From Cognitive Science to Dementia Assessment

Biological and Cognitive Markers of AD

Biological markers or “biomarkers” for AD include analysis of a sample of fluid taken from the spine and analysis of brain scans. These biomarker tests have been shown to be sensitive and specific for AD, although they are not fully reliable and have not been wholly satisfactory in differentiating AD from other forms of dementia (e.g., Blennow et al., 2015; Frisoni et al., 2010; Johnson, Fox, Sperling, & Klunk, 2012; Swan et al., 2015). There is a substantial research effort directed toward refining the use of such biomarkers, which are viewed as the best forms of assessment available. As such, clinical policy views biomarkers as a form of “gold standard” for assessment of AD. However, these techniques can be invasive, painful (e.g., a needle in the spine), and/or expensive (brain imaging). Moreover, they are only available in specialist centers and hospitals with highly trained staff. This means that they are not widely available, and the cost of setting up and running the necessary facilities, coupled with the very large volume of older people who seek medical help for their memory problems, make current biomarkers an impractical solution for routine assessment. Hence, clinical policy and practice cannot rely on current biomarkers. Moreover, due to their still limited specificity, biomarkers are currently recommended as research tools only rather than for routine clinical use (R. Sperling & Johnson, 2013).

Assessments of Mental or Cognitive Functions

Typically, diagnosis and monitoring of progression of AD is assisted by the use of generic assessments of “cognitive” functions, namely, memory, learning, attention, perception, thinking, and language. These assessments are non-invasive, inexpensive, portable, and, unlike assessment of spinal fluid samples and brain scans, they are widely available in clinical settings and can be used in primary care with minimal waiting time for the testing or for receiving the results. The outcome of this clinical assessment is often used to refer the patient to a specialist or for further assessment, such as a brain scan when the facilities are available and affordable. Or they may assist in the diagnosis of conditions such as depression that can be treated, at least initially, in primary care. However, cognitive assessments in current clinical use can only detect that there is a general impairment in mental ability. They cannot identify the underlying cause of that impairment. Therefore, these assessments may lead to uncertain diagnosis or inappropriate clinical follow-up.

Tests of cognitive function also have a range of other limitations. The most frequently used forms of assessment are dementia severity scales, notably the Mini-Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975). This can be a useful tool to categorize patients according to how severe is their general cognitive impairment, but it has been known for some time that it is not reliable (e.g., Cossa, Della Sala, Musicco, Spinnler, & Ubezio, 1997), nor is it specific for AD (e.g., Boustani, Peterson, Hanson, Harris, & Lohr, 2003; Tariq, Tumosa, Chibnall, Perry, & Morley, 2006) and has poor sensitivity for detecting impairments in individuals with high levels of education (Vertesi et al., 2001). Other dementia severity scales such as the Milan Overall Dementia Assessment (Brazzelli, Capitani, Della Sala, Spinnler, & Zuffi, 1994), and the Addenbrooke’s Cognitive Examination (e.g., Hsieh, Schubert, Hoon, Mioshi, & Hodges, 2013), have been shown to be more reliable and sensitive. However, poor performance on these collections of short tests is not an indicator of any particular disease.

More extensive batteries of tests assessing cognitive functions are also available, notably Alzheimer’s Disease Assessment Scale–Cognitive subscale (ADAS-Cog) or Cambridge Neuropsychological Test Automated Battery (CANTAB). These are commonly used in aiding diagnosis as well as in clinical trials of pharmaceutical treatments for AD. ADAS-Cog (Rosen, Mohs, & Davis, 1984) is widely used in the United States and consists of 11 tests covering a range of cognitive functions, not just those affected specifically by AD. However, it has been shown to suffer from unreliability and a lack of sensitivity (e.g., Connor & Sabbagh, 2008; Karin et al., 2014). CANTAB (e.g., Sahakian & Owen, 1992) is one of the best-known test batteries for cognitive impairment in the United Kingdom. This also is a suite of tests to measure a wide range of cognitive functions that offers a profile of which cognitive abilities are impaired and which are relatively spared in individuals. It is continually under development with an ongoing research program, but its strength lies in detecting that impairments are present, not what is causing those impairments, and so it is not specific for AD. It involves extended time for testing each patient, and performance on the tests is affected by healthy aging. It is also expensive relative to other forms of cognitive testing.

These test batteries were developed in the 1980s and 1990s, when theories of human mental abilities were much more preliminary than they are now, and the batteries involve multiple different tests that require extensive training to administer and score. The test batteries were also developed when there was limited understanding of the link between specific cognitive abilities and specific structures and networks in the brain that become less efficient with healthy aging, or fail to function as a result of specific brain diseases or brain injury. These widely used test batteries may be useful in measuring the severity of the cognitive impairment and in identifying the kinds of abilities that are impaired or are spared in a given individual. This information could be extremely helpful when advising individuals and their carers about the possible effect of cognitive impairments on everyday life. However, they are limited as aids to diagnosing the underlying cause of those impairments, and as such they are not effective aids to identifying that the individual is suffering from AD rather than showing the signs of healthy aging, or other disorders such as chronic depression or other forms of dementia.

There is widespread consensus that the main clinical criterion for referring a patient for further diagnostic tests is the presence of memory problems. Indeed, the hallmark of AD is widely considered to be a deficit in memory (Dubois et al., 2007; Perry, Watson, & Hodges, 2000), and memory problems are well established as being associated with, and being very sensitive to, the early stages of the disease. Unfortunately, memory problems also occur in many other disorders, including depression, and even in healthy normal aging, leading to diagnostic uncertainty (e.g., Pfennig, Littmann, & Bauer, 2007; Wright & Persaud, 2007). Moreover, the early sensitivity to AD of memory tests is coupled with performance that becomes too poor after the disease has progressed to allow for reliable or repeated measurement (e.g., Frisoni et al., 2010) to track the progression of the disease or evaluate efficacy of treatment. On the other hand, in less severe patients and in healthy volunteers, many memory tests tend to show effects of improvement simply due to practice across repeated test sessions (e.g., Rabbitt, Diggle, Holland, & McKinnes, 2004). This practice effect could mistakenly be interpreted as a benefit from a treatment (see discussion in Foley, Cocchini, Logie, & Della Sala, 2015). People who show early signs of memory deterioration do not necessarily go on to develop AD, and AD is not an inevitable consequence of aging: More than 50% of people who live into their 80s and 90s do not develop the disease, even if they show a decline in memory ability for other reasons. So, memory tests are sensitive but they are not specific to the disease. Indeed, there may be a danger of overdiagnosing dementia in patients who suffer from other disorders, including chronic depression, that have similar symptoms such as memory loss. Such overdiagnosis has implications for health care costs of inappropriate assessments and inappropriate treatments, and implications for the everyday lives and concerns of individuals and their families (Le Couteur, Doust, Creasey, & Brayne, 2013). We would argue then that one reason that cognitive assessments are seen by clinical policy makers as less promising than biomarkers as a means to help diagnose the disease is because biomarkers are typically designed to be specific to particular disorders and in many cases have been demonstrated to be so. Cognitive assessments typically are designed to be sensitive to the presence of a disorder, not to identify what that disorder might be. Therefore, even if they are much less expensive and more practical for large numbers of people, traditional tests of memory are not ideal for developing AD-specific assessments. More recent evidence from the cognitive and behavioral sciences has pointed toward approaches that might combine the convenience, sensitivity, and low cost of cognitive assessments with much greater specificity for the disease.

Free and Cued Selective Reminding Test (FCSRT)

One cognitive test that has been proposed as a candidate for a specific cognitive marker for AD is the FCSRT (Dubois et al., 2010; Grober, Buschke, Crystal, Bang, & Dresner, 1988). This is a memory test that typically involves first presenting a set of items that participants are then required to recall, first with no cues given, and then in response to specific cues for each item that they had previously failed to recall. The use of the cue procedure is a key characteristic that is not present in traditional memory tests. A consensus conference (Dubois et al., 2007) argued that a version of this task, involving memory for arrays of presented pictures of objects, has high predictive value for the progression toward AD. For example, in a study by Sarazin et al. (2010), 251 participants considered at risk of developing dementia were tested initially and followed at 6-month intervals for up to 3 years. A total of 59 of them developed AD. Out of the tests that were included in that study, the most sensitive and specific for diagnosis of early AD was the FCSRT. However, it remains unclear if the test would meet all of the criteria mentioned above. In particular, a well-established literature has shown that the FCSRT is affected by healthy aging (e.g., Grober et al., 2008; Grober, Lipton, Katz, & Sliwinski, 1998), and it is unknown as to whether or not performance is affected by repeated testing. Thus, poor performance in FCSRT may not offer the best indicator of AD (Carlesimo, Perri, & Caltagirone, 2011; Gainotti, Quaranta, Vita, & Marra, 2014).

Temporary Memory Binding

An alternative approach to the development of cognitive markers has come from clinical and experimental evidence showing that patients in the early stages of AD have problems dealing with multiple sources of information (e.g., Della Sala, Foley, Parra, & Logie, 2011; Foley et al., 2015; Logie, Cocchini, Della Sala, & Baddeley, 2004). A specific form of this online cognitive processing is known as temporary binding. This refers to the processes by which different aspects of stimuli such as colors and shapes are bound together on a temporary basis as an integrated object (i.e., a colored shape; Allen, Baddeley, & Hitch, 2006; Logie, Brockmole, & Vandenbroucke, 2009; Luck & Vogel, 1997; Treisman, 2006). To perform this kind of task, it is essential to keep track of the rapid changes in the environment such as the colors and types of cars about to overtake us on the freeway, or whether we have just taken the yellow oval pill or the white oval pill, or who said what during a conversation. People have to keep track of continually changing temporary bindings in their daily lives, so a breakdown in the ability to do so can undermine independent living. Holding and updating these rapid changes around us is thought to be a function of a system in the brain known as working memory (Baddeley, 2007; Baddeley & Logie, 1999; Logie, 2011) that can hold information for a few seconds and continually update its contents to help us reason, think, solve problems, navigate, hold conversations, and generally interact with the world moment to moment.

Holding color–shape combinations for just a few seconds is very different from learning stable associations that are a function of long-term memory, such as an object and its location, a face with a name, or that the typical receptacle for mail in the United Kingdom is red and cylindrical, but in the United States is blue and rectangular with a rounded top. Evidence suggests that temporary binding and the learning of associations are each supported by a different memory process (Colzato, Raffone, & Hommel, 2006; Logie et al., 2009; Parra, Della Sala, Logie, & Morcom, 2014; Parra et al., 2013; Treisman, 2006) and are differentially affected by brain damage and aging (Parra, Abrahams, et al., 2009; Parra, Della Sala, Logie, & Abrahams, 2009).

It has been demonstrated repeatedly that the learning of associations between stimuli (including objects and colors, word pairs, or patterns and spatial location) is affected in AD (e.g., Buschke et al., 1999; Fowler, Saling, Conway, Semple, & Louis, 2002; Gallo, Sullivan, Daffner, Schacter, & Budson, 2004). This associative learning has been shown to be dependent on a specific area of the brain known as the hippocampus in the temporal lobe (see Figure 1) and damage to this region is thought to be the hallmark of AD (e.g., Lowndes & Savage, 2007). However, associative learning (and hippocampal function) also declines as part of the normal aging process (de Jager, Blackwell, Budge, & Sahakian, 2005; de Jager, Milwain, & Budge, 2002; Old & Naveh-Benjamin, 2008) and in other disorders common in older adults namely chronic depression (Fossati et al., 2004; Gainotti, Marra, Villa, Parlato, & Chiarotti, 1998; Kaschel, Logie, Kazén, & Della Sala, 2009). The lack of specificity of this deficit limits the use of associative learning tasks in the assessment and diagnosis of AD.

OPEN IN VIEWER

Figure 1.

The major lobes of the cerebral cortex of the human brain.

In a series of studies, we have demonstrated that poor performance on tests of temporary binding is indicative of the very early stages of AD. Crucially, unlike other tests, impairments in temporary binding appear to be specific to the disease. We have developed a range of tasks to measure this function through the temporary retention of shapes and colors. The tasks enable the comparison of memory for single features (shapes or objects) and combinations of bound features (colored objects). A typical task is illustrated in Figure 2.

OPEN IN VIEWER

Figure 2.

Test for temporary memory of two shapes, of two colors, or of two-colored shapes.

We demonstrated a selective temporary binding deficit in a group of individuals with AD in comparison with healthy age-matched older adults (Parra, Abrahams, et al., 2009). The AD group showed a disproportionate deficit in the temporary retention of colored shapes in comparison with memory for a color alone or a shape alone. Remarkably, this selective temporary binding deficit was also present in people with a genetic mutation that invariably leads to early onset AD (age around 45-50 years), but who were tested at age 30 to 35 years, that is 10 years or more before they were showing symptoms of the disease (Parra et al., 2010). The task combined more sensitivity and specificity for AD than either associative learning tasks or other standard memory tasks. The task has the added advantage that, because of its reliance on simple shapes and colors, it can be used with people who have impaired language. Moreover, it is not affected by repeated testing (Logie et al., 2009), or by the level of education (Parra et al., 2010), so it can be used to test people with low levels of literacy as well as to test people who are highly educated.

Temporary binding impairments have also been shown to be specific to AD. This binding ability is preserved in healthy older adults who show no difference in memory ability for bindings between colors and shapes compared with memory for colors alone or shapes alone (Parra, Abrahams, et al., 2009; Parra, Della Sala, et al., 2009). This contrasts with associative learning that is vulnerable to the effects of age (e.g., Old & Naveh-Benjamin, 2008). Moreover, we have shown that the task also differentiates between AD and chronic depression as well as between AD and healthy aging (Parra et al., 2010). Depression is common in older adults, and poor performance on standard memory tests is characteristic of both AD and depression. We demonstrated that only the AD patients performed poorly on the shape–color binding condition, whereas people with depression and healthy older adults showed no difference between remembering single features (colors or shapes) and temporary bindings of features (colored shapes). Moreover our most recent work has demonstrated that this temporary binding deficit was not present in other forms of dementia, highlighting that it is specific to AD. Individuals with vascular dementia, dementia with Lewy bodies, frontal lobe dementia, and dementia associated with Parkinson’s disease showed a profile of impairment in memory, attention, executive function, and visuospatial function on standard neuropsychological testing (Della Sala, Parra, Fabi, Luzzi, & Abrahams, 2012). However, as shown in Figure 3, only those with AD showed a specific temporary memory binding deficit compared with memory for single features.

OPEN IN VIEWER

Figure 3.

A deficit in temporary binding is shown for AD but not in healthy older adults (controls) or in other forms of dementia.

Neuroanatomy of Temporary Binding

Cerebral changes very early in the course of AD are selective. Analyses of structural brain scans using Magnetic Resonance Imaging (e.g., Busatto et al., 2003) have shown shrinking of the brain in specific areas (known as focal brain atrophy), mainly in the medial and lateral temporal lobe regions. Figure 4 shows a scan of a healthy brain on the left, and of a brain with temporal lobe atrophy, including the hippocampus, on the right. The brain atrophy then spreads to other areas of the brain. However, it is not the only brain damage characteristic of AD.

OPEN IN VIEWER

Figure 4.

A scan of a healthy brain (left) and of a brain with focal atrophy (darker areas), particularly the medial temporal lobe, including the hippocampus.

The different areas of the brain normally communicate with one another via a network of connections known as white matter tracts; in the course of AD, such connections become progressively thinner and more inefficient particularly between the front (or anterior) and the back (posterior) areas of the brain (e.g., Bokde, Ewers, & Hampel, 2009). Colors and shapes are known to be processed by different areas of the brain (Van Essen & Drury, 1997), and binding colors and shapes together requires communication and integration between those areas (O’Reilly, Busby, & Soto, 2003). Therefore one plausible hypothesis as to why temporary binding shows a specific deficit in AD is because it is dependent on the integrity of communication along these white matter tracts (Parra et al., 2015).

As mentioned above, the hippocampus is well known to be crucial for memory and learning, and to be one of the structures affected first by the brain atrophy that accompanies AD (Braak & Braak, 1991). Due to its prominent role in learning and memory, and to the observation that the earliest symptoms referred to by patients with AD appear to reflect memory impairment, most memory tests developed to date have aimed at tests of the efficiency of the hippocampus (Small, Schobel, Buxton, Witter, & Barnes, 2011). However, focusing too much on the assessment of hippocampal-related memory may delay the diagnosis of AD (see R. A. Sperling et al., 2011). In particular, the hippocampus undergoes significant structural and functional changes as people grow older whether or not they will develop AD (Yang, Goh, Chen, & Qiu, 2013). This overlap with changes in healthy aging prevents the early identification of deficits specific to AD, even with highly validated tests. Nevertheless, recent guidelines continue to emphasize tests of associative learning and memory functions as promising markers for the early detection of AD (Auriacombe et al., 2010; Dubois et al., 2010; Dubois et al., 2007).

More recently, it has been suggested that memory tasks that target the stages of AD that occur prior to major changes in the hippocampus may be more promising in supporting early detection of the disease, than those that target the hippocampal stages (Das et al., 2015; Didic et al., 2011; Stark, Yassa, Lacy, & Stark, 2013; Wolk, Mancuso, Kliot, Arnold, & Dickerson, 2013). Before affecting the hippocampus, AD brain pathology seems to impact on other regions that are close to the hippocampus (extrahippocampal regions), known as the entorhinal and perirhinal cortex (Braak, Braak, & Bohl, 1993; Juottonen et al., 1998). Although this evidence is not entirely new, the role of these regions in supporting memory and development of cognitive tasks that are suitable to assess their function has become clear only over the last decade (Mayes, Montaldi, & Migo, 2007; Montaldi, Spencer, Roberts, & Mayes, 2006). The hippocampus supports learning for associative memory such as remembering details about an object or person encountered before. In contrast, the extra-hippocampal regions appear to support memory for more arbitrary memories for information not explicitly associated with particular details, such as recognizing something as familiar but not remembering where or when it has been encountered. The arbitrary combination of color and shape in temporary binding, shown to be sensitive and specific to AD, would also be an example of this latter form of memory. There is now evidence supporting the notion that associations in memory are different from more arbitrary memory for something being familiar. We will consider three lines of evidence for this difference.

First, studies using functional magnetic resonance imaging (fMRI) scans of the brains of healthy young volunteers have revealed that a temporary memory binding task does not rely on the function of the hippocampus (Rentz et al., 2013). This task activates a network involving specific regions of the brain but not including the hippocampus (Parra et al., 2014). This evidence converges with previous fMRI studies which also assessed similar forms of temporary binding (Shafritz, Gore, & Marois, 2002; Song & Jiang, 2006; Xu, 2007). When more associative information becomes relevant (e.g., the location of an object), the involvement of the hippocampus becomes apparent (Piekema, Rijpkema, Fernandez, & Kessels, 2010).

A second source of evidence for the difference in memory types comes from studies of individual cases of people with specific and focused brain lesions. One individual known by the initials A. E. (Parra et al., 2013), suffered from a stroke resulting in damage to the right hippocampus. The person had a striking deficit in associative binding, but had completely preserved temporary binding. Preserved temporary binding ability but impaired associative binding in the presence of damage to the hippocampus has also been observed in an individual known as Jon (Baddeley, Allen, & Vargha-Khadem, 2010).

A third source of evidence comes from studies, mentioned earlier, of groups of individuals who have a genetic mutation that is known to result in the development of AD when they are in their late 40s. Genetic testing can identify these individuals many years before they show any cognitive or other impairments associated with the disease, for example, when they are between 30 and 40 years of age. Despite showing no evidence of impairments on standard memory and other cognitive tests, these individuals perform poorly on temporary binding tests, specifically temporary binding of color and shape. However, they show normal levels of performance on associative binding tests, such as learning pairs of words (Parra et al., 2010; Parra et al., 2011).

In sum, this neuroanatomical evidence supports the recent proposal that temporary binding tasks that appear to require the use of extra-hippocampal regions in the brain may be more promising for the early detection of AD than traditional memory tasks that rely heavily on intact functioning of the hippocampus. This novel evidence may help explain why the temporary memory binding task has proved insensitive to healthy aging. In contrast to the hippocampus that shrinks as we grow older, brain regions along the lower areas of the brain, known as the ventral stream, such as the fusiform gyrus, entorhinal cortex, and ventral prefrontal cortex seem to undergo functional reorganization with age and allow the maintenance of performance on temporary binding tasks (Grady & Craik, 2000). Moreover, because temporary binding of color and shape does not rely on former associations, nor does it lead to learning or require support from previous knowledge, it is not influenced by repeated exposures (Colzato et al., 2006; Logie et al., 2009) or the cultural background or educational level of the assessed individual (Parra et al., 2011).