Abstract
Keywords
INTRODUCTION
The conceptual validity and utility of an early diagnosis of cognitive impairment rely on the hypothesis that preclinical cognitive changes may predict an increased risk of dementia, allowing the selection of subgroups as possible targets for preventive or therapeutic interventions.
Recent mild cognitive impairment (MCI) criteria include concern over a change in cognition, impairment in one or more cognitive domains, and preservation of independence in functional abilities, with mild problems in complex functional tasks [1]. The criterion of independence in functional abilities was defined as “challenging”. Very mild problems in instrumental activities of daily living (IADL) are generally consistent with MCI, while basic activities of daily living (ADL) should be preserved. Functional evaluation has a central role also in the diagnosis of minor neurocognitive disorder as defined in the 5th Edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) [2].
We have previously evaluated the frequency and 4-year progression to dementia of MCI and Cognitive Impairment, no Dementia (CIND) in the Italian Longitudinal Study on Aging (ILSA) [3]. CIND diagnosis was used in the Canadian Study of Health and Aging for individuals older than 65 years whose problems with memory or other areas of cognition were not sufficiently severe to meet the criteria for dementia, but who were judged to be distinct from those cognitively normal of the same ages [4].
In our original cohort, ADL or IADL impairment at baseline was present in MCI, CIND, and even in subjects with normal cognition. In elderly people, causes other than cognition may impact everyday function, requiring controlling for a series of comorbidities [5]. In this context, the contribution of motor performance (MP) may also require attention [6].
The present work extends the follow-up of the ILSA cohort, aiming to evaluate whether functional changes in ADL, IADL, or MP may have a role in predicting progression to dementia in subjects with normal cognition, CIND or MCI at baseline.
METHODS
Study design
This work is part of the ILSA, a large population-based survey described elsewhere [3, 7]. The ILSA aimed at evaluating age-related functional changes of cardiovascular and nervous systems, estimating, in a cohort of older Italians, prevalence and incidence of dementia, stroke, parkinsonism, peripheral neuropathy, angina pectoris, myocardial infarction, cardiac arrhythmia, heart failure, hypertension, peripheral artery disease, and diabetes. A random sample of 5,632 individuals aged 65–84 years, community-dwelling or institutionalized, stratified by 5-year age groups (65–69, 70–74, 75–79, 80–84 years) and gender, was extracted from population registers of eight municipalities across Northern, Central and Southern Italy, in urban and rural areas [8]. The ILSA included cross-sectional and longitudinal components. At baseline, cases were identified through a two-phase design. In Phase 1, involving all participants, a personal interview determined demographics, education (completed years of schooling), risk factors, and disease symptoms; a physical and neuropsychological examination, carried out by specifically trained physicians, and laboratory investigations were also performed. In Phase 2, a geriatrician or a neurologist confirmed diagnoses in suspected cases through direct examination and review of medical records.
The study cohort was identified within the ILSA population after the first cross-sectional study, carried out in 1992-93, that led to the detection of the prevalent cases of cognitive impairment and dementia (baseline survey) [3]. Prevalent cases of dementia were excluded from the present analysis. Starting on September 1, 1995, and on May 1, 2000, participants were re-examined following the same procedures of baseline examination [9]. Incident cases of dementia were individuals free of dementia at baseline, but who developed dementia during the follow-up.
Cognitive impairment and dementia
The diagnosis of cognitive impairment was based on a neuropsychological battery including the Mini-Mental State Examination (MMSE) [10], the Babcock Story Recall Test (BSRT) [11, 12], and the Digit Cancellation Test (DCT) [12]. The MMSE was used to evaluate general cognitive function by assessing orientation, immediate and delayed verbal memory, attention and calculation, constructional praxis, and language. To control for age and education on MMSE scores, our study subjects were stratified in four age groups (65–69, 70–74, 75–79, 80–84 years) and three education levels (≤5, 6–10, 11+ years), with 12 strata defined by the combinations of these variables. Educational levels were chosen considering the Italian educational system in the period of school attendance for the study population. For each group, mean and standard deviation (SD) of MMSE scores were calculated. According to the method proposed in the Kungsholmen Project [13], CIND was diagnosed in individuals scoring >1 SD below the age- and education-specific mean of the MMSE.
Memory was assessed using the BSRT. The subjects were asked to immediately recall a 21-unit story just read to them, then, after the story was read again, to recall it 10 minutes later. Score ranged from 0 to 16. An event-weighted, hierarchical scoring system, was used to reward the degree of organization of oral recollection by participating subjects. Selective attention was evaluated with the DCT (score ranging from 0 to 60). The subjects were asked, in a set time (limit: 45 seconds/matrix), to cross out target digits in three different and increasingly difficult matrices, made up of 13 strings of 10 digits (0 to 9 in random sequence). Each line included from 0 to 5 targets. Again, 12 different age- and education-specific strata were defined in our study population for the BSRT and DCT. Subjects scoring >1 SD below the age- and education-specific mean only in MMSE and with preserved attention and memory were diagnosed as CIND MMSE-defined. Subjects with CIND scoring >1 SD below the age- and education-specific mean only in BSRT and preserved attention were defined as CIND with predominantly memory deficits (amnestic-CIND) and those scoring >1 SD below the age- and education-specific mean only in DCT and preserved memory were defined as subjects with predominantly attentive (single non-memory CIND) deficits. Subjects with CIND scoring >1 SD below the age- and education-specific mean on both the BSRT and the DCT were considered to have multi-domain CIND.
Subjects not scoring >1 SD below the age- and education-specific mean of the respective stratum at the MMSE were considered to have nonimpaired general cognitive function and possible candidates for the MCI diagnosis. We identified three categories among these subjects with normal global cognitive function: individuals scoring >1 SD below the age- and education-specific mean only on the BSRT and preserved attention were defined as MCI subjects with predominantly memory deficit (amnestic MCI). Subjects scoring >1 SD below the age- and education-specific mean only on the DCT and preserved memory were defined as MCI subjects with predominantly attentive deficit (single non-memory MCI); subjects scoring >1 SD below the age- and education-specific mean on both the BSRT and the DCT were considered to have multidomain MCI. The remaining subjects were defined with normal cognition at baseline.
The diagnosis of dementia required a clinical assessment performed by trained neurologists, based on a neuropsychological battery including the sections B and H of the Cambridge Mental Disorders of the Elderly Examination [14], the Pfeffer Functional Activities Questionnaire [15], the Hamilton Depression Rating Scale [16], a complete neurological examination, and review of clinical records. The final diagnoses had to meet the DSM-III-R [17] or DSM-IV [18] criteria for dementia syndrome, the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association criteria for possible and probable Alzheimer’s disease (AD) [19], and the International Classification of Disease, 10th edition criteria for vascular dementia (VaD) and other dementing diseases [20]. This methodology was maintained across the different steps of the study surveys. Before the survey, clinical investigators participated in an interrater agreement study on the application of diagnostic criteria. The reproducibility of the clinical diagnoses proved substantial (kappa index = 0.82 for dementia syndrome, 0.80 for the diagnosis of AD, and 0.71 for the diagnosis of VaD) [9].
Information for individuals deceased before the two follow-up examinations was gathered by clinical investigators from general practitioners (GPs) and relatives with interviews on diagnosis of dementia prior to death, drug prescriptions, comorbidities, hospitalization, and institutionalization. Hospital records and death certificates were checked. With the same procedure, information on deceased individuals was also obtained throughout an interim follow-up performed after the baseline survey.
To increase diagnostic reliability across centers, and between living and deceased persons, all putative cases were independently reviewed by an adjudication panel of senior neurologists, blind to the final diagnosis made by centers. In case of disagreement, the diagnosis by the panel replaced the original diagnosis.
Study variables
Final diagnoses of diseases and risk factors were made by the ILSA specialists, using direct clinical evaluation, medical records, study data, and standard criteria, as detailed elsewhere [3].
Functional independence was evaluated using the index of ADL [21] and the IADL Scale [22]. The pertaining information was obtained by the study physicians at phase 1 through direct or proxy questioning. For ADL, questions aimed at the evaluation of the functional independence or dependence of subjects in bathing, dressing, going to toilet, transferring, continence, and feeding. IADLs refer to more sophisticated tasks of everyday life, such as using the telephone, shopping, preparing meals, doing housework or handyman work, laundry, transportation, taking medications, and managing money. The capacities of an individual to perform with or without supervision, direction or active personal assistance were investigated. The “not applicable” option was used for tasks never performed by subjects (e.g., doing laundry for men). The degree of dependence was categorized as none, or dependence in one, 2-3 or >3 for ADL, and none, or dependence in one, 2–4 or >4 for IADL.
MP was assessed using six tests from a battery measuring motor ability in older people [23]. Three texts explored dynamic balance and coordination: time to stand from a chair unaided and without using the arms, number of times a subject could step up onto a single 23-cm step in 10 seconds, and number of errors in a tandem walk along a 2-m line (5-cm wide). Three other tests assessed static balance: time standing on one leg, number of steps and time spent to walk 5 m, and number of steps to complete a 180° turn. A total score, ranging from 0 (worst performance) to 14 (best performance), was obtained by summing up single items’ scores. MP score categorization was 0–9, 10–13 and 14. Kappa agreement between independent examiners rating total score was 0.80 [24].
The study was approved by the Ethics Committees of participating institutions. Informed consent was given according to institutional guidelines.
Statistical analysis
Incidence rates of dementia were calculated as number of new cases/number of person-years at risk. Person-years of follow-up for non-demented persons were calculated as the time between baseline screening and follow-up examination, or death. The mid-point of the interval was used to calculate person-years for incident cases of dementia if precise information on dementia onset was not available. The rate ratios and exact 95% confidence intervals (CI) were estimated on the basis of the Poisson distribution, and differences between functional groups for total sample, normal cognition, CIND, and MCI were compared using the chi-square test. For predictive analyses, the statistical procedure of Cox regression was used to test the independent effect of ADL, IADL, and MP on the risk of developing dementia in total sample, normal cognition, CIND and MCI. Cox proportional hazard model was adjusted for demographic variables (age, sex, and education level), and health conditions evaluated in the ILSA (angina, myocardial infarction, heart failure, atrial fibrillation, diabetes, hypertension, stroke, smoking, alcohol consumption). p-values < 0.05 were considered statistically significant. Analyses were performed using IBM-SPSS (Statistical Package for the Social Sciences), Version 20.0 (Armonk, NY: IBM Corp.).
RESULTS
Out of the original ILSA sample, the study cohort included 2,830 participants who completed the evaluation of cognitive status at baseline (Fig. 1). Overall, 264 participants were diagnosed with CIND and 445 with MCI, while 2,059 were considered cognitively normal. Sixty-two subjects were diagnosed as having dementia at baseline and were excluded from further analyses. Therefore, the 1992 cohort at risk for dementia consisted of 2,768 individuals (mean age 73.6±5.5 years, 53.6% men, 46.4% women). At the 1995 examination, follow-up procedures were completed by 1,992 (81.2%; mean age 72.9±5.3 years, 52.8% men, 47.2% women) of the 2,452 survivors. At the 2000 examination, 1,810 subjects (65.4%) were alive, and 1,403 (77.5%; mean age 72.1±5.0 years, 50.2% men, 49.8% women) completed follow-up procedures. Overall, follow-up data for survivors were available for 1,674 (92.5%; mean age 72.2±5.1 years, 48.6% men, 51.4% women); 1,267 individuals (75.7%) completed both follow-up procedures, while 271 subjects (16.2%) were evaluated only at the first follow-up examination in 1995, and 136 (8.1%) only at the 2000 examination.
A total of 316 subjects died before they were contacted in 1995 and 642 before they were contacted in 2000 (total deceased, n = 958). Information from previous examinations, GPs, proxy-respondents, medical records and death certificates was considered reliable to define the occurrence or not of dementia for 712 of deceased individuals (74.3%). The final data used were for 2,386 individuals (1674 survivors and 712 deceased with reliable information; mean age 73.4±5.5 years, 53.6% men, 46.4% women) for a total of 16,545 person-years. Those included 213 (80.7%) of the 264 participants with CIND and 361 (81.1%) of 445 participants with MCI at 1992 examination, and 1812 (88.0%) of 2059 subjects cognitively normal at baseline. At baseline, IADL information was available for all subjects, ADL for 99.4% and MP for 98.7%.
The mean follow-up time was 6.9±2.7 years. A total of 210 new dementia cases were identified. Table 1 indicates the number of cases and the sex-specific incidence rates (new cases per 1,000 person-years) of dementia at follow-up in the overall study population, in cognitively normal subjects, and in the different subtypes of CIND and MCI. Incidence of dementia (per 1,000 person-years) was 12.69 in the total sample, 9.86 in subjects with normal cognition at baseline, 22.99 in CIND and 21.43 in MCI. In CIND, the lowest rates were found in subjects with deficits only in MMSE and normal BSRT and DCT, and the highest in subjects with multi-domain CIND. Among MCI subtypes, the lowest incidence was in amnestic and the highest in multi-domain. Incidence of dementia in both CIND and MCI was higher in women, while incidence in subjects with normal cognition at baseline was higher in men.
Incidence of dementia was significantly higher when an increasing number of baseline ADL and IADL were impaired, and with a worse MP (Tables 2 and 3). In particular, the rate-ratio for incident dementia with increasing IADL impairment ranged from 2.36 (95% CI, 1.58–3.51) to 8.10 (95% CI, 5.12–12.60) in total sample, from 2.42 (95% CI, 1.48–3.91) to 5.64 (95% CI, 2.62–11.18) in cognitively normal subjects, from 3.02 (95% CI, 0.73–17.69) to 8.71 (95% CI, 2.09–50.97) in CIND and from 1.83 (95% CI, 0.67–4.59) to 8.21 (95% CI, 3.67–18.25) in MCI subjects.
In Cox regressions controlled for age, sex, education level, and major age-related conditions, baseline IADL impairment was the stronger predictor of dementia at follow-up, with hazard ratio (HR) ranging from 2.03 (95% CI, 1.34–3.07) to 4.41 (95% CI, 2.62–7.45) in total sample, from 5.32 (95% CI, 1.01–27.85) to 10.93 (95% CI, 1.85–64.69) in CIND and from 2.16 (95% CI, 0.82–5.70) to 9.57 (95% CI, 3.40–26.91) in MCI. The effect was less pronounced in cognitively normal subjects at baseline, with HR ranging from 1.69 (95% CI, 1.02–2.79) to 1.99 (95% CI, 0.90–4.38) (Table 4).
DISCUSSION
We described an 8-year incidence of dementia in a large population-based study representative of Italian elderly. Incidence rates were significantly higher in subjects with a diagnosis of CIND or MCI than in total sample or in those with normal cognition. At univariate analyses, ADL, IADL, and MP were significant in the prediction of dementia at follow-up in total sample, in subjects with normal cognition, in CIND (only IADL and MP), and MCI. At multivariate analyses, IADL were the most powerful predictor of transition to dementia in total sample and in MCI.
ADL should not be compromised in subjects with preclinical cognitive impairment. However, dealing with population studies is different than selecting patients in dementia clinics. Functional problems due to comorbidities are very frequent in the elderly, even with normal cognition, and difficulties in everyday activities are more frequent in non-demented subjects with mild cognitive deficits than in the general population [25]. Therefore, if the choice is to avoid “ideal” conditions, as are often those in clinical trials, it is essential to control for demographics and comorbidities potentially explaining the lower performance in ADL or in motor abilities.
While ADL deal with more “basic” self-care behaviors, as ambulating, bathing, toileting, dressing,and feeding, IADL refer to more sophisticated tasks, more demanding in terms of cognitive function, such as using the telephone, shopping, transportation, medications management, and handling money [21, 27].
Individuals with MCI perform IADL significantly worse than matched healthy controls [27, 28]. However, functional status is a multidimensional construct. The percentage of variance explained by cognition may be surprisingly modest, ranging from 21% to 23.6% [5, 29]. Improved prediction of IADL functioning may be obtained including in models demographic variables and chronic illness, which may account for another 25.4% of the variability [5]. In our survey, we had the opportunity to control, apart from demographics, for major age-related and disabling diseases, such as stroke, myocardial infarction, heart failure, and diabetes. Controlling for demographics and major age-related diseases reduced the significance of ADL and MP in predicting the development of dementia. This was not the case for IADL, confirming the validity of their role on the MCI construct. It has been suggested that IADL may include elements of cognition not captured by neuropsychological tests, emphasizing the need to better understand the cognitive process that facilitate IADL enactment in the community, and the transition from functional independence to dependence, increasing prediction of persons at greatest risk for future decline [27].
So far, no prospective study has evaluated, in the same population, the possible role of ADL, IADL, and MP as possible predictors of progression to dementia in MCI subjects, while even IADL were analyzed only in a few longitudinal studies. In the PAQUID cohort, during a 2-year follow-up, even a mild restriction of IADL increased progression to dementia in MCI subjects, indicating that inclusion of IADL impairment in the MCI criteria improves the prediction of dementia [30]. In the 3C Study, adjusting for possible confounders, deficit in IADL in MCI subjects increased by 2.2 times in men and 3.5 times in women the risk of dementia over a 4-year follow up [31]. In the German Study on Ageing, Cognition, and Dementia in Primary Care Patients, and in the Leipzig Longitudinal Study of the Aged, during 4.5 and 8 years of follow-up, respectively, MCI combined with IADL impairment was associated with a higher conversion rate to dementia and a significantly shorter time to diagnosis [32, 33].
Among shortcomings of our study we include the problem of attrition, which may be very relevant in longitudinal studies. Missing data may lead to an underestimation of frequency of cognitive impairment, and in the ILSA non-response increased with advanced age and lower education level [3]. On the other hand, our data derive from a nationally representative, population-based sample, randomly selected, including urban and rural population, community-dwelling, and institutionalized. Including institutionalized people at baseline increased representativeness of resident population, but may have influenced rates of progression to dementia. However, institutionalization was not frequent in Italy in 1992, and the percentage in our sample was <1%, so we are confident that the effect on incidence, if any, should be minimal. The use of incident cases arising within a general population allows a major generalizability of results. The mean sampling interval was of 6.9±2.7 years. Studies with long intervals between the cross-sectional examinations are at risk of missing a relevant number of patients. We tried to reduce this effect using multiple sources of information, also for deceased individuals. Although these included interviews with GPs and relatives, and evaluation of hospital records, death certificates, drug prescriptions, comorbidities, and institutionalization, we cannot exclude an effect on predictive variables and outcome. A particular aspect of our survey was the opportunity to consider the occurrence of major age-related diseases. The assessment of the different conditions in the ILSA relied on direct examination performed by trained neurologists or geriatricians, and not merely on self-reported information. Among shortcomings of the survey, we consider also the limited neuropsychological evaluation, encompassing only global cognition, memory and attention. This could have been more relevant in patients with MMSE-defined CIND, although MMSE was the only tool used for CIND diagnosis in the Kungsholmen Project [13].
In conclusion, we confirmed the validity of inclusion of IADL in the MCI construct to significantly improve the prediction of dementia, showing also a reduced utility of ADL and MP measures when major age-related diseases are controlled for. Functional status has an initially subtle but still measurable impact on everyday life, and more refined instruments and guidelines for its definition are required for future research.
Footnotes
ACKNOWLEDGMENTS
As part of the Targeted Project on Aging, the ILSA was supported by the Italian National Research Council (CNR) with grants to each research unit from 1991 to 1995. The study was then funded by the Italian Ministry of Health (D.L. 502/92, 1998) through the programs “Epidemiology of the Elderly” (Istituto Superiore di Sanità) and “Estimates of Health Needs of the Elderly” (Tuscany Region).
The authors thank Maria Elena Della Santa (Institute of Neuroscience, Italian National Research Council, Florence, Italy) for technical support in preparing this article.
