Despite enormous efforts to develop disease-modifying drugs for Alzheimer's disease (AD), effective therapies have yet to be found. We have endured failure for a number of reasons, such as low specificity in drug candidates, inaccurate diagnosis and incorrect timing in the intervention. To help overcome these problems, modern research findings have been incorporated into new criteria and guidelines for the clinical diagnosis of AD. In addition, attempts to intervene during the earliest stages of the disease are planned, as this is probably when AD is most receptive to disease-modifying therapies. We discuss these issues and provide perspective into the future of drug development.
Alzheimer's disease (AD) is the most prevalent neurodegenerative disease among the elderly population. The latest figures estimate that up to 2 million AD dementia patients reside in Japan, with 30 million patients living worldwide. With the graying of society in both developed and developing countries, the number of AD dementia patients in the entire world is estimated to reach over 100 million by 2050, the impact of which will be economically and socially disastrous. The disease is characterized by the slow progressive loss of memory and a decline in executive functions, leading to impairment in the quality of life of patients. There are three burdens that AD dementia brings to society: (i) the first is a loss in the working population, as a significant number of patients lose their jobs as a result of the disease; (ii) the second is the cost for caregiving, losing the working population from otherwise different industries; and (iii) the third, and sometimes overlooked burden, is the care for the caregivers themselves, as they are frequently depressed as a result of their heavy caregiving and worried about their future. There are some symptomatic treatments available for AD dementia, such as choline esterase inhibitors or N-methyl-D-aspartate (NMDA) antagonist, but they do not affect the underlying disease etiology or halt the progression of symptoms. Thus, even if patients receive the ideal therapy available today, the effect would not be long lasting and the patient's condition would eventually return to basal levels after a certain period of time.[2, 3] Therefore, there is a huge demand for the development of disease-modifying drugs for AD, which if possible, will change the course of the disease and reduce its associated burdens. Disease-modifying therapy has a direct impact on the biology of the disease, thus changing the clinical course and delaying symptomatic progression (Fig. 1). Various estimates show that if we could delay the onset of AD dementia by only a few years, we could reduce the economic burden by tens of billions of dollars. However, although enormous effort has been expended to develop disease-modifying drugs for AD, no tested candidates have survived the evaluation of phase 3 trials. Potential reasons for these piled-up failures are the focus of the present review. Furthermore, we address possible solutions and future perspectives for the field of AD disease-modifying drug development.
Amyloid β as a key molecule in AD pathogenesis
Since the incorporation of molecular biological methodologies into the field of AD research, huge advances have been made regarding its pathophysiology. We now understand that the sequential cleavage of amyloid precursor protein (APP) by two different enzymes, beta- and gamma-secretases, produces toxic amyloid β (Aβ) peptides, especially Aβ1-42, that tend to aggregate and stick together, resulting in its abnormal accumulation in the brain. Toxic Aβ oligomers impair synaptic function, while long exposure to them promotes neuronal tau phosphorylation and accumulation, causing neuronal dysfunction and ultimately neuronal cell death. Most cases of AD are sporadic, whereas autosomal dominant familial AD, caused by mutations in PSEN1, PSEN2 or APP genes, comprises a tiny fraction of all cases. These “familial” cases have almost indistinguishable symptoms and pathological findings to those of sporadic cases, except that the disease onset is earlier and the distribution of senile plaques is slightly different. Interestingly, mutations found in these genes are usually associated with increased toxic Aβ production, which highlights the considerable role Aβ plays in sporadic cases as well.[9-11]
Failure of anti-Aβ therapy: How and why?
In 1999, Schenk et al. reported that Aβ vaccination successfully removed amyloid plaques and resulted in functional improvement in a mouse model of AD.[12, 13] This result had a huge impact on the research community, and a clinical trial was subsequently initiated; however, frequent adverse effects, such as encephalitis, forced the study to a halt, and disappointingly, the immunization had no effect on cognitive function decline. Indeed, so many AD dementia clinical trials have failed without extracting any beneficial effects. Researchers are disappointed by the results, while pharmaceutical companies are becoming increasingly discouraged by the notion of further drug trials given the huge cost involved.
Thus, it is important to investigate the reasons for these failures and to find sufficient solutions. One possible reason for failure is that targeting Aβ might be the entirely wrong approach. Indeed, Aβ accumulation might have nothing to do with sporadic AD pathogenesis, or at best, merely the result of an unknown upstream event. Nonetheless, the results from human genetic studies discussed earlier in the present review suggest that Aβ has at least some role in the disease pathogenesis. Furthermore, a recent study has even suggested a protective effect of low Aβ production against AD. From these reasons, we believe that targeting Aβ has at least some rationale. A second possibility is that a significant number of non-Alzheimer's disease dementia cases might have been misdiagnosed as AD dementia, and thus have contaminated the past trials. A third possibility is that we are targeting the disease at the incorrect time-point. The latter two possibilities are discussed in further detail.
Does misdiagnosis matter?
Unlike symptomatic drugs, disease-modifying drugs for AD are designed to target specific molecular pathways that underlie the entire disease process, which are thus destined to be ineffective for non-AD dementia patients. Thus, accurate diagnosis before enrolment in clinical trials is required in order to extract the maximum efficacy of disease-modifying drugs. As the final diagnosis of AD is based on neuropathological examination, there is always a possibility of clinical misdiagnosis. In most past clinical trials, National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA) criteria have been used to clinically define AD dementia. We have been using these criteria for almost 30 years, yet our knowledge has been quite dramatically refined in that time; thus, the criteria are obsolete. When the criteria were first developed, AD dementia was considered as dementia with AD pathology, whereas normal was considered cognitively normal without any AD pathology. We now know that there are other causes of dementia, such as frontotemporal lobar degeneration (FTLD) and Lewy body dementia, which are given little consideration in the criteria. We also know that AD pathology can exist in cognitively normal brains. NINCDS-ADRDA criteria have a policy of excluding other causes of dementia by negative findings, such as normal computed tomography scans or normal cerebrospinal fluid (CSF). Furthermore, positron emission tomography (PET) scans and magnetic resonance imaging (MRI) findings are not even mentioned. We now know that AD is associated with significant morphological changes, especially in the medial temporal region, and with Aβ1-42 and phosphorylated tau abnormalities in CSF, as well as reduced glucose metabolism in various areas of the brain. Thus, for an active diagnose of AD dementia, there has long been a demand to incorporate such vital research findings. The new criteria were developed with this in mind, and will hopefully lead to more accurate diagnoses of AD dementia[18, 19] (Fig. 2).
Right target but wrong timing?
As discussed earlier, we have at least some pieces of evidence that targeting Aβ could be beneficial in AD. Then why did clinical trials targeting Aβ fail? Unlike dementia as a result of stroke or head injury, it is impossible to identify the exact date of AD dementia onset. This is because the onset of AD dementia is insidious and the disease progresses very slowly. This is one of the characteristics of neurodegenerative diseases. For example, in Parkinson's disease, it is known that the appearance of Lewy bodies precedes symptomatic onset, and that the patient will be asymptomatic until there is more than 50% cell loss in the substantia nigra. These phenomena are possibly a result of neuronal plasticity that enables compensatory mechanisms to mask circuit dysfunctions and cell losses in neurodegeneration.
Incidentally, amyloid plaques are usually found at autopsy in individuals aged over 40 years, with their appearance increasing with age. This is approximately 10–20 years before the symptomatic onset of dementia.[21-23] From this result, it can easily be speculated that there are a significant number of people with amyloid plaques without any cognitive decline. This is confirmed by the discovery of Pittsburgh compound B-positive individuals among the cognitive normal population; that is, they have detectable amounts of fibrillar amyloid, but no cognitive decline.
The major symptom of AD is cognitive impairment. When patients are diagnosed with dementia, the deterioration is so advanced that they can no longer live independently; this is in fact the definition of dementia. Neuropathological changes are also quite prominent at this stage. Accumulation of amyloid plaques and neurofibrillary tangles are almost maximally developed when a diagnosis of dementia is made (Fig. 3). So, what if we could intervene in the disease process at an earlier stage? If so, when is it?
Mild cognitive impairment (MCI), especially of the amnestic type, is diagnosed when there is evidence of memory impairment, preserved general cognitive and functional abilities, and an absence of dementia. When an individual is diagnosed with MCI, they have ~50% chance of progressing (conversion) to dementia within 3 years; however, there are a certain number of patients whose symptoms are quite stable who will not show progression to dementia. This is because MCI is a syndrome that consists of various pathological backgrounds including AD, and MCI as a result of non-neurodegenerative disorder will not progress to dementia, and MCI as a result of non-AD neurodegeneration will have a different clinical course. Thus, if we could extract MCI caused by AD, it might be worthwhile planning an intervention therapy at this stage. To achieve this, we need to know which MCI patients are more likely to progress to dementia. Various studies, including cognitive tests, CSF, structural MRI, amyloid PET, 18F-fluoro-deoxy-glucose positron emission tomography (FDG-PET) or a combination of those have been carried out to showed their usefulness as predictors of conversion to AD dementia. By using these pieces of information, we might be able to treat AD at the MCI stage.
Another issue is related to clinical trial design for disease-modifying therapy. A recent trial on spinal and bulbar muscular atrophy showed that it is extremely difficult to prove efficacy of the therapy, even if it acts right on the target and efficacy has been well demonstrated in mouse models. These experiences highlight the issues of disease-modifying therapies in humans, and the details are discussed by the authors.
Importance of prospective cohorts and biomarker studies
Although our knowledge regarding AD is accumulating, it has mostly been obtained from cross-sectional studies. We have still not identified the triggering step in the disease course. Furthermore, longitudinal studies of patients are required with careful observations recorded in order to assess their disease progression in its entirety.
Biomarkers are measurements of biological states, such as blood pressure, blood glucose levels and QT intervals on electrocardiograms. Those measurements are highly reproducible and help evaluate or predict one's disease. What are the biomarkers for AD and how useful are they? At present, there are four major biomarkers that represent AD pathology. They are as follows:
- Aβ pathology, as detected in the CSF or by amyloid imaging PET.
- Neurodegeneration and dysfunction, as detected by CSF tau, phosphorylated tau, and FDG-PET imaging.
- Neuronal loss, as measured by MRI.
- Memory loss and cognitive decline, as measured by cognitive assessment batteries.
Recent studies have shown that the first three markers are useful for diagnosing AD, as they are abnormal even before the diagnosis of dementia.[36, 37] However, there are no confirmed biomarkers that represent the actual disease progression other than cognitive assessment. In other words, even if there were effective disease-modifying drugs, the outcome must be measured by cognitive assessment. In present clinical trials for AD dementia, the primary end-point is change in Alzheimer's disease assessment scale-cognitive subscale (ADAS-cog), which is a cognitive battery designed for mild-to-moderate dementia. As cognitive measurement is affected by the patient's physical condition, and the reproducibility of results is relatively low compared with biomarkers from imaging and CSF measurements, the numbers of participants required in phase 3 clinical trials would be high enough to achieve statistical significance, usually more than 500 participants per arm, thus increasing the cost of the study. So, what if there was a biomarker that had a high correlation with cognitive function, was highly reproducible and could represent the severity of AD? To achieve this, there are several issues that must be solved first. They are as follows:
- We need to standardize various cognitive batteries, MRI and PET data acquisition protocols, and CSF biomarker measurement protocols in order to eliminate laboratory-to-laboratory, site-to-site variations.
- We need to establish the time-course change of those biomarkers.
- We need to establish a method to identify individuals with AD pathology before they reach the symptomatic stage, preferably at the earliest stage possible.
By clearing these issues, we might be able to develop biomarkers that represent the disease progression, in other words, a surrogate marker for AD.
There are currently prospective multicenter cohorts in operation. One of them is the Alzheimer's disease neuroimaging initiative (ADNI; http://adni.loni.ucla.edu). The first phase of ADNI in North America dealt with biomarkers related to MCI to AD dementia conversion, and they are now at their second phase of study (ADNI2) to detect biomarkers related to the early phase of MCI. Similar studies have been carried out in the European Union (EU-ADNI; http://www.centroalzheimer.org/), Australia (AIBL; http://www.aibl.csiro.au/) and Japan (J-ADNI; http://www.j-adni.org/), and those protocols are designed to be partially compatible so that they can provide a worldwide platform for the next generation of clinical trials. The result from the ADNI study showed accelerated hippocampal atrophy rates in AD compared with MCI subjects. We believe that the most important part of these prospective cohorts is data sharing and project-to-project collaboration, as AD treatment is a worldwide issue, requiring the whole intelligence we humans have.
Is there any option to go earlier?
As discussed above, considering the natural course of AD, setting the disease onset at the time of dementia is too late. Then when should the onset of the disease be considered? Is MCI the beginning of AD? Or is there an earlier time-point for the disease onset? To settle this issue, two criteria for the early diagnosis of AD have been proposed[23, 39] (Fig. 4). According to these criteria, the appearance of amyloid plaques defines the disease onset. This can be detected by either amyloid PET or a low CSF Aβ1-42 profile. The patient undergoes an asymptomatic preclinical phase, possibly for more than 10 years, and neurofibrillary tangle accumulation marks the beginning of the second stage of preclinical AD. When there is a very subtle change in cognitive function, the patient enters the third stage of preclinical AD. Cognitive decline progresses until they are below 1–1.5 SD for their age and educational standards, after which they could possibly be diagnosed with MCI. According to previous understandings, MCI was thought to be a risk factor for AD dementia. However, the new criteria define MCI due to AD as a relatively late stage in AD, whereas dementia is defined as the end-stage of the entire disease process. This is quite a drastic way of redefining the disease course, but it is required for cutting-edge research purposes.
It is quite important to note that both criteria do not recommend or indeed oppose the incorporation of these preclinical stages into regular clinical practice, as there is a huge problem regarding ethical issues. We still do not have effective intervention against AD. Under these circumstances, a diagnosis of AD has a very serious impact on one's life. Thus, we need to be extremely careful about the concept surrounding this issue, and a meaningful public debate is required in order to obtain general consensus.
Attempts to carry out intervention at preclinical stages have been planned in the USA. The Alzheimer's Prevention Initiative (API) is designed for a large PSEN1 kindred in Columbia; the Dominantly Inherited Alzheimer's Network Trials Unit (DIAN-TU) is for PSEN1, 2 and APP familial cases in USA, Europe and Australia; and the Anti-amyloid in asymptomatic AD (A4) trial is for sporadic cases of amyloid-positive cognitively normal cases in the USA. In Japan, the J-ADNI2 study is planned as an observational study to include amyloid-positive and cognitive normal individuals.
Limitations in AD drug development
Aβ is like a small piece of garbage that neurons produce along with their activity. Accumulation of Aβ requires a certain amount of time, thus the lifespan of an individual is important for their accumulation. Animals, such as rodents, have a short lifespan that is not long enough for Aβ accumulation, thus there are no rodents that develop AD in natural conditions. Canines might develop dementia in their second decade of life, but they do not develop neurofibrillary tangles in their brains, probably because neuronal exposure to toxic Aβ is not long enough. Animal models of AD that we are currently using for drug development are mostly transgenic mice that produce very high amounts of Aβ and tau. Preclinical development using these animals has its own problems, such as: (i) the animals are generated under the hypothesis that Aβ and tau play the central role in the disease pathogenesis, which is likely but not yet proven; and (ii) the effect of drugs are measured by cognitive batteries developed for rodents, such as successfully navigating through mazes and finding safe spots, which could be different from human cognitive function. Thus, we need to be very cautious when we interpret the results from preclinical experiments.
We would like to propose some future perspectives for AD drug development in this last section.
- Dementia is the end-stage phenotype of AD, which begins with the accumulation of Aβ in the 40s or 50s; therefore, intervention targeted towards Aβ should be initiated as early as possible, even in patients without symptoms. We still need to determine at which stage of the disease intervention is most efficacious. If drugs, such as statins, that which have very few side-effects, could be developed, we might recommend taking those drugs after some screening tests for Aβ accumulation in cognitively normal individuals.
- Surrogate biomarker development is required. Objective biomarkers, such as MRI shrinkage rate or CSF biomarkers, would be useful to determine if the intervention is effective or not. The cost of clinical studies using cognitive batteries is enormous, but if we could incorporate those biomarkers as an end-point for clinical studies, it would help reduce the cost of running clinical studies. As a result, we could test more drugs.
- Tau might be the next target for drug development, as it can be effective even after the development of dementia; also, there are other neurodegenerative diseases that possess tau pathology, but do not show Aβ accumulation.
We reviewed and discussed the current achievements and problems in the field of AD disease-modifying drugs. We hope these issues can soon be solved and that the successful development of drugs for this debilitating disease will soon be achieved.
- 1World Alzheimer Report 2010. Alzheimer's Disease International, 2010; 1–54. Available from: http://www.alz.co.uk/research/files/WorldAlzheimerReport2010ExecutiveSummary.pdf (accessed 19 March 2012)..
- 8High striatal amyloid β-peptide deposition across different autosomal Alzheimer disease mutation types. Arch. Neurol. 2009; 66: 1537–44., , et al.
- 18The diagnosis of dementia due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 263–9., , et al.
- 20Ageing and Parkinson's disease: substantia nigra regional selectivity. Brain 1991; 114 (Pt 5): 2283–301., .
- 22Alzheimer's disease facts and figures. Alzheimers Dement 2009; 5: 234–70..
- 23Toward defining the preclinical stages of Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 280–92., , et al.