Prevalence of age-associated memory impairment in a randomly selected population from eastern Finland

Older adults frequently report memory disturbances in their everyday life that are generally thought to be associated with aging. There is also scientific evidence that aging affects various aspects of memory in normal subjects ^[1-3]. Kral ^[4] introduced the term "benign senescent forgetfulness" to describe memory dysfunction related to advancing age, but this concept has been criticized as being semantically inappropriate, imprecisely defined, and intended to describe a particular subset of older people with memory impairment ^[5]. The National Institute of Mental Health (NIMH) Work Group on Aging and Memory proposed research criteria for age-associated memory impairment (AAMI) to describe memory decline with age (table 1) ^[5,6]. Briefly, these criteria include the presence of complaints of gradual memory loss in tasks of daily life in persons older than age 50, objective evidence of impairment on a standardized memory test as compared with the mean established for young adults, evidence of adequate intellectual function, and absence of dementia or any medical condition that could produce cognitive deterioration.

The concept of AAMI has been criticized. The criteria for measuring objective memory impairment are postulated to lack reliability, ^[7,8] and the whole concept is said to be too broad a clinical entity ^[9,10]. Blackford and La Rue ^[11] also suggested subsequent revisions of the AAMI criteria. Although the validity of this disorder is controversial, Brayne and Calloway ^[12] suggested that AAMI forms a continuum from normal aging to Alzheimer's disease (AD). However, such a continuum was not implied by the NIMH Work Group criteria for AAMI ^[5].

Information on the prevalence of AAMI is limited. Some researchers have estimated that most people older than age 50 are affected ^[13]. Two recent epidemiologic studies with different methodologies reported quite variable prevalence rates for AAMI: 34.9% in an Australian study ^[14] and only 4.6% in a Spanish study ^[15].

The aim of this study was to evaluate the prevalence rate of AAMI by applying the research criteria proposed by the NIMH Work Group in a randomly selected population of 1,049 elderly subjects combined from three separate subpopulations. We also evaluated the impact of age and sex on the prevalence rate for AAMI.

Methods. Subjects. Population 1. A random sample of 592 persons born from 1912 to 1921 was drawn from the Kuopio, Finland, population register, which included all 5,286 inhabitants of the town of Kuopio in this age group on January 1, 1986. Of the 592 persons randomly selected, 403 were completely evaluated. A total of 79 subjects had died before the investigation, 11 had moved outside the study area, 17 were too ill to participate, 55 refused to participate, and 27 could not be contacted. Because the data for the assessment of AAMI were missing for one patient, 402 completely evaluated subjects were included for analysis. Thus, the participation rate in this substudy was 80.3%. The examinations were carried out from January to August 1989.

Population 2. Another random population sample of 250 persons born from 1922 to 1930 was drawn from the Kuopio population register to evaluate the prevalence rate for AAMI in a younger age group. Of 250 randomly selected persons, 176 (70.4%) participated in this substudy. Examinations were done from January to February 1990. No data on nonparticipants were collected.

Population 3. To increase the power of the study on the prevalence of AAMI, we also examined a third population sample of 667 subjects born from 1917 to 1927. This random population sample was also taken from the Kuopio population register; however, all persons included in subpopulations 1 and 2 described above were excluded. Eventually, 482 persons participated in this substudy. A total of 108 were not willing to participate, 62 could not be contacted, 11 were too ill to participate, and four had died before the study. Thus, the participation rate was 72.7% in this substudy. Of those who participated, data for the assessment of AAMI were available from 471 subjects (97.7%). These examinations were done from May to September 1992.

After combining the three subpopulations, a total of 1,049 subjects aged 60 to 78 years were completely evaluated, giving an overall participation rate of 74.4% (table 2).

Methods. All participants were interviewed for demographic information, medical history, current medication, a history of smoking and alcohol consumption, and subjective assessment about memory disturbances or depression during the past year. Subjective complaints of memory loss were recorded by a memory complaint questionnaire (MAC-Q) ^[16] consisting of six questions regarding one's current ability to remember everyday things, eg, names of other people or phone numbers, as compared with instances when the person was young, by a five-stage scoring scale for each question. Total scores on the MAC-Q can range from 7 to 35, higher scores reflecting more complaints of memory loss. A score of 25 or more is regarded as indicating subjective memory impairment. Memory capacity was evaluated by the Benton Visual Retention Test (Administration A) (BVRT), ^[17] in which a score of 6 or less was a cutoff score for AAMI, and by the Associate Learning subtest of the Wechsler Memory Scale (WMS-AL), ^[18] in which a score of 13 or less in total pairs was a cutoff score for AAMI in subpopulations 1 and 2. In addition, the Logical Memory subtest of the Wechsler Memory Scale (WMS-LM) ^[18] was applied in subpopulation 3, in which intellectual function also was evaluated by using the Vocabulary subtest of the Wechsler Adult Intelligence Scale (WAIS-V) ^[19]. To examine mental capacity and to rule out the presence of dementia, all subjects of the three subpopulations were tested by the Mini-Mental State Examination (MMSE), ^[20] in which a score of 24 or more was regarded as excluding the presence of dementia. The NIMH Work Group research criteria were used for the final diagnosis of AAMI (table 1) ^[5]. In subpopulation 1, brief neuropsychological tests recently described elsewhere ^[21] were also used to screen out dementia.

The interview and neuropsychological tests were carried out by trained personnel (nurse, psychologist, or doctor) during a single visit at the Memory Research Unit, Department of Neurology, University of Kuopio. Standardization of the interview and tests was accomplished by using an instruction manual. All participants or their family members gave informed consent. Each of the substudies was approved by the Ethics Committee of the University of Kuopio.

Statistical analyses were carried out at the Computer Center of the University of Kuopio by using the SPSS programs. The results for continuous variables are given as a mean +-\SD and, for categorical data, as percentages.

Results. Of the 1,509 persons included in the various subpopulations, 1,049 were completely evaluated (table 2). Table 2lists the clinical characteristics of the subjects in the three subpopulations who participated in the interview and neuropsychological tests of the whole population. The mean age of the whole population was 68.8 years (SD+-\3.9; range, 60 to 78). The most frequent previous diagnoses among the whole study population were hypertension (38%) and coronary heart disease (23%).

Subjective memory impairment was found according to the MAC-Q score in most of the subjects in all three subpopulations (79.8% in subpopulation 1, 79.0% in subpopulation 2, and 73.9% in subpopulation 3) (table 3). Objective memory impairment was assessed by the BVRT and the WMS-AL in all study populations. In the youngest group (subpopulation 2), 67.4% of persons scored 6 or less in the BVRT as compared with 78.4% in the oldest group (subpopulation 1) (table 3). According to the WMS-AL, the prevalence of objective memory impairment was lower: 31.9% scored 13 or less in the youngest population and 43.6% in the oldest population.

All objective memory tests proposed by the NIMH Work Group were performed in substudy 3 (table 3). In the WMS-LM, 36.3% of subjects scored below the cutoff point of 6 or less. A raw score of 32 or more in the WAIS-V was reached by 57.7% of persons in study 3. According to the MMSE, dementia was suspected (total score of 23 or less) in 14.2% of subjects in subpopulation 1, 2.3% of subjects in subpopulation 2, and 8.7% in subpopulation 3.

A total of 564 subjects in the whole population (239 men, 325 women) were classified as having AAMI by the inclusion criteria alone, ie, they had subjective memory impairment, memory test performance below the cutoff score for AAMI in at least one of the tests, and no dementia, resulting in a prevalence rate of 53.8% (men, 57.4%; women, 51.3%) (table 4). The prevalence rate was highest (58.6%) in the youngest age group of 60 to 64 years (men, 58%; women, 59.1%) and lowest (48%) in the oldest age group of 75 to 78 years (men, 53.8%; women, 45.9%). The prevalence rate for men was higher than for women in all age groups except the youngest age group of 60 to 64 years (men, 58.0%; women, 59.1%).

When the exclusion criteria for AAMI were taken into account, including illnesses that could cause memory impairment, the prevalence of AAMI decreased to the rates presented in table 5. A total of 403 subjects (177 men, 226 women) were diagnosed as having AAMI according to the proper criteria of the NIMH Work Group. Thus, the prevalence rate for AAMI in the total study population was 38.4% (men, 42.5%; women, 35.7%). Also, by these criteria the prevalence rate was highest (45.7%) in the youngest age group of 60 to 64 years (men, 50.0%; women, 42.4%) and lowest (33.0%) in the oldest age group of 75 to 78 years (men, 34.6%; women, 32.4%).

Discussion. We found a high prevalence rate (38.4%) for AAMI, as defined according to the proper criteria of the NIMH Work Group, in a randomly selected population of 1,049 elderly subjects aged 60 to 78 years. The prevalence rate was higher in men (42.5%) than in women (35.7%), and it declined with age in both sexes. By the proposed inclusion criteria alone, ie, complaints of gradual memory impairment in tasks of daily life, evidence of memory impairment in objective memory tests, and absence of dementia, the prevalence rate for memory impairment in the elderly was more than 50%.

In a recent longitudinal epidemiologic study from Sydney, Australia, the prevalence rate for AAMI was 34.9% and the annual incidence rate was 6.6% in the age group 65 years and older ^[14]. However, in a Spanish community-based survey of the population aged 40 years and older, the prevalence rate for AAMI was much lower: only 3.6% in individuals 40 years and older and 7.1% in individuals 65 years and older ^[15]. However, these two studies used an unstandardized methodology and definition of AAMI. In the Australian study, the study population was small and highly select: only 146 subjects from 1,000 dwellings completed their AAMI assessment. The Spanish study used a short screening test with 10 items aimed primarily at identifying potentially demented patients during the door-to-door phase; this screening test identified only 52 of 476 subjects (10.9%) interviewed for further examinations including the assessment of AAMI. The criteria for AAMI were not clarified satisfactorily in either study, ^[14,15] eg, the Spanish study evaluated the prevalence of "amnestic syndrome of unclear etiology, which is best categorized as AAMI".

The present study is the first epidemiologic evaluation of the prevalence rate of AAMI defined according to the criteria of the NIMH Work Group ^[5]. It showed a high prevalence rate of AAMI (38.4%) in the age group 60 to 78 years. Thus, AAMI seems to characterize a phenomenon of normal aging rather than any clinical entity or generally accepted diagnosis. In this respect, we share the critical opinion of many researchers regarding the usefulness of the AAMI concept in clinical practice ^[7,9-11,22] and in the research of aging ^[8].

Brayne and Calloway ^[12] postulated that AAMI and AD could form a continuum, but no clinical or neuropathologic follow-up studies of AAMI subjects are available to support this. In a longitudinal study of elderly community residents, 95% of the individuals who were in the "forgetfulness phase" according to the Global Deterioration Scale (GDS) remained clinically unchanged during the mean follow-up of 3.6 years ^[23]. A possible connection between AAMI and AD could be evaluated by comparing the prevalence and incidence rates of these entities in the same study population, because the prevalence of AD varies markedly in different populations ^[24]. Although such studies with large populations are not available, our study showed that the epidemiology of AAMI is not the same as the epidemiology of AD. The overall pattern of age and sex differences in AD is consistent in recent studies: the prevalence of AD increases exponentially with advancing age, and is higher in women than in men ^[25-27].

The NIMH Work Group presented their proposed diagnostic criteria for AAMI to "facilitate communication among investigators and stimulate research into the epidemiology, clinical characterization, and treatment of later-life memory loss" ^[5]. However, the AAMI concept per se ^[9,10,22] and all essential parts of the proposed criteria including subjective memory complaints, ^[11,28] objective memory tests, ^[7,11] intellectual capacity testing, ^[11] exclusion of dementia, ^[11] and medical exclusion criteria ^[7,11] have been criticized.

The subjective estimation of memory loss can be affected by other factors in addition to actual memory deficit, such as the affective state and personality traits ^[28-30]. Supporting this finding, we reported recently that subjective memory impairment correlates more closely with personality traits than with actual memory performance ^[31]. Perhaps a more detailed questionnaire or a questionnaire also addressed to a close relative would better differentiate subjects having age-associated memory disturbances from those having no memory impairment. Indeed, relatives' ratings of a subject's memory have been recently shown to correlate better with objective memory scores than the subject's own assessment of memory problems ^[32].

The prevalence rates for AAMI can vary widely, depending on which objective memory tests and cutoff points are used ^[7]. Our results are in agreement with such findings. The BVRT was the most inclusive measurement (78.4% in subpopulation 1, 67.4% in subpopulation 2, and 70.9% in subpopulation 3), whereas the WMS-AL was the least inclusive (43.6% in subpopulation 1, 31.9% in subpopulation 2, and 36.7% in subpopulation 3). This may be explained by the impairment of visual memory before verbal association learning memory with aging or, on the other hand, by unsuitability of the proposed cutoff points for diagnosing AAMI.

The prevalence of AAMI also depends on intellectual capacity. We applied the WAIS-V only in subpopulation 3 because a validated Finnish translation of the test was not available before 1992. A cutoff score of at least 9 (raw score of at least 32) has been estimated to exclude 37% of the lower end of the population continuum ^[11]. In our population, using the same cutoff score, 42.3% had intellectual function inadequate for assessment for AAMI. We agree with Blackford and La Rue ^[11] that the proposed WAIS-V cutoff score provides a potential skewing of the subject population.

Absence of dementia is essential for the diagnosis of AAMI. By itself, an MMSE cutoff score equal to or less than 24 is an inadequate method for the assessment of dementia if no other, more accurate methods for dementia diagnosis are used. Because the score in the MMSE correlates positively with education and inversely with age, a cutoff score of 23 or less will overestimate the prevalence of dementia in less-educated normal older adults and underestimate the prevalence in younger, more- educated subjects ^[21,33-35]. Indeed, Folstein et al ^[36] found a false-positive ratio as high as one-fourth by using a cutoff point of the total score of 23 in comprehensively examined subjects. Compared with the MMSE alone, a more accurate method for assessing dementia is to use widely accepted criteria for dementia, such as the DSM-III-R ^[37] or the NINCDS-ADRDA ^[38] criteria.

Finally, the AAMI medical exclusion criteria may be too restrictive and thus exclude persons whose medical conditions do not affect their memory performance ^[11]. Smith et al ^[7] reported exclusion rates ranging from 7.5 to 19.2% when they applied the medical exclusion criteria to two samples of community-dwelling, healthy elderly persons. In the present study, 15.7% (63/402 in subpopulation 1) to 15.9% (75/471 in subpopulation 3) were excluded because of medical conditions. The number of excluded subjects could probably have been even higher by using stricter methods than the medical history alone to detect diseases that can cause a decline in memory.