A methodological critique of the National Institute of Aging and Alzheimer’s Association Guidelines for Alzheimer’s disease,dementia,and mild cognitive impairments

Introduction

In 2011, the U.S. National Institute on Aging, in collaboration with the (United States) Alzheimer’s Association published new guidelines on Alzheimer’s disease (AD) (Jack et al., 2011). In doing so, the guidelines redefined the research agenda for AD. This paradigm change has received very little discussion. Although “These recommendations are solely intended for research purposes and do not have any clinical implications at this time” (Sperling et al., 2011, p. 280), the clinical implications are hard to dispel (Mattsson, Brax, & Zetterberg, 2010). This paper aims at contributing to an open discussion about the guidelines’ methodological validity before such clinical implications are adopted.

While the previous 1984 guidelines only recognized one stage—Alzheimer’s dementia—the new guidelines propose that AD progresses on a continuum with three stages (Jack et al., 2011). The first is an early, pre-clinical stage with no symptoms (Sperling et al., 2011), followed by a middle stage of mild cognitive impairment (MCI) (Albert et al., 2011), and a final stage of AD dementia (McKhann et al., 2011). An additional guideline updates the criteria for documenting and reporting Alzheimer’s-related changes observed at autopsy (Hyman et al., 2012).

The stated purpose for these new guidelines is to enable researchers to incorporate new neurological insights and technological advances, specifically in the area of biomarkers and techniques for the detection of neuronal damage. Some of the techniques that have been identified for AD biomarkers are brain-imaging studies using magnetic resonance imaging (MRI), positron emission tomography (PET), and proteins in cerebrospinal fluid (CSF). The advances in these techniques include not only hardware and software but also new methodologies for imaging such as with the functional MRI (fMRI), and for interpreting data as from volumetric MRI (vMRI).

By allowing these advances in science to be incorporated within a general framework of AD, these guidelines channel efforts to elucidate the early neurological basis of the disease. The ambition for the guidelines was to act as a consolidating force. Despite the intention of the guidelines to summarize a wealth of existing data, allowing the development of new hypotheses and predictions, and to establish a forum for sharing of ideas, it is not clear that these guidelines are more inclusive than their 1984 predecessor.

The new guidelines argue that when the neurological damage—as identified by the biomarkers—becomes severe, the disease turns first into MCI and then, with increasing severity of neuropathology, into AD. MCI refers to memory lapses that do not affect daily activities, while AD is the expression of memory loss and cognitive capacity that affect engaging in day-to-day activities. What was since 1984 a probability of association is now seen as a causal progression. Because finding a cure for AD has been illusive—mirroring the enigmatic behavior of cancer—there is a palpable yearning for such a linear system of causality. Except for one shortfall. At every stage in this proposed causal links, there is no definitive evidence that one stage causes the other. What the literature shows is that there is an increased probability that being in one stage increases the likelihood of experiencing subsequent stages. However, probability does not determine causality and there is no one study that reports causality in any stage of AD.

The guidelines assert to have been established to “improve current diagnosis, strengthen autopsy reporting of Alzheimer’s brain changes, and establish a research agenda for future progress in earlier detection and even greater diagnostic accuracy” (Alzheimer’s Association, 2014, p. 1). This critique similarly supports establishing a valid and comprehensive “… research agenda for future progress,” but one based on firmer science.

Methodological shortfalls

The guidelines both implicitly and explicitly suggest a causal pathway for AD. It is important to note that the refutation—rather than the acceptance—of a hypothesis is the basis of the scientific method (Popper, 2004).

One methodological design to test for causality would be to follow a randomly assigned group of young adults—one group with and one group without the biomarkers—and then see who develops AD. The guidelines do not report of such studies (Sperling et al., 2011). The assumption is that since AD is a neuropathological disease, biomarkers that identify early stages of this neuropathology can therefore act as proxies for the early stages of AD. Sperling et al. (2011) identify that there are empirical limitations in making these linear assumptions about the disease. The authors argue that it is important to note several potential confounding issues in the majority of pre-clinical dementia studies. One of these issues relate to the design being susceptible to “cohort biases” due to the convenience sampling methodology in all of these pre-clinical studies. It is very rare that people who have no memory or cognitive issues are ever studied. There might also be people who have the neuropathology but never become “symptomatic during their lifetime” (Sperling et al., 2011, p. 2).

Theoretical criticisms

Overall the guidelines are prone to theoretical isolation. Studying AD in isolation to other neurological diseases or psychosocial processes assumes that AD is a neurologically specific disease. Studying the same or different biomarkers in other neurological diseases can contribute to identifying the uniqueness of AD, while including psychosocial studies will help explain some of the research anomalies that exist. Gaining a better understanding of how the brain accommodates neuropathology without any diminished cognitive affect should be a key feature of AD research which needs to be incorporated in future guidelines. Theoretical isolation is part of the biological determinism that is implicit in the old guidelines. The notion that specific neurological degradation will solely result in dementia might be logical but still untested. In contradiction to this biological determinism, numerous studies continue showing the importance of social intervention as mediating and moderating factors.

Threats to construct validity

There is variability in measuring the biomarkers, MCI, and AD. This variability poses a threat to the construct validity. If what we are measuring is not stable, then it is more difficult to establish the relationship between two moving targets. The lack of reliability obscures the relationship being studied. There are numerous biomarkers competing for a role in AD formation. A coherent approach to identifying each of these biomarkers and how they could possibly interact among themselves is an important omission in the guidelines that erode the validity of these biomarkers as constructs. In addition, the lack of reliability in measuring MCI diminishes the soundness of using generic MCI as an indicator. One example for future guidelines is to focus on aMCI as defined by Petersen et al. (1999, 2001).

The guidelines fail to define the constructs primarily because we are still uncertain about what we are trying to measure. Is AD the neuropathology as defined by the biomarkers, or, as is assumed, the expression of the disease through loss of memory and change in capacity and personality? Therefore, there is an instrumentation issue in defining what AD is. Whether it is the neuropathology or whether it is the expression of the disease. Although AD was named because of the characteristic neuropathological progression associated with severe cognitive decline, the guidelines use any type of dementia as AD. In one example, Esiri (2010) indicates the likely confusion that might arise with respect to AD and AIDS-related dementia and also AD and other dementing illnesses. There are numerous types of dementia. AD is the most common form of dementia to be found in over 60% of cases followed by vascular dementia, dementia with Lewy bodies, fronto-temporal dementia, Korsakoff syndrome, Creutzfeldt–Jakob disease, and HIV-related cognitive impairment. The rare forms which occur in 5% of cases relate to corticobasal degeneration: Huntington’s disease, multiple sclerosis, Niemann–Pick disease type C, normal pressure hydrocephalus, Parkinson’s disease, posterior cortical atrophy, and progressive supranuclear palsy. Different types of dementia might have different and very specific causes. Focusing on a specific type of dementia will limit the variability in the dependent variable (Fratiglioni, De Ronchi, & Agüero-Torres, 1999). Up the chain of causality is the instrumentation issue with MCI. There is great variability in defining MCI, allowing researchers to select types of MCI that are more likely to be pre-dementia and then using that as the predictor, which makes for a tautological argument (Fratiglioni et al., 1999).

Since AD is mainly diagnosed behaviorally—and confirmed histologically—there is still the individual physician’s reliability in diagnosing AD. In a clinical setting, a physician’s ability to assess patients may improve over time. Some physicians are more competent at diagnosis than others, while some type of patients might be more likely to receive one diagnosis or severity of diagnosis then another patient (Christakis, 1999). Instrumentation issues in AD research are capricious and require new guidelines to provide some practical suggestions.

There are also threats to internal validity in establishing causality. Because these guidelines have sensitized people to associate MCI directly with dementia, patients might present themselves much later for diagnosis when AD is more likely to be a comorbidity. There will be a change in prevalence due to history as it relates to sensationalizing of MCI and AD (Metlife, 2011). There are also the ethical implications of early diagnosis (Mattsson et al., 2010) and legal implications of a diagnosis of AD on the individual, needs to be further considered in the literature.

Especially with much older older-adults, the confluence of morbidities means that not only is there selective sample attrition due to mortality, but there is increased confounding of variables. In cases of dementia in the oldest-old, when compared to younger people, it is more likely to be related to mixed disease pathologies (Brumback-Peltz et al., 2011).

Studies that are cited in support of the use of biomarkers for AD are based on people who initially present themselves as candidates for AD (Sperling et al., 2011). This methodology is not only based on convenience sample but is also self-selecting. Such non-representative sampling precludes generalization to a wider audience. Having mostly White patient base poses a non-representative research threat since minorities are unrepresentative in AD research and when they are included, show different outcomes (Tang et al., 1998). In the real world, a heavier toll and at earlier ages falls on minority populations (Clark et al., 2005; Haan et al., 2003; Valle et al., 2013). A homogenously White population explains the guidelines’ assertion to connect MCI with AD since with ethnically diverse groups MCI was found to be more unreliable (Low et al., 2012).

Discussion

There is ample evidence now showing that there are mediating and moderating factors in the expression of the disease. AD is mediated by education through a process Snowdon identifies as “…cognitive reserve—the capacity of the brain to resist the expression of symptoms in the face of existing neuropathology” (Snowdon, 2003, p. 453). These studies support the hypothesis that brain plasticity offers a possible mechanism through which the brain might be induced to repair itself (Gage, 2003). Overwhelming evidence exists that cognitive exercises have a positive influence on the neurology of the brain. Cognitive training programs have also been associated with delays in the onset of dementia and age-related cognitive diseases. Why the guidelines chose to exclude such bourgeoning studies is unclear. What is relevant is to identify the importance of social factors in future guidelines and in any model of the etiology of AD. Non-cognitive variables such as SES, education, literacy, social class—might all have a mediating and/or moderating role in the etiology as well as in the expression of the disease.

In addition, there is evidence suggesting that some people do not conform to the statistical norm in terms of neurological shrinkage. It seems some people escape the correlates of aging. Evert, Lawler, Bogan, and Perls (2003) support this view when they examined death from heart disease, non-skin cancer, and stroke; 87% of male and 83% of female centenarians that they studied delayed or escaped these diseases. Similarly when Harrison, Weintraub, Mesulam, and Rogalski (2012) looked at two groups of older adults 80 years and older compared to 50 to 65 year olds, these exceptional older adults had a thicker outer layer of the brain important for memory, attention, and thicker anterior cingulate—which is responsible for attention. Exceptional people are the Popperian nemesis of these guidelines (den Dunnen et al., 2008).

If we apply the concept of escapers to the brain, then we can say that exceptional older adults escape from damaging their brain, and if they have neuropathology, they have enough reserve to cope with the day-to-day cognitive demands.

Social impact

These guidelines usher in a new era of fatalism. “Because no treatment is available for the condition…the prognosis significance of these tests eclipses their therapeutic utility” (Christakis, 1999, p. 12). The effect of these guidelines is to instill a lonely fear of declining cognitive ability: “On the one hand, the future course of an illness is actually for the patient to experience, and the experience indeed often takes place outside of the professional domain—removed from the doctor spatially, temporally, and physiologically” (Christakis, 1999, p. 63). It is a lonely hopeless journey. Our task as scientists is to define a methodology that allow for an exploration of anomalies in order to fully understand the disease and provide preventing, mitigating, or moderating strategies. Unintended consequences of these new guidelines are stoking the fear of dementia, without hope of any forthcoming remediation.

Conclusion

The inductive method of observing the disease and then looking for biomarkers is a false syllogism (Fuller, 2011). Just because everyone with dementia might have neuropathology, we cannot assume that the neuropathology determined the dementia. The theoretical postulation that there is one linear causal pathway from a biomarker causing MCI which then matures into a dementing illness is not supported by a mounting body of evidence. These anomalies are pointing to the necessity of expanding the concept of biomarkers and showing that these biomarkers might be moderated or mediated by other, likely psychosocial, factors. Moderating variables are amiss in the these guidelines. Rather, any future guidelines need to include such variables as education, comorbidities, brain plasticity and neurogenesis, diet, exercise, and meditation (Ornish et al., 2008) and not to bypass culture and ethnicity (Clark et al., 2005; Haan et al., 2003; Valle et al., 2013). It is also amiss that there is no discussion of how epigenetics might modulate the genetic predisposition to AD. Are biomarkers the neuropathology of the disease, and is the “disease” the expression of dementia or the neuropathology? Dubois et al. (2010) recommend changing the lexicon “to consider AD solely as a clinical and symptomatic entity that encompasses both predementia and dementia phases” (Dubois et al., 2010, p. 1118) leading to potential situations where despite having normal memory and behavior you might still get diagnosed with dementia.

There is also a need to identify the type of MCI that leads to AD. The relationship between so-called Alzheimer’s changes and normal aging processes should be distinguished. One of the distinguishing features is the speed of attrition. Even though we still do not know enough about normal aging processes, sudden attrition might be indicative more of disease pathology. A longitudinal perspective is crucial for making such distinction. There are emerging tests that show promise in differentiating memory abilities affected by cognitive impairment from those affected by aging (Brainerd et al., 2013).

These are all important theoretical consideration that need to be part of an established “research agenda for future progress.” The only way to define a future agenda is to treat anomalies in the data not as errors but as shedding light on hidden and yet unknown dynamics. The NIA guidelines do not establish a research agenda for future progress. Instead the guidelines represent a dying proposition of biological determinism that exclude social and environmental factors as reflected in the emerging science of epigenetics and neuroplasticity.