Posted 23 July 2010

The Meaning of Alzheimer’s Disease Across Time and Place: From the Bedside to the Desktop

Plenary Lecture delivered at the International conference on Alzheimer’s Disease, 2010, Honolulu, Hawaii, by Jason Karlawish, University of Pennsylvania, Philadelphia

Click on the image below to view PowerPoint slides

Today I want to talk about what we mean when we say “Alzheimer’s Disease.” I am especially keen to convey to you how this meaning is changing. We are entering the 4th Age of Alzheimer’s Disease. I want to summarize this new age and to examine its implications for physicians, researchers, patients, their families and society.

(Slide 3) Why do we care about Alzheimer’s disease? We care about it because it causes disability, a loss of productivity not only in the persons who have the disease but the persons who care for them. We care about Alzheimer’s disease because people don’t just die with it but from it. We care about it because when we add up each of these case histories, we add up substantial costs to families and society. We can say these things because we have an understanding of Alzheimer’s as a disease that is caused by pathology in a person and this pathology presents with symptoms and signs. The symptoms and signs include memory loss; losses in other cognitive abilities; changes in mood, behavior and motivation; and, perhaps most importantly, losses in the ability to perform usual and everyday activities such as managing money, medications, and traveling outside the house.

This meaning of Alzheimer’s—this understanding of the disease—is similar to many other diseases that physicians diagnose using clinico-pathological correlation. This term describes how a physician approaches a patient who presents with a chief complaint. The physician obtains a history and then a physical, orders tests and then fits or correlates these data to select which of our known diseases best explains the patient’s chief complaint.

This is the model of bedside medicine. It is a powerful model that over the last century and a half has organized how medicine selects students, trains physicians, and sets up clinical practice. The chief complaint initiated the history and physical (H&P), which is the cornerstone of medical training and practice.

In the beginning, Alzheimer’s disease was a protean concept, loosely tied up with ideas of aging, mental illness, insanity, and sin and the suffering of the bitter fruits of a dissipated life. That was the First Age of Alzheimer Disease.

(Slide 5) The Second Age of Alzheimer’s disease began in the early 20th century, when Alois Alzheimer’s case report “On an unusual illness of the cerebral cortex” (Alzheimer, 1907) inaugurated the concept of Alzheimer’s disease as a disease in a person diagnosed using the H&P and clinical pathological correlation. The story began with a 51-year-old woman who was jealous of her husband. It ended with a post-mortem examination of the brain showing “only a tangle of fibrils indicates the place where a neuron was previously located.”

It was a classic of the genre of bedside medicine: a careful clinical history initiated with a chief complaint and concluding with anatomic examination of the brain.

(Slide 7) Some seventy years later, the Third Age of Alzheimer’s disease began, when Robert Katzman published his editorial in the Archives of Neurology: “The Prevalence and Malignancy of Alzheimer’s Disease: A Major Killer” (Katzman, 1976 http://www.alzforum.org/pap/annotation.asp?powID=81966).

In this 1,000-word essay, Katzman expanded the category of Alzheimer’s disease to include not simply persons who were middle aged, but, he argued using clinical pathological correlation, persons who were elderly with senile dementia. He wrote: “Both Alzheimer disease and senile dementia are progressive dementias … that are indistinguishable by careful clinical analyses. The pathological findings are identical.”

This shift in meaning, saying senile dementia is Alzheimer’s disease, had a substantial impact on the way we think about what is normal aging versus what is disease. By removing large quantities of older adults from the category of the senile to persons with a disease, it rapidly expanded the size and therefore the urgency of the problem of Alzheimer’s disease. It was now a major killer.

It is entirely sensible that countries with aging populations began to pay attention to Alzheimer’s disease. In the United States at the time of Katzman’s editorial, for example, the less than a decade old National Institute on Aging within the National Institutes of Health took on Alzheimer’s disease as a disease in need of research and established a national network of Alzheimer’s Disease Centers.

Their core was two cores—a clinical core to describe the patients and a neuropath core to describe their pathology. In short, these centers were grounded in the meaning of Alzheimer’s disease as a disease in a person diagnosed through careful clinical-pathological correlation.

Because Alzheimer’s disease was this (Slide 8)—a person with a history of chronic progressive and disabling declines in multiple cognitive abilities and a brain full of plaques and tangles.

(Slide 9) We are now leaving this Third Age of Alzheimer’s disease and entering The Fourth Age of Alzheimer’s disease—the disease as a risk factor.

We can date the dawn of this Fourth Age to the end of the 20th century when investigators at the Mayo Clinic in Rochester, Minnesota led by Ron Petersen published “Mild Cognitive Impairment: Clinical characterization and outcome” (Petersen et al., 1999).

Central to this paper was this figure (Slide 10)—not a figure of static pathology but of change over time, specifically a figure of risk. In this case, the risk of transforming from having a largely isolated memory disorder to having Alzheimer’s disease.

(Slide 11) The meaning of Alzheimer’s disease over time has changed. In this century it is changing again: from pathology in a person that is the cause of disability, to the risk of a future event that a person faces. The question that frames the remainder of my talk is what are the clinical and policy implications of this new meaning of Alzheimer’s disease?

After a brief summary of this new meaning, the approach I will take to answer this question will be to look at other risk-factor diseases. In particular, I will talk about diseases like hypertension, osteoporosis and dyslipemia. What do their discovery, development and models tell us about the future of our disease?

Central to the transformation of Alzheimer’s disease into a risk-factor diagnosis are recent advances in science and technology, specifically the rise of epidemiology, information and computing sciences, and fast and powerful networked computers that permit researchers to quantify the kinds of persons who, over time, are likely to develop a clinical event that is disabling and is costly, such as a stroke, heart attack or fracture of hip.

One of the greatest interests of researchers who use these sciences and technologies has been the discovery of biological measures that predict the risk of developing that clinical event. So-called “biomarkers.”

(Slide 12) You can see on this slide how this term is relatively recent, first appearing in the medical literature in 1990, but since then it has gained substantial use. In the case of Alzheimer’s disease, we see here (Slide 13) how the term biomarker has captured the attention of researcher just in this past ten years. This new meaning has captured considerable scientific and clinical attention.

(Slide 14) The 2007 publication of “Research criteria for the diagnosis of Alzheimer’s disease: revising the NINDS-ADRDA criteria” by Bruno Dubois and 18 colleagues illustrates this attention.

(Slide 15) A key feature of these criteria is not simply what is here, but what is not here. History is deemphasized. Dementia is not discussed. The clinical assessment is simply whether there has been memory decline. The focus is “objective evidence”—a measured score on a memory test and a biomarker of the disease.

Whether these criteria are right or wrong is not the issue. The issue is that by 2007, some 100 years after Alois [Ah - loyss] Alzheimer’s’ case report, 19 authors from six countries were writing that we can uncouple Alzheimer’s disease from dementia. It is to a degree a clinical diagnosis as it still requires report of memory loss, but central to the definition and its defense is the concept of risk. Specifically, among persons who meet this set of criteria, they are at risk of developing disability. Subsequent iterations of these criteria have suggested dropping the requirement for a symptom of memory loss and simply focus on a score on a cognitive test and a biomarker.

The model of Alzheimer’s disease based upon clinical-pathological correlation is beginning to be replaced with a model of Alzheimer’s disease based on the risk of future disability. We are moving from the bedside diagnosis of Alzheimer’s disease to the desktop diagnosis of Alzheimer’s disease.

I use the term “desktop” to describe how a desk with a networked computer equipped with software capable of longitudinal, multivariate data modeling is beginning to transform medical science and, in turn, medical practice. The desktop is the space where medical researchers discover risk-factor based diseases such as hypertension, osteoporosis, and dyslipemia and where physicians diagnose and treat patients with these diseases.

But the field of Alzheimer’s disease is not there yet. The histories of other diseases explain why the field is not there and how the field will get there and the clinical and policy challenges we are beginning to face.

(Slide 16) This slide summarizes core features of one of these diseases: dyslipemia or high cholesterol. In brief, it describes a disease of a level of cholesterol that puts a person at sufficient risk of a future heart attack that this level should be reduced.

The discovery of this desktop disease was made possible by a two step validation process. First, biological and epidemiological data showed that a factor, elevated LDL cholesterol, is associated with the risk of a negative health event, a heart attack. Second, a randomized and controlled trial showed that an intervention upon that risk-factor reduces the likelihood of that event.

The history of dyslipemia is illustrative of the discovery of desktop diseases. Merck selected as the first patient group to target with simvastatin persons with familial hypercholesterolemia, an autosomal dominant disease. Such persons not only had such high risk of coronary vascular disease that effects could be detected in small sample trials, but they had clinical signs—such as subcutaneous cholesterol plaques. Hence, these were persons who had a disease with both desktop and bedside features. It was a useful bridging diagnosis.

Success with an intervention in this patient population would allow Merck and its co-investigators to pursue studies of persons in lower risk groups, persons with elevated cholesterol but without familial hypercholesterolemia. Successive trials did just that. That is, as trials showed effects upon persons with a certain level of cholesterol, subsequent trials went after persons with even lower levels. The result is a kind of looping mechanism between the measure of the disease and the drugs that treat the disease. They shape and reshape each other.

Within a decade the multitude of terms describing different types of lipid disorders rapidly disappeared. All the various types of hyperlipidemia soon vanished. They were replaced with this (Slide 17)—the National Cholesterol Education Project’s “Risk assessment tool for estimating your 10-year risk of having a heart attack.”

It is a Web-accessed risk calculator. You enter your data into the various boxes—age, gender, total cholesterol and so on -- press this button, and get back an estimate of your ten year risk of a heart attack. Using this actuarial calculation, you decide whether to recommend a risk reduction intervention such as taking a statin drug. This exercise is called clinical actuarial correlation, to distinguish it from clinical-pathological correlation.

The point of this brief history is to illustrate how the clinical trial and drugs are essential tools not simply to develop a treatment for a disease but in the very discovery of the disease itself.

I cannot emphasize enough how critical are drugs in the very definition of desktop diseases such as hypertension, dyslipemia and osteoporosis, and someday, perhaps soon, Alzheimer’s disease.

Next, I will discuss some of the challenges presented by these kinds of desktop diseases.

Dyslipemia is one desktop disease. (Slide 18) This slide summarizes two others: hypertension and osteoporosis.

One feature to note is the discrete clinical event. This discrete clinical event is not only a clinically sensible concept, it is a very useful endpoint of a clinical trial to develop treatments. The result is a clear and coherent language of both clinical care and clinical research. A picture says a thousand words.

(Slide 19) This event, a fractured bone, is the discrete clinical event that is part of the definition of osteoporosis.

(Slide 20) Here is a slide of the WHO FRAX criteria that define one’s 10-year probability of a major osteoporotic fracture or a hip fracture.

The discrete clinical event that defines the meaning of Alzheimer’s disease in its 4th Age is not as clear as a bone fracture or heart attack or stroke. We know that the onset of Alzheimer’s disease is gradual and between experts there is substantial inter-rater variability in whether someone has Alzheimer’s disease. The lack of a discrete clinical event presents a substantial challenge to the Fourth Age of Alzheimer’s Disease.

Diabetes may be our model for how we talk about treatment. Diabetes has become a disease defined by an abnormal biomarker: hemoglobin a1c, or glycosylated hemoglobin.

This single number is a powerful tool to promote awareness and define a person’s risk.

(Slide 21) This is a bus kiosk in my home city of Philadelphia. It reads along the bottom “Know your risk. Know your A1C” and the picture is of an ambulance whose side panel reads “Hank - Diabetes complications are coming to get you” and here is Hank sitting in his living room and beside him is a woman.

Clinical trials of persons at risk for diabetes—such as persons with impaired glucose tolerance—are examining whether an intervention reduces the risk of developing diabetes as defined by an elevated HgA1c. The discrete clinical event is simply a level of A1C that is above normal.

The limitation of the diabetes model for Alzheimer’s disease is that we have not established a biomarker that has a threshold above which we say is disease and below which we say it is not disease. And we have not yet established a relationship between changes in a biomarker and changes in a clinically meaningful measure such as capacity to perform activities.

The Clinical Dementia Rating has been proposed as an endpoint. The one limitation of this measure is that it is not routinely used in clinical practice. Valid as it may be, it is opaque. The typical clinician does not use the CDR. They do use HgA1C, LDL, blood pressure, stroke, MI, bone fracture because they are well understood across clinical and clinical research communities. They are efficient and reimbursable tests.

Another challenge of desktop diseases is the shift in the control of who defines the disease and with this shift in control, a shift in power.

(Slide 22) This is a screen shot from the on-line FRAX website. Like the National Cholesterol Education Project risk calculator, the FRAX calculator is an online site to perform a risk assessment to determine whether a person has sufficient risk of fracture that an intervention is recommended. There are drop down menus to indicate what country your potential patient is in and then a series of 12 data fields to fill in—one of which is bone mineral density. The result is the 10-year probability of a fracture.

It should be apparent from this and the NCEP risk calculators that the central authority of the expert clinician in defining whether a patient has the disease is waning. Anyone can go to these websites. It should also be apparent that as clear as these websites are, what is behind them is relatively opaque. The creators of FRAX have been criticized for not making their formula public. The formula is privately owned, not public, and these owners are in negotiations to license the formula to two leading manufacturers of bone-scanning equipment that would allow them to incorporate it into their software.

What’s my point? Disease -- its definition and its treatment -- and the ownership of these things are more and more tightly enmeshed into the private sector.

In our field, Alzheimer’s disease, confidentiality agreements, for profit information brokers, and propriety data have created asymmetries of information that threaten our ability to select drugs in competitive head-to-head trials to validate biomarkers. We have an urgent need to renegotiate the public-private distinction.

A third challenge of desktop diseases is that they are about time. Time to a future event. But people—both patients and doctors—have a hard time thinking about time. We are very present biased in our decision making. We prefer to enjoy a tasty high fat snack and consequent weight gain and risk of diabetes, to a less pleasurable activity future oriented activity such as exercise to reduce the risk of diabetes.

Even when we are told what is our risk—the fact, the number—we often take away a different risk.

A recent paper out of the Risk Evaluation in Alzheimer’s Disease study group—the REVEAL study—led by Robert Green at Boston University gives us a vivid example of the problems of thinking about disease as risk. Cognitively normal first-degree relatives of persons with AD reported a very different risk of developing Alzheimer’s disease than the calculated estimate of their lifetime risk based on age and ApoE allele.

(Slide 23) The title of the paper sums up how people think about risk information: “’I know what you told me, but this is what I think:’ Perceived risk of Alzheimer disease among individuals who accurately recall their genetics-based risk estimate.”

It is easy to feel a symptom—pain hurts. It is harder to feel a risk. We have a host of biases, heuristics, and misrepresentations that lead us to a gist understanding of what we think is our risk and therefore what we choose to do or not to do. Within six months following a heart attack, nearly half of the once again asymptomatic patients started on a statin are no longer taking the drug and persons who do adhere are more likely to report absenteeism from work.

(Slide 24) I’ve discussed how drugs define disease. They also and obviously treat the disease. Another challenge in desktop medicine is deciding who to treat. Once a disease is defined using risk as its conceptual model, it moves from discrete category to a continuum, as risk is a continuum.

As a result, the borders between who should versus who should not be treated become porous. Why should someone who is just below the threshold of risk of persons treated in the trial, not have access to the treatment? Answering this question engages issues of what kinds of evidence should have what weight, what is cost-effective and our ethics.

The NCEP controversy over whether to treat women at high risk for heart disease illustrates this. In brief, NCEP’s treatment recommendations include the recommendation to treat women defined at moderately high risk of coronary vascular disease, although the available clinical trials did not show a benefit of statin treatment for these persons. NCEP was criticized for making this recommendation without adequate clinical trial evidence to support it.

NCEP’s response to this criticism spoke of a different way of thinking about who to treat. They reject privileging the results of randomized and controlled trials as the arbiter of who to treat. Instead, they argue for an approach that reviews the entire body of scientific evidence in formulating recommendations, including animal, pathologic, genetic and epidemiological studies, and clinical trials.

This approach obliterates the logic of conceptually separating primary and secondary prevention. NCEP wrote: “insisting on waiting until clinical trial evidence becomes available specifically in moderately high risk women before regarding them as eligible for cholesterol lowering would mean that many women would have a potentially preventable heart attack before they are accorded the benefits of therapy. For many women, the first sign of heart disease is sudden death. Sound public health policy demands that the significant risk for illness and death in women be addressed with science-based prevention recommendations.”

The message here is that decisions about whom to treat—framed as risk reduction—use not only data from clinical trials but other studies, such as epidemiology, to decide that a group is at sufficient risk to warrant intervention. It is evidence-based, but it is evidence based in a manner that relaxes the primacy of the gold standard of the clinical trial.

In this approach, the clinical trial does not functional as a kind of autonomous machine to tell us who to treat, instead human judgment does. This new paradigm increases the urgency to develop methods to govern who should participate in guideline writing and how the process should operate.

These methods should be as rigorous as our clinical trials. These methods must recognize that expertise in the disease is essential, so too is expertise in other desktop diseases, and in policy making. This new paradigm further reiterates the present problem of asymmetric information and the private control of data.

I have spoken about Alzheimer’s disease across time—its history and its incorporation into history as risk. I want to make a few brief remarks about Alzheimer’s Disease across place.

(Slide 25) This slide depicts the top ten countries for a troubling statistic in 2020—the dependency ratio. This ratio describes the numbers of persons who are over 65 to the number of persons who are between 15 and 64. In other words, it is the ratio of those who are not working, to those who are working thus paying taxes and producing goods, in part to help care for the elderly. Only one country—France—has a national plan. Three others—Belgium, Germany and Sweden—are working on plans. The rest are simply waiting as the numbers count up the burden of risk they face without a national plan to manage that risk.

(Slide 26) This is our future. I’ve talked about the challenges we will face.

(Slide 27) Key points/questions are:
What will be the discrete clinical event?
How can we break the logjam that proprietary control and asymmetric information create?

(Slide 28) Who will control the drugs and models that define disease?
How will we foster better risk perception and communication?
How will we foster appropriate and sustained risk reduction behaviors (e.g., retention in a RCT and adherence to an Rx)?
How will we develop guidelines that are fair and unbiased and use all the data?

The promise of the risk factor approach to medicine is that it will take us from the suffering patient at the bedside to the relatively healthy person at the desktop. It will reduce the costs of disability. But the histories of other risk factor diseases tell us that this approach has its own unique challenges as well.

Perhaps the most contentious debates in desktop diseases is over how much risk reduction is worth what costs and who should decide this?

One key cause of this discord is the prisoner’s dilemma created by proprietary control over drugs and data.

Another cause is over how much risk in what kind of person for how long is worth the costs of workup, monitoring and intervention?

In policy making, this question has a different emotional valence than it does when the disease is a bedside disease and the issue is the burden of suffering in present patients.

The cold numbers tallying the costs of risk-factor assessments and treatment over many years replace the heart wrenching stories of the suffering patients. These economic debates become part of the disease itself.

Let us be secure and unified in our resolve that we do not forget the patients who started us on this journey into the 4th Age of Alzheimer’s disease.

Thank you.