 million people worldwide by 2050. This is because dementia is predominantly a condition of aging, and people are living longer thanks to biomedical science and healthier lifestyles. Less well known is the fact that, adjusted for age, new diagnoses of the condition have been going down, at least in some of the countries studied, over the past decades. Again, this most probably reflects aspects of general health that reduce risk—better control of blood pressure, for instance, or increased exercise. Still, there are now more than 55 million people estimated to be living “out of mind,” and tackling diseases of neurodegeneration is one of the more intense efforts of biomedical science. There have been recent blips of hope, but it’s proving a stubborn challenge.

In How Not to Study a Disease, a lucid, knowledgeable, but not entirely objective critique of the field of Alzheimer’s research, Karl Herrup, a professor of neurobiology at the University of Pittsburgh School of Medicine, analyzes this challenge. He describes the complexity of even defining the disease; tells of the excitement of Alzheimer’s research in the 1990s, when its cause was thought to be understood and treatments seemed imminent; and charts the long and expensive trickle of disappointment in the decades since, as it has become clear that the disease is much more complicated than was initially thought. As his title signals, Herrup criticizes the field for its adherence to a causal theory of Alzheimer’s that he argues does not hold up.

Between 1906 and 1911 Alois Alzheimer, a doctor and researcher heading a laboratory of neuroanatomy at the Nervenklinik in Munich, reported pathological structures in the brains of a handful of people with presenile dementia, the early-onset form of the condition. Alzheimer and his colleagues had observed, under a microscope, unusual structures in their patients’ brains after their deaths; these are now called plaques and tangles. The scientists did not know of what material these structures were made, and this would remain unknown for well over half a century.

The answer emerged in an exhilarating story of discovery in the 1980s and 1990s, a convergence of different lines of inquiry on a particular macromolecule as a central player in the rare familial type of Alzheimer’s disease. One set of scientists were biochemical sleuths, studying what is physically in the plaques; they found bits of the clump-prone amyloid precursor protein. A second group, the geneticists, pursued the genes that are mutated in families in which Alzheimer’s runs, and found just three disease genes that all pointed to the same culprit, amyloid precursor protein. These (and related) discoveries produced a powerful conviction that the cause of Alzheimer’s had been identified, and formed the basis of the amyloid cascade hypothesis, for many years the dominant theory about what goes wrong in the disease.

Simply put, the hypothesis says that in the brains of some people, short, toxic forms of the amyloid precursor protein clump up into the dense, sticky protein masses called plaques and set into motion the events that cause Alzheimer’s disease. These events involve the mysterious spread of tangled-up versions of another protein called tau, which in turn, in ways unknown, results in the death of neurons and in the traumatic, drifting loss of mind and self that is the human experience of the disease.

It’s now clear that the amyloid cascade hypothesis in its simplest form is incomplete. The immune, vascular, and lipid systems in the brain are also involved, and we are only beginning to understand how. Crucially for the development of treatments, we don’t know for certain which pathological events must be prevented to stop brain cell death and cognitive decline in a way that is meaningful for people with the disease. That we don’t know this and yet press on with costly clinical trials is testament to the formidable complexity of dementia, the suffering it causes, and the vast amounts of money to be made by those who might alleviate it.


There have been two anti-amyloid drugs that have received some form of FDA approval for treating Alzheimer’s disease in the past two and a half years, and one that is being considered for approval. The drug aducanumab was approved in June 2021. As the first drug that attempts to target a cause of Alzheimer’s rather than its symptoms, this was historic. It was also flagrantly controversial, an instance of what one might call drug development as farce. The FDA made this momentous decision against the unanimous recommendation of its scientific advisory committee not to approve and in a manner incongruent with its own statisticians, who concluded that there was no evidence that the drug was effective. Uproar followed, advisory committee members resigned, and more than one respected hospital said it would not be offering the treatment. In the coup de grâce, the government then declined to cover the treatment except as part of a clinical trial. This produced the absurd situation in which one US regulatory body approved aducanumab to treat people with early Alzheimer’s while another essentially vetoed it. The European counterpart of the FDA, the European Medicines Agency (EMA), has not approved the drug.

Part of the problem with aducanumab was procedural. The companies that had developed it, Biogen and Eisai, initially halted their clinical trials as futile, then later announced that a new analysis of the results gave them grounds to apply to the FDA for provisional approval.1 Even odder, as a recent congressional inquiry into FDA mishandling of the decision has also reported, the reanalysis was done in unconventionally close collaboration with scientists at the FDA itself. Naturally, none of this built confidence in the drug. But this was just the first layer of the problem.

Aducanumab and other drugs like it are antibodies: molecules derived from the immune system that can be designed to bind tightly to a target of interest but not to anything else in the body, so that the immune system then rids the body of the targeted molecule. Aducanumab binds to amyloid plaque. The question is no longer whether such antibodies can clear amyloid from the brain—intravenous infusions of aducanumab and other such drugs do this very well. What we don’t have very good evidence for is that this produces a meaningful cognitive improvement in people afflicted with the disease.

The two trials on which the FDA based its decision did test this, but the results were middling at best. One trial showed a slight slowing of cognitive decline over eighteen months in people with early Alzheimer’s, buying them about five months of mental function compared with those who got a placebo. The other trial showed no effect at all. So, in approving aducanumab even provisionally, the FDA gave its imprimatur to a drug that effectively removes amyloid from the brains of people with Alzheimer’s disease but that may or may not improve their lives.

There is nothing inherently wrong with using a surrogate for the true outcome of interest in a drug trial, in this case an amyloid brain scan as a proxy for the mental function of a person with Alzheimer’s. Such a proxy may even be practically necessary in trials for a disease that unfolds over years, as Alzheimer’s does. But for a proxy to be useful it must be an excellent predictor of the outcome that matters, as high cholesterol levels are of cardiovascular disease, or blood HIV levels of AIDS. The FDA explicitly based its approval of aducanumab on a judgment that amyloid plaque removal is a good proxy for stopping cognitive decline. Its decision may in the long run be seen as a precedent for the approval of other Alzheimer’s drugs based on their effects on amyloid rather than on cognitive function. I hope that the strong flare of disagreement that ensued will slow down this sort of thinking.

The furor has since been subsumed to some extent by a later FDA approval of another anti-amyloid Alzheimer’s drug. This was less theatrical, in part because the drug in question, lecanemab, more unequivocally slowed cognitive decline in trials on people with early Alzheimer’s. Lecanemab now has full FDA approval, will be covered by Medicare under certain strict terms, and is under consideration by the EMA. Another such drug, donanemab, is under consideration for full approval by the FDA, with a decision expected in the next few months.


These developments have caused much excitement in the Alzheimer’s field, beleaguered as it has been by the repeated failure of other potential treatments in clinical trials. But the unfortunate fact is that the effects of even these newest drugs on mental function are painfully small. People who received treatment declined, over eighteen months, to the level that those in the placebo group did over ten to fifteen months, depending on the particular trial and set of patients. In other words, lecanemab and donanemab by no means halt the disease. They may simply buy people a sliver of time.

The question then becomes, tragically, whether a few months of stable cognitive function is worth some $26,500 a year and possible side effects. All of these drugs come with the risk of brain bleeding or swelling, in most cases minor and resolving without symptoms, but in rare cases serious and potentially fatal. The drugs also produce a mysterious loss of brain volume on MRI scans, an undesirable phenomenon for which neither the cause nor the long-term consequences are known. So, despite these trials being hailed as resounding successes, a purely anti-amyloid route to curing this disease does not look very promising.

How Not to Study a Disease was first published in late 2021 and does not comment on any of the FDA approvals directly. But Herrup makes a pointed critique of the amyloid cascade hypothesis and so disputes the very basis of drugs like these. The momentum of Alzheimer’s research in the 1980s and 1990s, he says, “had gathered so much force that the ideas in the amyloid cascade hypothesis were nearly unstoppable—so much so that more and more people stopped listening to their own data.” He acknowledges just how convincing the evidence for amyloid involvement in Alzheimer’s disease was, and still is. But he points out crucial inconsistencies in the story.

The first troublesome matter is that it isn’t only people with Alzheimer’s who have amyloid plaques in their brains. A hefty fraction of older people with normally functioning minds have them too. While a person with lots of amyloid plaques is more likely to develop Alzheimer’s than a person with few or none of them, the disease typically doesn’t develop very quickly; it could take many years, and it could also not happen at all. One can conclude from this that there must be something besides the accumulation of amyloid, or some combination of things, that sets some people along the path to Alzheimer’s. We still don’t really know what this is.

The second incongruity Herrup mentions is that many clinical trials attempting to reduce or remove amyloid clumps from the brain have not slowed the decline in cognition in Alzheimer’s patients. This sounds like an equally damning critique, but it’s trickier. There have been many reasons for trial failure, and not all of them are equally informative. Some trials were halted because of bad side effects, others didn’t actually get rid of brain amyloid very well, and in at least one trial the patients on a placebo didn’t decline either, making it impossible to test the drug’s effects. Of the two dozen or so completed phase 3 clinical trials (large trials that enroll many hundreds of patients), by my count we can be confident for only three of them that brain amyloid was strongly reduced.2 These were the trials of the now approved drugs and of donanemab.

In support of Herrup’s argument, the effects on cognition have been small. It’s possible that with trials as short as eighteen months we are seeing only the beginning of an effect that will grow over time. This should become apparent as the thousands of people who have by now participated in trials with strong amyloid-reducing drugs are observed over longer periods.

But a very plausible reason for the lack of meaningful cognitive effects is that the timing of treatment really matters. In familial, early-onset Alzheimer’s, plaque accumulates for a startlingly long time, over two or even three decades, before the person mentally slips. By the time these mental problems manifest, it’s very likely that there have been irreversible changes in the brain that can no longer be affected by getting rid of amyloid. So even in familial Alzheimer’s, where the evidence for a causal role of amyloid is strong, targeting it to stop the disease once a person is afflicted may be like banning matchboxes in a forest already burned down.

Herrup appears to take a dim view of invoking timing to explain failures in anti-amyloid treatment. He is of the view that, in particular in the common late-onset form of the disease, amyloid is simply not very important, and that other drug targets must be found. Much of the evidence for amyloid in Alzheimer’s, he points out, has come from studying the rare early-onset form. Though that’s true, I think a simple and equally important point is that, even for early-onset Alzheimer’s, one might have to lower amyloid for years or decades before a person is even sick to have a fighting chance at changing the course of the disease. Whatever the exact role of amyloid in late-onset disease, this timing is unlikely to be shorter or easier to predict. This is reason enough to seek alternative or additional therapies.

Lifelong lower levels of amyloid are protective against Alzheimer’s. We know this not from a clinical trial but because of a genetic mutation identified in the Icelandic population, which fortuitously reduces the amount of amyloid fragments by 40 percent and confers a reduced risk of disease on people who carry the mutation compared with the general population. But people are born with this mutation, so it’s the equivalent of taking an amyloid-reducing drug from birth. If this is the sort of timescale that a truly effective anti-amyloid treatment must replicate, we would be stuck: even clinical trials to test this would be practically and ethically unjustified. The more workable scenario is one where a preventative treatment is begun before a person has irreversible damage but when it is very likely that they will progress to the disease. What this therapeutic sweet spot is and whether it can be identified are as yet unknown.

For those with the misfortune of carrying an Alzheimer’s mutation, meaning they are very likely to get early-onset disease, or even for those who simply carry strong risk genes, a decades-long preventative treatment may be acceptable, even desired, and at the very least a source of precious hope. And if one’s goal is profit, the prospect of long-term preventative treatment against a feared disease doubtless has some charm. Accordingly, a crop of clinical trials for preventative anti-amyloid treatments are now in progress in healthy but genetically at-risk people. Many of these trials will begin reporting results in the next few years.

One is tempted to be hopeful, but there was bleak news in 2022 from the first such large preventative trial, a heroic effort involving a Colombian extended family in which members carrying an Alzheimer’s mutation start showing signs of disease in their mid-forties. Treatment with the anti-amyloid drug crenezumab was begun about seven years before this age, well before symptoms were expected to manifest, and was then continued for five to eight years. But the trial failed: the treated people began to decline cognitively at roughly the expected age and, despite some faint positive signs, there was no difference from those who received the placebo. There are possible reasons for this failure, among them that crenezumab is probably not very good at getting rid of brain amyloid, so the newer drugs being tested now may do better at prevention. Still, it raises the worry that even many years of preventative treatment may not be enough.

Herrup argues that Alzheimer’s research has lost its way, having become enamored of the amyloid hypothesis to the exclusion of other good ideas. He attributes this to three events over the course of its history. The first was in the early twentieth century, when the psychiatrist Emil Kraepelin, Alois Alzheimer’s boss and an arbiter of received opinion in the form of his textbook Psychiatrie, elevated Alzheimer’s observations on the unusual postmortem brain structures in a few patients from mere case studies to a newly defined, textbook-worthy disease. Neither Kraepelin nor Alzheimer could have known that the unusual structures they’d seen were the cause of the disease, but Kraepelin pushed this hypothesis forward.

Herrup holds that this thinking has been hard to shake off. But one good reason we haven’t abandoned it is that human genetics, some of the strongest evidence available, has insistently implicated the amyloid precursor protein in early-onset disease. Even though amyloid is far from the whole story, and even though targeting it alone almost certainly won’t give us a cure, Kraepelin’s and Alzheimer’s instincts weren’t all that bad.

The second event Herrup cites is the expansion of the definition of Alzheimer’s disease in the 1970s from the relatively rare presenile dementia it had referred to until then to senile dementia more broadly, which strikes a good two or three decades later but afflicts vastly more people. In a fascinatingly cynical interpretation, Herrup puts this down to a simple jostling for resources, as the leaders of the newly formed National Institute on Aging sought to secure the scale of funding that would accrue if one could claim to be tackling a widespread disease rather than simply studying aging. By rebranding senile dementia as Alzheimer’s disease, Herrup argues, the leaders of the National Institute on Aging had their enemy, they could wage their war, and they would be funded to do it.

While this is certainly plausible, another way to think about it is that our attempts to classify dementia have been cyclical over the past century. Familial early-onset and sporadic late-onset forms of Alzheimer’s roughly map onto the presenile and senile dementia of a hundred years ago, and theories about whether these are the same or different diseases have had either lumpers or splitters in the ascendant at any given time. Today, as molecular understanding deepens, the picture is gaining nuance.

Early- and late-onset Alzheimer’s disease are considered similar enough that learning about one is informative about the other. But clinical trials are run separately, and there are now attempts to model the two forms of the disease separately in laboratory mice. The human genes involved in the two forms are also different: the genes that put a person at risk for late-onset Alzheimer’s are not as directly linked to amyloid. But it is certainly not unexpected that there could be many routes, some indirect, to the pathological changes that lead to Alzheimer’s. Much effort is now being expended on working out these indirect routes to the disease.

To confound things further, classifying the pathological patterns in the brains of people with dementia is a messy business. Postmortem brain tissue of old people with cognitive problems often bears evidence of other conditions, not just the tangles and plaques that define Alzheimer’s disease. This is hardly surprising, but it makes the task of understanding and treating dementia that much harder.

The third event Herrup argues has derailed Alzheimer’s research is the definition of a presymptomatic phase of the disease, in which people have amyloid plaque in their brains but no disease. This was done, pace Herrup, expressly by proponents of the amyloid hypothesis to explain the presence of amyloid plaque in people with normal mental function. I share his unease at labeling someone with no symptoms as incipiently sick based on an imperfect surrogate like amyloid, which for late-onset disease is more of a risk factor than a deterministic one. But in fairness it is also reasonable, in a slow-burning disease like Alzheimer’s, to try to identify at-risk people at early stages, when they may be more amenable to treatment.

Although amyloid is still intensely studied, it has certainly not been lost on researchers that deeper biological understanding and alternative targets are needed if Alzheimer’s is to be meaningfully treated, and Herrup does not quite acknowledge this shift. His book strikes me as being fueled by many years of sailing against the prevailing winds of the amyloid hypothesis. This is a courageous position, and it’s self-evident that people who speak against consensus are needed. Herrup also makes sensible proposals for what should be done instead of orienting around amyloid. He argues that we ought to pay more attention to (and better fund) the biological study of aging, age being the single largest risk factor for Alzheimer’s disease, and that the key to understanding neurodegeneration is to think of it as a disease “of the neighborhood,” meaning one that involves not just neurons but also their interactions with the many other cell types in the brain. But the book pays only fleeting attention to more recent developments, mainly in the study of the brain’s immune and lipid systems, stating simply that change is too slow in coming. A reader may therefore not fully appreciate that Alzheimer’s research is becoming, even if too slowly, the broader church that Herrup says it should be.

As of about a year and a half ago there were well over a hundred drugs being tested against Alzheimer’s disease; less than a fifth of these target amyloid. Most of the rest target immune processes like inflammation, aim to protect the contact points between neurons, or are designed to interfere with tangles of tau. This last is an attractive drug target because, unlike amyloid, tau tangles appear roughly concomitantly with the earliest cognitive symptoms and in the neurons that actually die. Drug development efforts targeting tau have nonetheless lagged, in part because of the early focus on amyloid that Herrup points to but also because tau is a complex protein that comes in many forms, with the relevant toxic ones still murky. Early trials targeting tau have been discouraging, but many others are in progress.3

Herrup is quite correct that even for the variables we know about—tau, amyloid, apolipoprotein E (the major risk factor for sporadic Alzheimer’s disease), brain immune or vascular cells—the field is still working out how, molecularly, dysfunction contributes to the death of neurons. There is still therefore an element of guesswork in designing Alzheimer’s drugs. In contrast, the newest classes of cancer drugs are typically designed to interfere with the precise molecular pathways that drive the uncontrolled growth of tumor cells. It will take some time and effort before we get to that point with neurodegeneration.

More optimistically, it is possible to reduce the risk of dementia even without understanding all the biological details. Several facets of general health—high blood pressure, smoking, obesity, lack of physical exercise, diabetes, loss of hearing, excessive alcohol consumption, depression, low education level, low social contact, air pollution, and traumatic brain injury—have been observed to be risk factors for developing dementia.

And while these links are observational, there is some interventional evidence too. The 1,200-person, two-year FINGER trial (Finnish Geriatric Intervention to Prevent Cognitive Impairment and Disability) reported in 2015 that an intensive combined intervention in diet, exercise, cognitive training, social contact, and active monitoring of vascular risk factors improved cognitive function compared with standard health advice. The people recruited for this study were between sixty and seventy-seven years old and with normal or very slightly impaired cognition, so what this says is simply that lifestyle can improve mental function in the elderly. But several ongoing trials should soon give more direct information about whether lifestyle interventions can also meaningfully slow the onset of dementia.

Not all such factors are easily modifiable in real life; try telling a recent economic migrant to polluted Mumbai, or a lifelong elderly resident, that they need to move to the countryside for the sake of their brain. But many of them are modifiable by individuals even outside of an intensive program, and most are amenable to collective change in the longer term. This would simultaneously benefit general health and well-being, does no harm, and at least for factors amenable to individual choice, does not cost very much compared with developing new drugs.

Neurodegeneration is a very hard problem, probably harder than clogged arteries. Although there’s no reason to think it’s an insoluble one, there will almost certainly not be a single silver pill that gives most of us a nimble mind well into the tenth decade of life. Rather, there will be small advances and combined strategies that will have to be painstakingly worked out. Optimism, I’ve been told, is a defining trait of centenarians with clear minds. I’d like very much to be optimistic about our ability to cure Alzheimer’s in the next decades; for this to happen we must continue to broaden and deepen our inquiry into exactly what happens to brain cells in the disease, adjudicate clinical trials by what they mean for actual human lives, and consider that a drug is not always the only option.