CERAD Test Battery

Abbreviation: CERAD

Since August 15, 2019, TDB2Online has had new norms for scoring all CERAD subtests. They are accessed in the TDB2Online app under the name “CERAD”. We recommend these norms for all initial assessments of patients. The new norming is based on a broader basis of random samples of healthy examinees without dementia.

The norms used before August 15, 2019 relied on only two norm sources: for the linearization of the scales on the old work by Welsh et al. (1994)¹⁾ and for the age reference groups on the Basel norms²⁾. We are keeping these old norms in the app under the name “CERAD old” until further notice, so that patients examined earlier can continue to be followed longitudinally with them.

The classic subtests of the CERAD include the Mini Mental State Examination (MMSE), and the subtests of the CERAD-Plus (see below) include the Trail Making Tests. For these procedures there is a substantially broader normative basis than for the other CERAD subtests; in particular, there are also norms for young adults for them. They are therefore listed twice in the app and in the documentation: once under their own names MMSE and Trail Making Tests A and B with the broad normative basis, in which the performance scores are based on the values of young adults, and once as subtests within the CERAD battery, as documented in the following.

In the mid-1980s, several North American clinics and research institutions, under the aegis of the National Institute on Aging, joined together under the name Consortium to Establish a Registry for Alzheimer's Disease (CERAD) to jointly collect clinical, neuropsychological, laboratory and imaging data from patients with Alzheimer's dementia. The neuropsychological test battery on which the institutions agreed³⁾ is known among neuropsychologists by the abbreviation CERAD Test Battery. It is a screening battery for the domains of orientation, language, memory and (to a limited extent) visuoconstruction, whose level of difficulty is adapted to patients with incipient dementia.

The CERAD battery quickly became established in other countries and languages. In the German-speaking region, the Memory Clinic of the Geriatric University Hospital Basel, under the aegis of Hannes B. Stähelin, presented an authorized translation. After the American norms of the battery had been used for some time, a multicenter norming study was started in German-speaking countries in 1997, which is described in most detail in the electronically published dissertation by Chantal Aebi ⁴⁾.

Since 2005, the Basel version of the CERAD battery, under the name CERAD-Plus, has had three additional subtests. This version is certainly the most widely used in the German-speaking region.

Tables 1a and 1b give an overview of all subtests of the CERAD-Plus, their assignment to cognitive domains, the abbreviations used here, the rank order in test administration and the possible raw score range.

Table 1a: Domains and subtests of the classic CERAD

Table 1b: Domains and subtests of the “Plus” additions

The German test material for the CERAD battery, authorized by the American authors, can be either purchased as a paper version by professionals after registering on the website of the Basel Memory Clinic or downloaded free of charge as a pdf file. In addition, the Memory Clinic also offers scoring software and a scoring service that are based on the Basel norms.

The CERAD Consortium claims rights to the CERAD materials. They can be ordered there by individuals and used for scientific and/or clinical purposes. Use outside these two purposes, and in particular passing them on to third parties, is prohibited.

There are a number of foreign-language versions, including the approved German version of the Basel Memory Clinic. Obtaining and using the German version is subject to similar restrictions as the original US version.

The subtests of the CERAD battery include as standard both the Mini Mental State Examination (MMSE) and a short form of the Boston Naming Test. The MMSE was included in the battery at a time when the authors of the original article did not enforce copyright for it. In the meantime, however, copyright is enforced for the MMSE. The same applies to the Boston Naming Test. A search in the WorldCat catalog for “Boston Naming Test” shows that since its first publication in 1967 it has been published by at least three different publishers and additionally in several language versions, but currently probably exclusively by Pro-Ed in Austin, TX. The above-mentioned American website for obtaining the CERAD materials now carries a note that obtaining both the MMSE sheet and the Boston Naming Test is permitted only if the requester can show a license agreement with the rights holders. The request form links to it. The German-language version is currently available without these restrictions.

The CERAD is administered as an individual test. Patients examined with the CERAD typically have at least a suspected dementia, often a manifest dementia. For administration of the CERAD, therefore, not only knowledge of the test material and the administration instructions but also a certain amount of experience with patients with dementia is necessary.

The Basel test manual describes the administration and scoring of the subtests very clearly and in detail. TDB2Online does not score all raw scores described in the Basel manual. Since we always place particular weight on the largest and most representative possible normative basis, we restrict ourselves to those parameters for which the best normative samples exist.

For practical use as a screening test for (Alzheimer's) dementia, a measurement or estimate of premorbid intelligence is additionally necessary. In the memory clinic of the Department of Psychiatry at the University Hospital of LMU Munich, the CERAD battery is regularly supplemented by Schmidt & Metzler's Vocabulary Test (WST)⁵⁾, whose items require recognizing one word among several pseudowords. At least in milder dementias, this measurement of passive verbal abilities allows an estimate of premorbid intelligence. As an alternative or also a supplement, the use of a social formula is also useful, with which premorbid intelligence can be calculated from data obtainable from the patient history. Social formulas for German-speaking patients were most recently proposed by Jahn et al. 2013.⁶⁾

We use the following raw scores for the classic CERAD subtests:

• Mini Mental State Examination: points (range 0-30)
• Word List Learning: number of words learned across all three learning trials (range 0-30)
• Word List Recall: number of correct words named spontaneously (0-10)
• Word List Recognition: number of correctly assigned words minus 10 (0-10)
• Semantic word fluency: number of correct words named in one minute (open-ended, range tabulated from 0 to 36)
• Naming: number of correctly named drawings (0-15)
• Constructional Praxis: number of points across all four figures (0-11)
• Constructional Praxis Recall: number of points across all four figures (0-11)

and for the subtests of the CERAD-Plus:

• Trail Making Test A: time in seconds (open-ended, range tabulated from 10 to 500)
• Trail Making Test B: time in seconds (open-ended, range tabulated from 20 to 500)
• Lexical word fluency: number of correct words named in one minute (open-ended, range tabulated from 0 to 36)

For most CERAD subtests, no norming studies in young adults are available. We therefore cannot relate the performance scores of the CERAD subtests to the performance of young adults, as we otherwise do in TDB2Online. For all CERAD subtests, the performance scores are therefore related to the performance of 60-64-year-olds.

For two procedures, the MMSE and TMT A and B, good norms for young adults exist. If these procedures are administered within the CERAD battery, however, it would not be sensible to score them on the basis of different norms than the remaining subtests. Both procedures are therefore available twice in TDB2Online: once as subtests of the CERAD with a uniform norming for persons aged 60 years and above, and once as separate tests (MMSE and Trail Making Tests A and B) with the usual norming of the performance scores on the basis of young adults.

For the TDB2 norming of the complete CERAD, the studies listed in Table 2 were taken into account. The selection criterion was that healthy persons in the older age range (drawn more or less randomly and/or representatively) were examined and that the results were published separately by age group. Details on the studies are explained below; Table 2 first gives an overview of the norming studies considered and the corresponding reference. For the sake of readability, these studies are later referenced only by the name of the study or the first author.

Table 2: Overview of the norming studies used

On the studies in detail:

The ADAMS study (overview in Langa et al., 2005)¹⁶⁾ comes very close to the ideal of a norming study. It has the following background: Since 1992, the Institute for Social Research of the University of Michigan has been conducting the “Health and Retirement Study” (HRS), which now represents the entire US population over 50 years of age with more than 30,000 persons (5000 died after entering the study). In connection with the HRS, the Aging, Demographics, and Memory Study (ADAMS) was conducted, whose aim is a population-based dementia study. In a sample of 856 persons aged 70 and older, the CERAD and the Trail Making Tests, among others, were administered at home by trained professionals between 2001 and 2003. This is the only study that actually contains, at first, the data of a random selection without further selection. Many further data are available that make it possible to specify the results for particular groups, for example to exclude persons with existing dementia. For one's own research, the data are available on request via the HRS website.

In the second study in Table 2 (Welsh-Bohmer), too, a random sample of persons over 65 years was examined, albeit within a spatially limited population (Cache County in Utah). All persons were carefully clinically evaluated. Dementia or mild cognitive impairment (MCI) (diagnosed in the context of an illness) led to exclusion.

In the study by Luck et al., a random sample of non-demented patients from 138 general practices was examined with the CERAD. The age range was set at 75 years and older; results are reported for four subtests of the CERAD, but given the large sample, the age groups are little differentiated (only two groups: 75-79 and 80+).

Fillenbaum et al. compiled norm results for the subtest “Constructional Praxis Recall” from three studies, two of which are based on random samples from limited populations, the third on a convenience sample.

The other five studies in Table 2 did not claim to examine random samples. Welsh et al., Collie et al., and the Basel norming of the German CERAD-Plus (entry “Monsch” in Table 2; manual and norms retrievable after registration and login on the website of the Basel Memory Clinic, where further literature is also found) examined mainly healthy relatives of patients in memory clinics, supplemented by volunteers recruited in different ways. Murphy examined volunteers in Cork (Ireland) and the surrounding area; Beeri et al. examined residents of senior centers and other recruited persons as part of a longitudinal study on cardiovascular risk factors. In all five studies, persons with manifest dementia were excluded.

Studies with the CERAD in developing countries (for example ¹⁷⁾,¹⁸⁾) were not considered because the persons examined there had, in the majority, less than 10 years of schooling.

The processing of the classic CERAD subtests for TDB2Online has some special features, which are explained in the following.

Normally, in TDB2Online test performances are referenced to the average performance of young adults (“performance scores”). This is not possible for the CERAD battery because norming studies with young adults were not carried out for most subtests. For the CERAD battery, the performance scores refer to healthy persons aged 60-64 years.

Normally, TDB2Online relies primarily on norming studies based on random samples. For tests used predominantly in the age range between 20 and 65, such samples provide the best basis for test standardization. It becomes more difficult with tests intended for old and very old age. Age-associated pathological cognitive disorders account for a relatively large proportion of the sample in random samples at advanced age; the older the group, the greater the proportion. In most norming studies for the CERAD (see Table 2), persons with age-associated pathological cognitive disorders were excluded from the outset, even though the exclusion criteria are often poorly defined. In the ADAMS study (described in Langa et al. (2005)¹⁹⁾), in contrast, no one was excluded at first. Here, a random sample of 856 persons, representative of the US population over 69 years, was visited at home and examined with the complete CERAD including MMSE and TMT. In addition, extensively documented, systematic clinical examinations were carried out for existing illnesses with effects on cognition, which make it possible to test the effects of various exclusion criteria. The clinical examinations took place in several waves, so that it is also possible to identify persons who met the criteria for dementia a few years after the initial examination with the CERAD. If the original data of this study are processed accordingly, the effect of different exclusion criteria on test performance can be demonstrated well. For this purpose we use the MMSE data of the ADAMS study. On the basis of the original data, we formed four groups:

• Group 1: the complete population-representative (US population 2001) initial sample at the initial examination (time point A), N=833
• Group 2: the initial sample without those with dementia at time point A, N=738
• Group 3: the initial sample without those with dementia at time point A and without those who developed dementia over the following 4 years (measured at time point B, four years after time point A), N=649
• Group 4: only those who were neurologically and psychiatrically completely healthy at time point A, N=551

In Figure 1, the mean MMSE raw scores of these four groups across the covered age range are plotted as dashed lines with the labels ADAMS_M1 to ADAMS_M4.

Figure 1: Age trajectories of the mean MMSE raw scores of the four ADAMS groups (see text)

In addition to the four ADAMS groups (ADAMS_M1 to ADAMS_M4), Figure 1 also contains the data of two further random samples in which there was no selection of participants with cognitively relevant illnesses. The two studies, Crum et al. (1993)²⁰⁾ and MRC CFAS (1998)²¹⁾, are among the norming studies of the MMSE and are described in the corresponding section of the test documentation on the MMSE. It can be seen that the means of the pure random samples (Crum_M, MRC_M and ADAMS_M1 in Figure 1) agree well with one another across the age course from 70 to 94 years. The proportion of persons with pathological cognitive impairment increases from about age 70, the older, the more. This has the following effects on the sample sizes in Groups 2 to 4:

• In Group 2, in which only those with dementia at the time of testing were excluded, the group size is reduced compared with the total sample by 4% in the age group 70-74, by 6% in 75-79, by 16% in 80-84, by 29% in 85-89 and by 28% in the age group 90-94.
• For Group 3 the corresponding figures are 12%, 16%, 28%, 41% and 46%,
• for Group 4 20%, 22%, 48%, 57% and 70%.

The effects on the mean trajectories are considerable: Figure 1 shows that the exclusion of those currently with dementia has no appreciable effect yet in the 70-74-year-olds. Here the MMSE means of Group 2 lie very close to the three random samples. In the following age groups this changes: the considerable decline in performance that we observe in the random samples is reduced to a only very gradual decline in performance, which in the higher age groups does not look much different from that in the two decades before. The additional exclusion of persons who develop dementia in the next four years (definition of Group 3) has a certain, but not very large effect on the age trajectory. The effect is somewhat stronger if one excludes from the outset all persons who are in any way psychiatrically or neurologically abnormal (Group 4). This selection obviously leads to an overall higher-performing group.

The informative value of these findings is very high, since they were obtained in large (ADAMS) and very large (Crum and MRC-CFAS) random samples. The interpretation is clear: The considerable decline in performance of persons over 70 years that is observed in the random samples, and that increases with higher age, is attributable to those with dementia, and indeed to a proportion that increases with age. It is thus based on an age-associated change in the composition of the sample.

For the norming of the CERAD subtests, we therefore fall back on the data of Group 2 for the ADAMS study; in most other norming studies persons with dementia are excluded anyway.

The subtests of the CERAD yield raw scores with very different distributional properties. The raw scores of the Semantic word fluency subtest, for example, extend over a sufficiently large range of values and are also approximately normally distributed. Quite different, by contrast, is the Word List Recognition: here the range of values goes only from 0 to 10. Because most persons recognize 9 or 10 words, the distribution is also extremely skewed.

The further measures for the psychometric processing of the raw scores into meaningful performance scores are therefore presented below for each subtest individually. The starting point for the distribution analyses is always the ADAMS sample without those currently with dementia (= Group 2), since the complete distribution of values for all subtests is available there. The other random samples from Table 2, Welsh-Bohmer²²⁾, Luck²³⁾ and Fillenbaum²⁴⁾, are taken into account in different ways for the subtests.

For the assessment of the normative samples, the means of all other samples from Table 2 are also displayed together with the random samples, regardless of whether they enter into the norms.

see the detailed test documentation

see the detailed test documentation

see the detailed test documentation

see the detailed test documentation

see the detailed test documentation

see the detailed test documentation

see the detailed test documentation

see the detailed test documentation

see the detailed test documentation

see the detailed test documentation

see the detailed test documentation

At the Psychological section of the Department of Psychiatry and Psychotherapy, University Hospital, LMU Munich, the CERAD battery has been used since the mid-1990s, predominantly in the context of the memory clinic. The American norms were processed and linearized early on for this purpose; a profile sheet was used for presentation that anticipated some features of the presentation principle of TDB2Online,²⁵⁾ but is now outdated.

Until August 14, 2019, the CERAD norms used in TDB2Online were based exclusively on two sources: The performance scores were derived from the percentile distributions of Welsh et al. (1994)²⁶⁾; for the age norm limits, the data of the norming study of the Basel Memory Clinic were additionally used. This now outdated norming remains available in TDB2Online for comparison purposes until further notice under the name CERAD old.