The pathophysiology of Alzheimer's disease

The underlying pathophysiology of Alzheimer's disease is important in understanding why particular biomarkers have been developed and is depicted in Fig. 1. Alzheimer's disease involves abnormal functioning of amyloid precursor protein, which leads to an excess of amyloid-β in the cortex. This excess of amyloid-β is thought to lead to faulty accumulation of tau proteins, which in turn leads to synaptic dysfunction and neuronal death (Jack Reference Jack, Knopman and Jagust2010). The pathological hallmarks of Alzheimer's disease are amyloid-β plaques and neurofibrillary tangles of hyperphosphorylated tau. Compared to amyloid-β plaque formation, hyperphosphorylated tau and neurofibrillary tangles are more closely correlated to neurodegeneration and clinical symptoms (Bennett Reference Bennett, Schneider and Wilson2004).

FIG 1

A hypothetical model of the pathophysiological cascade in Alzheimer's disease (from Sperling Reference Sperling, Aisen and Beckett2011).

The difference between genetic profiles, other risk factors for dementia and cognitive reserve may explain the variation in lag time between the development of Alzheimer's pathology and dementia itself. Cognitive reserve is given as one explanation of why there is such individual variability in the time taken between developing Alzheimer's pathology and clinical symptoms (Fig. 2). The greater your cognitive reserve, the more insult you can endure before displaying cognitive symptoms. Cognitive reserve is an active process which can be strengthened by, for example, education and mental activity (Stern Reference Stern2012).

FIG 2

Cognitive reserve in ageing and Alzheimer's disease (AD) (Stern Reference Stern2012). With high cognitive reserve, more Alzheimer's disease pathology can be tolerated before symptoms develop.

Biomarkers

The main biomarkers of Alzheimer's disease can be divided by two major pathological processes: amyloid-β deposition and neurodegeneration. Hence, main biomarkers focus on:

• • measures of brain amyloid-β deposition

• • measures of markers of neurodegeneration.

The two main methods of detecting amyloid-β deposition are by measuring amyloid-β levels in the CSF and through position emission tomography (PET) amyloid imaging (Berti Reference Berti, Polio and Lombardi2016). Amyloid-β₄₂ is the most likely type of amyloid to aggregate and is the most commonly measured amyloid variant. Where there is significant amyloid-β accumulation, levels of amyloid-β₄₂ in the cerebrospinal fluid (CSF) are low and there is increased amyloid tracer retention on PET imaging (Sperling Reference Sperling, Aisen and Beckett2011).

Plasma amyloid can also be detected, but plasma amyloid-β₄₂ has hitherto been used less frequently than CSF owing to its lower sensitivity and specificity (Anoop Reference Anoop, Singh and Jacob2010). Recently, however, there have been further advances in finding a suitable method to detect biomarkers in the blood (Nakamura Reference Nakamura, Kaneko and Villemagne2018). Acquiring blood, unlike CSF, requires a much less invasive procedure and, unlike PET scanning, is less costly. The researchers, by using immunoprecipitation and mass spectrometry, were able to achieve roughly 90% accuracy (using ¹¹C-labelled Pittsburgh Compound B ([¹¹C]PiB) PET as their gold standard) in the detection of amyloid-β biomarkers. They concluded: ‘These results demonstrate the potential clinical utility of plasma biomarkers in predicting brain amyloid-β burden at an individual level. These plasma biomarkers also have cost-benefit and scalability advantages over current techniques, potentially enabling broader clinical access and efficient population screening’ (Nakamura Reference Nakamura, Kaneko and Villemagne2018).

Neurodegeneration is measured by various indicators: increased concentrations of CSF total tau (t-tau) and phosphorylated tau (p-tau), hypermetabolism on fluorodeoxyglucose (FDG) PET imaging and atrophy in structural magnetic resonance imaging (MRI): t-tau is a more direct marker of neuronal degeneration and p-tau is a marker of neurofibrillary tangles (Dubois Reference Dubois and Albert2014).

Both CSF amyloid-β₄₂ and tau protein have been found to reflect the degree of amyloid load and neurofibrillary abnormality accurately at autopsy (Tapiola Reference Tapiola, Alafuzoff and Herukka2009). Similarly, amyloid imaging strongly correlates with the pathological burden of disease at autopsy (Backsai Reference Bacskai, Frosch and Freeman2007) and with concentrations of amyloid-β in CSF (Fagan Reference Fagan, Mintum and Mach2006).

A significant number of cognitively healthy older people will have evidence of amyloid-β deposition both at autopsy and in biomarkers of CSF and PET amyloid imaging. The number of individuals who are biomarker ‘amyloid positive’ but ‘clinically negative’ varies from 20 to 40% (Sperling Reference Sperling, Aisen and Beckett2011), which is similar to autopsy findings (Morris Reference Morris, Storandt and McKeel1996). One theory is that if ‘amyloid-positive’ individuals lived longer they would eventually develop symptoms of Alzheimer's disease.

Jack et al (Reference Jack, Knopman and Jagust2010) have proposed a model to represent the progression of Alzheimer's pathology (Fig. 3). In this model, amyloid-β deposition occurs first, years before the onset of clinical symptoms. The duration of this phase may vary, depending on the individual's cognitive reserve and risk factors for Alzheimer's disease. Next, tau-mediated neurodegeneration begins, which is evident from changes on structural imaging. Finally, there is progression to cognitive impairment and clinical symptoms become evident. Important to the biomarker model is that Aβ accumulation alone is not sufficient to cause dementia (Jack Reference Jack, Knopman and Jagust2010).

FIG 3

Accumulation of markers of disease over time (Jack Reference Jack, Knopman and Jagust2010). In this hypothetical model, amyloid-β markers are the first to become abnormal, followed by markers of neurodegeneration, followed by clinical symptoms. MCI, mild cognitive impairment.

Ocular biomarkers are less well-known biomarkers of Alzheimer's disease. Ocular biomarkers have been proposed for several reasons. First, individuals with Alzheimer's disease often present with visual deterioration, and this appears to pre-date cognitive changes. Second, the retina is easily visualised and ocular biomarkers do not require invasive tests such as lumbar puncture. There are many proposed ocular biomarkers, which we shall not discuss exhaustively here. For instance, micro-RNA can be found in tear fluid and is implicated in the regulation of amyloid. Other important approaches include detecting retinal amyloid-β accumulation, or an assessment of functional and clinical changes within the visual system (Lim Reference Lim, Li and He2016). Amyloid-β detection in the retina has been performed with or without contrast enhanced imaging. Further work is required to determine whether amyloid accumulation detected in the retina is reflective of brain deposition and, indeed, predictive of future cognitive impairment. Functional changes to the visual system include changes in neuronal responses, such as reduced visually evoked potentials over the occipital cortex and interrupted neurotransmission between photoreceptors in the retina. Further ocular biomarkers include ‘fixation and movement errors’, where there is failure to fixate or follow a target and reduced eye movements (Lim Reference Lim, Li and He2016).

Do CSF biomarkers accurately predict who will develop Alzheimer's dementia?

Using biomarkers to help predict who will develop dementia already has a significant history, particularly in determining which individuals with MCI are most at risk of developing Alzheimer's disease. In this population, there is evidence that biomarkers enhance diagnostic specificity and prognostication (Hardy Reference Hardy and Selkoe2002; Klein Reference Klein, Stine and Teplow2004; Hansson Reference Hansson, Zetterberg and Buchhave2006; Parnetti Reference Parnetti, Lanari and Silvestrelli2006; Mattsson Reference Mattsson, Zetterberg and Hansson2009; Visser Reference Visser, Verhey and Knol2009; Chetelat Reference Chetelat, Villemagne and Pike2011).

The combination of high CSF t-tau and p-tau with low CSF amyloid-β₄₂ has been termed the ‘Alzheimer's disease signature’ (De Meyer Reference De Meyer, Shapiro and Vanderstichele2010) and is highly predictive of Alzheimer's dementia, as confirmed in three large multicentre studies: the Alzheimer's Disease Neuroimaging Initiative (ADNI) study, the Development of Screening Guidelines and Criteria for Predementia Alzheimer's Disease (DESCRIPA) study and the Swedish Brain Power (SBP) project. The CSF Alzheimer's disease signature has been shown to increase diagnostic accuracy significantly even at a prodromal stage, with a sensitivity of 90–95% and a specificity of about 90% for Alzheimer's disease (Snider Reference Snider, Fagan and Roe2009; De Souza Reference De Souza, Lamari and Belliard2011).

De Meyer e t al (Reference De Meyer, Shapiro and Vanderstichele2010) studied participants from the ADNI. The ADNI was established in 2004 to determine whether biomarkers could predict progression from MCI to Alzheimer's disease. Their sample included cognitively normal individuals, individuals with MCI and individuals with Alzheimer's disease. The Alzheimer's disease signature was detected in 90% of individuals with Alzheimer's disease, 72% of those with MCI and 36% of cognitively healthy individuals. In individuals with MCI and an Alzheimer's disease signature, the diagnostic sensitivity of progression to Alzheimer's disease was 90%, with a specificity of 64%. In addition, the combination simply of high CSF p-tau and low CSF amyloid-β₄₂ correctly identified 100% of individuals with MCI who progressed to Alzheimer's disease.

Hansson et al (Reference Hansson, Zetterberg and Buchhave2006) found that the Alzheimer's disease signature had a sensitivity of >90% and specificity of >85% for individuals with MCI who subsequently developed Alzheimer's disease. Similarly, Van Rossum et al (Reference Van Rossum, Vos and Handels2010) found that the combination of CSF amyloid-β₄₂ with either p-tau or t-tau was highly predictive of the development of Alzheimer's disease from MCI, with an odds ratio of 18.1 (95% CI 9.6–32.4).

Most of this evidence relates to individuals who already have MCI. If pathological changes really do start decades before the onset of clinical symptoms, is it possible to tell at this very early stage who will develop Alzheimer's disease in their later years? There have not been enough longitudinal studies to answer this conclusively, although some studies have found that amyloid-β positivity confers an increased risk of progression to Alzheimer's disease (Fagan Reference Fagan, Mintum and Mach2007; Li Reference Li, Sokal and Quinn2007; Villemagne Reference Villemagne, Pike and Darby2008; Morris Reference Morris, Roe and Grant2009; Storandt Reference Storandt, Mintun and Head2009; Resnick Reference Resnick, Sojkova and Zhou2010; Chetelat Reference Chetelat, Villemagne and Pike2011), and two (Vemuri Reference Vemuri, Wiste and Weigand2009; Yaffe Reference Yaffe, Weston and Graff-Radford2011) found that plasma biomarkers and imaging (MRI and FDG PET) can predict cognitively normal individuals who are at risk of cognitive decline. Fagan et al (Reference Fagan, Roe and Xiong2007) found that non-cognitively impaired individuals with the Alzheimer's disease signature progress to symptomatic cognitive impairment more quickly than do the remainder of the cohort.

Are CSF biomarkers useful in people who already have Alzheimer's disease?

A marked reduction in CSF amyloid-β₄₂ has consistently been noted in patients at different stages of Alzheimer's disease. On the one hand, an isolated low amyloid-β₄₂ is not sufficiently specific for a diagnosis. Non-Alzheimer's disease dementias, such as Lewy body disease and vascular dementia, are also associated with low CSF amyloid-β₄₂ (Dubois Reference Dubois, Feldman and Jacova2014). CSF p-tau appears to be the most specific CSF biomarker, distinguishing Alzheimer's disease from non-Alzheimer's dementias (Koopman Reference Koopman, Le Bastard and Martin2009). On the other hand, combinations of CSF biomarkers appear to enhance the sensitivity of an Alzheimer's disease diagnosis, but there is no consensus as to which specific combination has the greatest accuracy (Dubois Reference Dubois, Feldman and Jacova2014). CSF biomarkers may also have a role in predicting the course of Alzheimer's disease. Snider et al (2009) found that individuals with Alzheimer's disease and the CSF Alzheimer's disease signature progressed more rapidly than those without. Markers of amyloid load tend to remain constant throughout the course of Alzheimer's disease, whereas t-tau and p-tau rise as Alzheimer's disease progresses (De Meyer Reference De Meyer, Shapiro and Vanderstichele2010).

Neuroimaging

Studies investigating the value of FDG PET (Box 2) in diagnosing prodromal Alzheimer's disease are few and have short follow-up periods. Metabolic reductions in the medial, temporal and parietal cortices, as well as the anterior and posterior cingulate, detected prodromal Alzheimer's disease with an accuracy of 75–84% (Dubois Reference Dubois, Feldman and Jacova2007). Differences between MCI and Alzheimer's disease have been consistently found on FDG PET (Shivamurthy Reference Shivamurthy, Tahari and Marcus2015).

There are PET techniques that provide in-vivo detection of amyloid and, potentially, of neurofibrillary tangles. Imaging using [¹¹C]PiB (N-methyl-[¹¹C]2-(4'-methyl aminophenyl)-6-hydroxybenzothiazole) and [¹⁸F]FDDNP (2-(1-{6-[(2-[¹⁸F]fluoroethyl](methyl)amino]-2-naphthyl}ethylidene)malononitrile) have detected enhanced radio-ligand retention patterns in patients with Alzheimer's disease compared with controls; and positive [¹¹C]PiB patterns are correlated with low CSF amyloid-β₄₂ (Dubois Reference Dubois, Feldman and Jacova2007). There is evidence of enhanced [¹¹C]PiB retention in cognitively normal people and in those with MCI (Dubois Reference Dubois, Feldman and Jacova2007). Longitudinal follow-up is required to determine whether these are people with a preclinical or prodromal dementia.

Diagnostic criteria

In 1984, a workgroup from the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association (NINCDS–ADRDA) (McKhann Reference McKhann, Drachman and Folstein1984) published criteria for the diagnosis of Alzheimer's disease. These criteria allowed only a ‘probable’ diagnosis of Alzheimer’s disease to be reached while the person was alive: a definitive diagnosis could be made only if Alzheimer’s pathology was found at autopsy. Furthermore, the (probable) diagnosis required the presence of significant disability and impact on daily living, thus excluding MCI (which was not recognised as such at the time). The specificity of these diagnostic criteria, in terms of distinguishing Alzheimer's disease from other types of dementia, was low (Dubois Reference Dubois, Feldman and Jacova2014).

In 2011, the National Institute on Aging and the Alzheimer's Association (NIA–AA) published guidelines (Jack Reference Jack, Albert and Knopman2011) revising the 1984 diagnostic criteria. All-cause dementia continues to be diagnosed by history and neuropsychological testing, with impairment in function required for a diagnosis to be made. This is then further subdivided into ‘probable’ or ‘possible’ Alzheimer's disease, and biomarkers may be used as supportive information to help guide a diagnosis and to help distinguish Alzheimer's disease from other forms of dementia. There is also a shift away from concentrating only on memory problems as the clinical phenotype of Alzheimer's disease towards also including impairments of other cognitive domains, such as language (Dubois Reference Dubois, Feldman and Jacova2014).

The NIA–AA guidelines of 2011 also recognise the spectrum of Alzheimer's disease, such as preclinical Alzheimer's disease and MCI, and the use of CSF and imaging biomarkers in identifying these states. Dubois & Albert (Reference Dubois and Albert2004) describe prodromal MCI as a heterogeneous state that may or may not lead to dementia of multiple different aetiologies. They describe prodromal Alzheimer's disease as ‘MCI of the Alzheimer type’ and advocate the use of CSF and imaging biomarkers to help make this distinction. For preclinical Alzheimer's disease, however, the NIA–AA guidelines only apply in a research setting (Watkin Reference Watkin, Sikdar and Majurndar2013). Research criteria for diagnosing preclinical states of Alzheimer's disease developed by the International Working Group (the IGW-2 criteria) require the individual to be asymptomatic, have a marker of Alzheimer's disease pathology or an Alzheimer's disease autosomal dominant mutation on chromosome 1, 14 or 21 (Dubois Reference Dubois, Feldman and Jacova2014). This is describing an at-risk state where progression to Alzheimer's disease is not inevitable.

Limitations in the use of biomarkers

Biomarker results are often taken from highly selected populations, so that the predictability of the results is likely to be exaggerated (Livingston Reference Livingston, Sommerlad and Orgeta2017). A major limitation is that participants in these studies are generally volunteers and one wonders who volunteers for invasive trials such as these. They may be individuals with concerns about their cognitive performance or with a family history of dementia. This is supported by the higher rates of apolipoprotein E (ApoE) ε4 carriers detected in many of these cohorts (Sperling Reference Sperling, Aisen and Beckett2011). This would lead to an overly high prediction of biomarkers in the general population.

Lumbar punctures are not without risk and there is no standardised analytic technique or reference range for CSF biomarkers (Livingston Reference Livingston, Sommerlad and Orgeta2017). CSF biomarkers show 20–30% inter-laboratory and inter-assay variability (Berti Reference Berti, Polio and Lombardi2016). The Alzheimer's Association has funded the Alzheimer's Association QC [Quality Control] Program for CSF Biomarkers to try to reduce this variability by using reference laboratories to compare samples (Mattsson Reference Mattsson, Andreasson and Persson2013).

Ethical implications

With this new concept of preclinical dementia come numerous ethical implications. Finding out that a cognitively healthy individual has Alzheimer's pathology leaves the researcher with a difficult question regarding an unknown risk that the participant will develop an incredibly debilitating disease. With such a significant number (20–40%) of cognitively normal individuals with evidence of amyloid-β plaques, more information is needed on the sensitivity of such a test, so that the implication of this result can be communicated to participants. Currently, establishing preclinical dementia has no clinical utility, and an unknown prognosis therefore provides unsettling and uncertain information for the patient (Harkins Reference Harkins, Sankar and Sperling2015).

Hughes et al (Reference Hughes, Ingram and Jarvis2017) carried out a scoping review to explore some of the issues that arise following detection of preclinical dementia and identified four main themes: stigma, ethical questions, psychological burden and language/terminology.

First, individuals with preclinical dementia may face stigma, including difficulties at work and insurance implications. In addition, social stigma can lead to isolation and cause the individual to form negative self-perceptions (Hughes Reference Hughes, Ingram and Jarvis2017).

Second, the idea of non-maleficence has recurred in the literature. Harm may be done by telling someone about biomarker positivity, particularly if, as is currently the case, there are no effective treatments for Alzheimer's disease. However, the Risk Evaluation and Education for Alzheimer's Disease (REVEAL) project reported that participants were no more likely to develop depression, anxiety or distress on being told their APoE genotype in addition to general risk factors for Alzheimer's disease (such as family history, age and gender) than those who were told the general risk factors alone (Green Reference Green, Roberts and Cupples2009). In terms of beneficence, individuals with a positive amyloid profile can alter their lifestyle choices to reduce their risk of dementia in the future or have time to make advance care planning decisions.

Third, there are psychological burdens associated with these studies, including anxiety about the test results. Psychoeducation to help participants and their families understand and reflect on the results may help.

The final theme in the literature is the importance of language. How things are communicated will shape how the person understands the meaning of results that may be ambiguous. Indeed, language itself can cause harm. A question can be raised, given the level of uncertainty, about whether the label of ‘preclinical Alzheimer's disease’ should ever be used, except perhaps in the context of research. The situation described by ‘preclinical Alzheimer's disease’ can also be called ‘non-dementia Alzheimer's disease’. It could be said that those affected are not ill, but have a disease. This is not uncommon in medicine: for example, many people may have cancer but no symptoms. It is more problematic when people have the putative pathology, but never go on to have symptoms. So, for example, many normal older people will have amyloids in their brains and it seems odd on this account alone to describe them as diseased. The worry here is a broad one about the medicalisation of ageing (Estes Reference Estes and Binney1989).