Abstract
Evaluation of mental status in healthy and cognitively impaired elderly is often based on brief instruments that do not comprehensively assess many areas of cognition [1]. As most neuropsychological tests are affected by confounding covariables such as age, culture, language, gender, level of education, intelligence, and floor and ceiling effects [1–5], the ability of these brief examinations to distinguish early cognitive decline from normal age-associated changes is limited [6]. Considerable evidence now exists for a long prodromal period in the Alzheimer's disease (AD) process, during which very subtle cognitive impairment may be detectable [7–8]. Therefore, a more extensive and more sensitive approach to the assessment of cognition in populations ‘at-risk’ for AD is necessary.
Recently, the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) has defined a brief but comprehensive battery of neuropsychological tests aimed at assessing the areas of cognition commonly observed to decline early in the course of AD. These include memory, learning, language, praxis and general cognition [1],[9]. This battery has been adopted in many research and clinical centres as the primary neuropsychological assessment battery for the detection and diagnosis of AD [10]. The CERAD battery has been shown to have high validity [9],[11], reliability, and sensitivity in mild and very mild cases of dementia [12],[13] in older American populations [10]. However, the suitability of the battery for similar use in an Australian setting has not yet been determined. In addition, although tests were selected for inclusion in the CERAD battery on the basis that they measured specific cognitive abilities, the relationship between the different areas assessed by the CERAD battery have yet to be determined, either in the United States or elsewhere.
Previous investigations using the CERAD battery also indicate that neuropsychological test performance in a normal elderly population varies significantly with age, gender and level of education [1]. More specifically, males, older, and less-educated subjects perform more poorly on most neuropsychological measures than female, younger and more educated subjects. In other studies, high levels of anxiety and depressive symptoms have also been shown to decrease cognitive performance in older people [14]. Considering that increased levels of depression and anxiety are common in older people who are referred for investigation of changes in cognition, the investigation of the effect of mood on CERAD performance is also important. However, the effects of mood variables on cognitive factors measured by the CERAD battery has not been established.
The current study measured the performance of a large group of healthy older individuals on the CERAD neuropsychological battery in order to: (i) provide normative data in an Australian setting; (ii) determine the general domains of cognition that the battery assesses; (iii) confirm the effect of age, gender and level of education on performance; and (iv) investigate the effects of anxiety and depressive symptoms on neuropsychological performance in a healthy elderly sample.
Method
Participants
Two hundred and forty-three Australian volunteers older than 44 years of age (mean age = 63.1 years, SD = 9.02 years; 79 males) participated in the study. All were recruited through the media or through contact with the research institute and all spoke English as their first language. All subjects were assessed by a neurologist, a neuropsychologist and a psychiatric nurse. Exclusion criteria included meeting the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA) criteria for dementia [15], a score below 28 on the Mini-Mental State Examination (MMSE) [16] or above the age-appropriate limit on the Short Blessed Test (SBT) [17], the presence of two or more abnormal neurological signs indicated by the CERAD neurological examination [18], past or current history of stroke, hypertension, major vascular disorder, heart disease, head injury, epilepsy, diabetes, thyroid disease, major depressive or anxiety disorder or any other psychiatric illness. Factors such as education, race, intelligence and socioeconomic status were not included in the criteria for selection. However, as can be seen in Table 1, the majority of subjects were well educated and had above average estimated full-scale intelligence.
Materials
A comprehensive medical history was taken to determine health status. Information regarding the age, gender and years of education of each subject obtained in the first interview was used for the purposes of the current study.
The CERAD battery of neuropsychological tests was administered to all participants. The battery consists of the Modified Boston Naming Test [19], Categorical Verbal Fluency, Word List Learning, Word List Recall, Constructional Praxis ‘line drawing’, Word List Recognition, and the MMSE [16]. The State-Trait Anxiety Inventory (STAI) [20], and the Centre for Epidemiological Studies in Depression (CES-D) [21] rating scale were also administered, while the National Adult Reading Test (NART) was administered to estimate full scale IQ. The SBT [17], a cognitive screening test not unlike the MMSE, on which higher scores indicate more errors and a greater likelihood of cognitive decline, was also administered.
Procedure
All examinations were undertaken at the Mental Health Research Institute of Victoria. Upon arrival at the institute subjects gave informed consent, and were asked to complete the CES-D and the STAI. A detailed self-report medical history was taken by a clinical psychiatric nurse, followed by a neurological examination. The CERAD battery, the SBT and the NART were then administered by a neuropsychologist or trained research assistant.
In order to examine the effects of education, age and gender on performance, participants in the current study were divided into the same groups as the original CERAD sample. More specifically, subjects were categorised according to age (≤ 69 years, and > 69 years), education (< 12 years, ≥ 12 years), and gender (see Table 1).
Group means and (SD) for education, age and gender categories on Consortium to Establish a Registry for Alzheimer's Disease measures, the Nelson Adult Reading Test and the Short Blessed Test
Data analysis
A number of different methods of analysis were used in the current study. First, the group means and standard deviations were calculated for all measures for each age, gender and education subgroup. These were compared to the original CERAD sample. Any mean from the current sample that was more than one standard deviation from the same CERAD category was considered to be significantly different.
A principal components analysis (PCA) of CERAD measures was conducted in order to identify the broader domains of cognition assessed by the battery. Factor scores for each participant were calculated for all factors identified. Finally, the independent variables age, gender, education, level of anxiety and depression were then included in a linear multiple regression in which factor scores were the dependent variables. Age, gender and education were included in the regression analysis at stage 1 to determine their effects, with the mood variables added at stage 2, in order to determine whether they increased the predictive power of the model.
Methods
The means of all measures are within 1 SD of the equivalent CERAD normative study results (see Table 1). The PCA generated a five-factor solution with eigenvalues > 1, which accounted for 64.50% of the variance in cognitive test performance (see Table 2). The first factor received high loadings (r > 0.70) from measures of delayed recall, word learning trials 1–3 and word recognition-yes, and accounted for 21.29% of the variance in overall test performance. The second factor had high loadings from the MMSE language/copying, orientation and registration subscales and accounted for a further 15.08% of the variance in test scores. The praxis and word recognition-no measures loaded on the third factor, which accounted for another 9.80% of the variance. The fourth factor had high loadings from both the verbal measures (categorical naming and overt verbal fluency) and accounted for a further 9.66% of the variance of test scores. Finally, the fifth factor received high loadings from the MMSE recall and attention/calculation subscales, accounting for another 8.66% of the variance of test scores.
Principal Components Analysis for the Consortium to Establish a Registry for Alzheimer's Disease neuropsychological test battery (healthy ageing cohort, 243 subjects)
A multiple linear regression of demographic (age, gender, education) and mood variables (state anxiety, trait anxiety, depression) on these five factors indicated that age, gender and education were the most consistent and significant predictors of performance on three of the five PCA factors (see Table 3). However, adding the mood variables at stage 2 did not result in significant increases in predictive ability (r2 change). More specifically, for the first factor (Learning and Recall), both age and gender recorded significant Beta values at stage 1. No other variables reached significance when included in stage 2. On the third (Praxis) and fourth factors (verbal fluency and naming), education reached significance at both steps. On factors 2 and 5, both of which consisted wholly of MMSE subscales, no Beta values reached significance (see Table 3).
Multiple linear regression of the independent variables age, education, gender, anxiety and depression on the five Consortium to Establish a Registry for Alzheimer's Disease factors
Discussion
The current study investigated cognitive performance in a sample of healthy aged individuals using the CERAD neuropsychological test battery in order to establish norms, and investigate the effects of the confounding covariables age, education, gender, anxiety and depression. The mean test scores of the current Australian cohort were all within the normal range as defined by the normative sample [1], allowing direct comparison with the CERAD normative study when investigating the effects of age, gender, education and mood. A PCA revealed five factors with eigenvalues greater than one. These factors loaded independently on measures of memory and learning, verbal fluency/naming, praxis, orientation/registration and executive function. These factors represent the areas most commonly observed to decline early in the course of AD [1],[5],[12]. Age, education and gender all had significant effects on neuropsychological performance in these cognitive domains. However, the mood variables state and trait anxiety and depression had no effect on performance.
The means for neuropsychological tests in the current study were all within one standard deviation of those found in the CERAD normative study. As these measures have been widely adopted in healthy aged populations in which the battery has previously been validated [1], the current results indicate that the CERAD neuropsychological battery is a valid and reliable tool for use in the detection of cognitive impairment in an Australian setting. Unfortunately, the number of participants within the older age, high education group in the current study are limited and may not provide an accurate representation of normative data. Further data may need to be collected in this demographic group before the battery's applicability to such aged individuals can be investigated.
Five factors were extracted from the CERAD battery via a PCA. These factors loaded on measures of memory and learning (word list recall, word list learning trials 1–3 and word recognition-yes), language (verbal fluency, categorical naming), praxis (constructional line drawing, word recognition-no), executive function (MMSE subscales attention/concentration and recall) and registration/orientation (MMSE subscales orientation, registration and language). These factors are consistent with earlier research which has also found deterioration of these areas of cognition in AD patients. For example, Becker et al. [22] identified deficits in visuoconstructive (praxis) and lexical/semantic abilities in AD patients. Many other studies have reported deficits in visuoconstructive and verbal abilities in AD patients [11],[22–26]. More recently, other researchers have advocated the role of measures of verbal abilities in predicting the onset of dementia [27],[28].
Other measures included in the CERAD battery that have been shown to deteriorate in AD include delayed recall of word lists [12],[13] and measures of global cognitive function, such as the MMSE and the SBT. The latter is perhaps the most commonly reported area of decline in AD [11–13],[17],[22],[29–32]. For example, Fillenbaum et al. [29] found that the MMSE was effective at discriminating between stages of AD (mild, moderate and severe) and also at discriminating AD from normal controls. In a longitudinal study of a large cohort of community residents aged over 85, Izaks et al. [31] found that MMSE score declined on average 4 points over a 3-year period. Other studies have included measures of global function in batteries aimed at detecting preclinical AD [8],[27],[33–36]. The fact that the CERAD battery assesses these most often cited areas of decline in AD confirms its utility for the detection and diagnosis of AD.
The effects of the demographic variables age, education and gender were clearly evident in the current study (see Table 2). For example, on three of the five factors identified either age, gender or education, or a combination of these, were significant independent predictors of performance. These results support previous research which has suggested that age is a risk factor for loss of cognitive capacities indicated by the onset of dementia [1–5],[29],[37–39]. For example, Benton et al. [40] found that the frequency of defective performances on neuropsychological measures among an elderly population increased with age. Similarly, lack of education has also been previously cited as a risk factor for poor cognitive performance [1]. While gender differences are not as commonly reported, the normative study of the CERAD battery did report differing neuropsychological performance between males and females [1].
While the current results suggests that mood variables have no predictive value over and above that demonstrated by variables such as age and gender, previous research has found that mood may play an important role in the determination of the individuals cognitive performance [14]. However, there is also support for the current findings among the literature [41]. For example, Waldstein et al. [41] found that levels of anxiety measured by the STAI did not have an effect on neuropsychological performance among older community dwelling subjects. When the demographic and mood variables are considered together, our results suggest that mood had no effect on neuropsychological performance. Further, mood variables did not enhance a model for predicting cognitive performance that comprised a combination of the variables age, gender and education. It is important to emphasise that the absence of any effect of mood on the cognitive performance in the current study should not be taken to indicate that mood will have no effect when the CERAD battery is used in more general settings. In the current study, individuals with mood disorders or any history of psychiatric illness were excluded from participation. Therefore, levels of anxiety and depressive symptoms were generally very low. Further study of healthy individuals with high levels of depressive or anxiety symptoms will be necessary to establish the precise nature of any effects of mood on cognitive performance. However, the current data provide a sound foundation for such studies.
In conclusion, it can be stated that the CERAD battery is a valid and reliable measure of cognition in an elderly Australian sample, and is also applicable for use in detecting and diagnosing AD. The norms of the current sample are very similar to that of a large American sample [1], where the battery has been used for such purposes. The CERAD neuropsychological battery assesses a number of general domains of cognition, all of which have been consistently shown to be at risk in the AD process: memory, language, praxis, general cognition and executive function [9], [12], [22], [23], [30], [31]. This finding enhances the validity of the battery for use in individuals with AD. Finally, the confounding variables age, gender and education all had significant effects on neuropsychological test performance in a normal elderly sample, while the effects of mood were negligible. This finding is generally consistent with previous research investigating neuropsychological performance in healthy and dementing elderly populations [1],[41],[42], and further demonstrates the validity of the battery for use in such populations.
Footnotes
Acknowledgements
The authors would like to acknowledge Kellie Rice, Sheree Cairney, Justin Curtin, Rebecca Summerfield and Rishma Vidyasagar for their assistance with data collection.