Body mass index, lifestyles, physical performance and cognitive decline: The “Treviso Longeva (Trelong)” study

Abstract

Objectives

The relative contributions of risk factors, as body mass index (BMI), depression, chronic diseases, smoking, and lifestyles (as physical and performance activity, social contacts and reading habit) to cognitive decline in the elderly are unclear. We explored these variables in relation to 7-year cognitive decline in long-lived Italian elderly.

Design

Secondary data analysis of a longitudinal study of a representative, age-stratified, population sample.

Setting

The TREVISO LONGEVA (TRELONG) Study, in Treviso, Italy.

Participants

120 men and 189 women, age 77 years and older (mean age 80.2 ± 6.9 years) survivors after seven years of follow up.

Measurements

Cognitive decline measured as difference between Mini-Mental State Examination (MMSE) score in 2003 and in 2010; Body mass index (BMI), handgrip, Short Physical Performance Battery (SPPB) score, social contacts, reading habit, sight, hearing, schooling, mediterranean diet and multiple clinical and survey data recorded at baseline in 2003.

Results

In separate univariate analyses, age, SPPB score < 5, depressive symptoms (GDS) and more comorbidities (CCI) were associated with greater cognitive decline. Otherwise higher BMI, higher handgrip, reading habit, non-deteriorated sight and hearing, and schooling were protective. In a final multivariate model, age and higher BMI were associated with greater cognitive decline while reading habits was protective. SPPB score < 5 tends, though weakly, to be associated with greater cognitive decline. These associations remained with multivariate adjustment for gender, schooling, Charlson co-morbidity index (CCI) and baseline MMSE.

Conclusion

Age and higher baseline BMI, independent of gender, and other confounding factors, are risk factors for cognitive decline. Reading habit plays a protective role seven years later among northern Italian adults aged 70 years or older. Low physical performance tends, though weakly, to be associated with greater cognitive decline.

Objectives

Currently about twenty-four million people in the world have dementia and this amount will double every 20 years to 42 million by 2020 and to more than 80 million by 2040, assuming no changes in mortality, and no effective prevention strategies or curative treatments (1). According to the Global Burden of Disease estimates from the 2003 World Health Report, dementia contributed 11.2% of years lived with disability in people aged 60 years and older, more than stroke (9.5%), musculoskeletal disorders (8.9%), cardiovascular disease (5.0%) and all forms of cancer (2.4%) (2). In recent years, a growing body of evidence suggests that vascular risk factors and lifestyles may affect the risk of dementia in old age. Being overweight during midlife and late-life has been related to dementia (3., 4., 5., 6.). Differently some studies have not shown this relationship (7), or reported late-life decline in body mass index (BMI) prior to dementia onset (8., 9., 10., 11., 12.). Differences in study designs and in lengths of observation periods contribute to make unclear the relationship between BMI and cognitive status. Some studies provide evidence for the potential role of physical exercise in promoting cognitive health later in life (13., 14., 15., 16.). We examined the change in cognitive status measured between baseline in 2003 and follow up in 2010 and its relationships with baseline measures of BMI, lifestyles, physical functioning and other factors in the “TREVISO LONGEVA” (TRELONG) Study, a prospective cohort study of women and men, aged 70 years and older in Northern Italy. The TRELONG Study was realized to study the human aging in an area, such as Treviso, where the highest longevity in Italy, especially among women, is observed (17).

Methods

The TRELONG Study, was initiated in 2003. Eligible participants lived in the municipality of Treviso in Northeast Italy. At the beginning of this study in 2003 Treviso had a resident population of 81,700, of whom 13,861 (17%) were over the age of 70 years. Participants were systematically sampled from the list of residents of the Registry Office of

Treviso, based on an initial plan to include 100 participants per sex and 10 year age group, with 125 women and 125 men aged 70-79 years and all of those over 100 years be included. Thus 670 adults were eligible. Of these, 668 participated (99.7% response rate), 311 men and 357 women, aged 70 years and older (mean age 84 ± 8 years, range 70-105.5 years). An interviewer-administered questionnaire and a blood sample were collected at participants’ homes (18). Information on cognitive performances was also obtained from caregivers. Baseline characteristics of this study population and methodological details have been published previously, and form the basis of these analyses (18). The study protocol was approved by the Ethical Committee of the National Institute on Research and Care of the Elderly (INRCA, Italy). All participants and/or their caregivers provided written consent. A follow up interviewer-administered questionnaire and a blood sample were collected at home from the 309 survivor participants in March 2010. The only follow-up data used in the current study is the change of MMSE score between 2003 and 2010. In Table 1 the main baseline (2003) data of 309 survivors at follow up in 2010 are reported.

Table 1.

Baseline characteristics of survivors at follow up in 2010

Variable	Median (interquartile range) or n (%)	Number
MMSE (2003)	27.70 (2.75)	309
Age (years)	79.10 (9.65)	309
Sex= F	189 F (60%) 25.3 (5.1)	309
BMI (Kg/m2)		298
Handgrip average (Kg)	15.0 (12.3)	309
SPPB score	8.0 (5.0)	309
GDS score	1.0 (4.0)	303
Socialization (yes/no)	166 y (53.7%)	309
Reading (yes/no)	259 y (83.8%)	309
Sight (yes/no)	224 y (72%)	309
Hearing (yes/no)	212 y (75%)	281
Schooling (years)	6.6 (5.1)	308
Mediterranean diet (yes/no)	284 y (92%)	309
Albumin (gr/dl)	4.3 (0.3)	272
Cerebrovascular disease & Stroke	30 y (9.7%)	309
CCI	4.9 (1.7)	309
Smoking (yes/no)	122 y (39.4%)	309

MMSE=Mini Mental State Examination; BMI= Body Mass Index; SPPB=Short Physical Performance Battery; GDS=Geriatric Depression Scale; CCI=Charlson Co-morbidity Index

Global cognition measure

Global cognition was measured using the 30-point mini mental state examination (MMSE) (19) at baseline in 2003 and at 2010 follow up. Raw scores were corrected for age and education (20).

Anthropometric Measures

Body weight and body height were measured (18). Body mass Index (BMI) was calculated as kg/m².

Physical performance measures

The Short Physical Performance Battery (SPPB) was used to measure standing balance, walking, and chair stand tests. Standing balance tests included tandem, semitandem and side-by-side stands. To test walking speed, each participant was timed for two walks on an 8-foot walking course. The faster of the two walks was used. The chair stand test assesses the ability to rise from a chair five times as quickly as possible. Performance categories were created for each set of performance measures to permit analyses that included those unable to perform a task, using criteria already established and used in the literature (21., 22.). A summary performance scale was created by adding up the category scores for walking, chair stand and balance tests (21). Physical strength was measured via handgrip using a Collins dynamometer (Witte GMBH Chirurgische Instrumente, Adult Size, 12.7 x 5.7 cm, Solingen, Germany) and taking the average of three assessments with the stronger hand (18, 22).

Biochemical Measures

A fasting morning blood sample was collected at home from each participant and transported to the Clinical Chemistry

Laboratory of Treviso Hospital, within 30 minutes. Routine hematological and clinical chemistry tests were performed immediately using standard laboratory methods. Serum biochemical markers, such as albumin, were determined by commercial methods (Roche Diagnostics GMBH, D-68298 Mannheim Germany) on a Modular Analyzer (18).

Assessment of chronic disease

In-person interviews and examination of medical records provided information on history of chronic diseases as cerebrovascular disease, chronic renal insufficiency, tumors and others (18, 22). Using this information, a Charlson co-morbidity index (CCI) was determined (23., 24., 25., 26., 27.). The CCI contains 19 categories of pathologies, which were defined using ICD-9-CM (2). The categories are: acute myocardial infarction, congestive heart failure, peripheral vascular disease, cerebrovascular disease, dementia, chronic obstructive pulmonary disease, connective tissue disease, gastric ulcer, mild liver insufficiency, diabetes, hemiplegia, moderate-to-severe renal insufficiency, diabetes with end organ damage, solid tumors, leukemia, lymphomas, medium-to-severe liver insufficiency, tumor metastases, acquired immune deficiency syndrome (AIDS), and age. Each category is weighed according to the adjusted risk of 1-year mortality. The overall co-morbidity score is related to the likelihood of mortality within one year; the higher the score, the more severe the co-morbidity burden (23).

Depressive symptoms were assessed using the geriatric depression scale (GDS) (28, 29). Sight and hearing were also considered. The sight was classified as: (a) normal, (b) slightly impaired, (c) greatly impaired, and (d) blind, asking the subjects and showing them colored tables. Hearing was determined to be: (a) normal, (b) slightly impaired, (c) greatly impaired, and (d) total deafness. The interviewer pronounced the sentence “Today is a sunny/rainy day” slowly and softly from a 4 meters distance. In the analyses presented in this paper, we considered sight and hearing categories (a) versus (b-d) (18, 22).

Lifestyles measures

Lifestyle measures included physical activity, socialization, reading, and smoking. To assess physical activity, participants were asked whether they took walks and/or did gardening every day. In the current study, we considered physical activity categories “at least one activity” versus “none”. Socialization was assessed via: (a) regular visits to friends, (b) participation in recreation and social centres, (c) involvement in socially useful activities, or (d) volunteer work. Cultural activities included regular reading activity (18, 30). Each participant reported if he smoked. Number of years of education was recorded. Dietary patterns were also assessed, particularly intake of the Mediterranean diet: cereals, fish, vegetables and fruit (31., 32., 33.). It was asked if the cereals were usually eaten at least once a day, vegetables and fruit at least twice a day, and fish at least twice a week.

Statistical analysis

Participants were initially characterized on a variety of clinical and surveyed variables. Medians and interquartile ranges, percentiles and percentages were calculated. Moreover an univariate linear regression on each covariate that was taken into account was calculated. The multivariate statistical analysis consists of a logistic regression on the dichotomic variable taking the value 1 if the individual underwent between 2003 and 2010 a MMSE decline greater than its standard deviation (5.9) and zero otherwise. Information on MMSE decline was available only for participants who survived up to 2010, thus individuals who died before 2010 have been disregarded. No data were inputted and missing data were removed. Participants were stratified on the basis of SPPB score (<5 vs. ≥5), schooling (≥8 years vs. ≤8 years), albumin (<4 gr/dl vs. ≥4 gr/dl). We also examined possible interaction effects between covariates. In the final multivariate model we kept the interaction between BMI and age. The final model is shown in Table 3. This is the model that ensures the best fit with data (in terms of residual deviance) with a relatively low number of independent variables. Statistical analyses were conducted using the R software (34). Results were considered significant at a level of p<0.05. However, results significant at 0.10 level are not disregarded, considering the low sample size, which makes the statistical tests less powerful.

Table 3.

Multivariate model of variables considered with regard to cognitive decline between 2003 and 2010

Variable			Confidence interval
	Coefficient	Std. Error	2.5%	97.5%	Wald x	p-value
					2 statistic
MMSE (2003)	0.064	0.056	−0.044	0.176	1.341	0.247
Age	0.476	0.152	0.182	0.781	9.828	0.002 **
Sex= M	−0.063	0.338	−0.733	0.600	0.035	0.852
BMI	1.049	0.480	0.104	1.998	4.779	0.029 *
SPPB score < 5	0.678	0.409	−0.132	1.479	2.749	0.097 .
Reading (y/n)	−0.902	0.449	−1.783	−0.012	4.028	0.045 *
CCI	−0.071	0.110	−0.292	0.143	0.416	0.519
Schooling ≥ 8 ys	−0.605	0.376	−1.371	0.111	2.595	0.107
BMI*Age	−0.014	0.006	−0.026	−0.002	5.410	0.020 *

A positive effect indicates that the independent variable is considered a risk factor (increased cognitive impairment) and a negative effect indicates that the variable is a protective factor. Signif. codes: 0 ‘ *** ’ 0.001 ‘

^**’ 0.01 ‘**’ 0.05 ‘ ’ 0.1 ‘ ’ 1 MMSE=Mini Mental State Examination; BMI= Body Mass Index; SPPB=Short Physical Performance Battery; CCI=Charlson Co-morbidity Index

Results

The sample was composed of 309 survivors after seven years of follow, 120 men and 189 women, age 77 years and older (age median 79.1 years). Participants were just slightly overweight (BMI median 25.3) and had a moderate degree of physical functioning (SPPB median score 8.0). They showed a good inclination to read (83.8%), good adherence to mediterranean diet (92%), moderate socialization (53.7%) and no high tendency to smoke (39.4%). Depressive symptoms were not relevant (GDS median 1.0) (Table 1).

An average decline by about three MMSE points during follow up, was found. The distribution of MMSE follow up scores was much more variable and “moved” to lower scores on the worst cognitive profiles. In 2003 half of the sample had a score between 25.70 and 28.45, in 2010 the same interquartile range was between 21.60 and 28.70 (Figure 1).

Figure 1.

Box-plot of MMSE score in 2003, in 2010 and of MMSE decline

Univariate analysis (Table 2) showed that age, low SPPB score, depressive symptoms (GDS) and more comorbidities (CCI) were significant and positive. Whereas BMI, handgrip, reading habit, preserved sight and hearing, schooling and also, though weakly, socialization were significant and negative. BMI coefficient was low and the true role of BMI was revealed by the multivariate analysis. Also the correlation between MMSE decline and BMI change over time is modest (-0.12), albeit significant (p-value=0.041) and reveals that increasing BMI is negatively associated with cognitive decline.

Table 2.

Univariate analysis of variables considered with regard to cognitive decline between 2003 and 2010

Variable	Coefficient	Std. Error	p-value
MMSE (2003)	−0.006	0.037	0.881
Age	0.135	0.022	0.000	***
Sex= M	−0.472	0.283	0.095
BMI	−0.110	0.037	0.003	**
Handgrip (average)	−0.061	0.016	0.000	***
SPPB score < 5	1.0895	0.315	0.001	***
HDL cholesterol	0.003	0.011	0.754
GDS	0.095	0.041	0.021	*
Socialization (y/n)	−0.479	0.270	0.076
Reading (y/n)	−0.860	0.349	0.014	*
Sight (y/n)	−0.870	0.296	0.003	**
Hearing (y/n)	−0.794	0.307	0.010	**
Schooling ≥ 8 ys	−0.737	0.322	0.022	*
Mediterranean diet	0.205	0.666	0.758
Albumin < 4 gr/dl	0.400	0.458	0.383
Cerebrovascular disease	0.583	0.440	0.185
& Stroke
CCI	0.212	0.079	0.008	**
Smoking (y/n)	−0.256	0.277	0.355
Diabetes (y/n)	0.191	0.404	0.637
Hypertension (y/n)	−0.140	0.274	0.610

A positive effect indicates that the independent variable is considered a risk factor (increased cognitive impairment) and a negative effect indicates that the variable is a protective factor. Signif. codes: 0 ‘

^***’ 0.001 ‘

^**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1. MMSE=Mini Mental State Examination; BMI= Body Mass Index; SPPB=Short Physical Performance Battery; GDS=Geriatric Depression Scale; CCI=Charlson Co-morbidity Index

In the multivariate model, the BMI was a risk factor as it was directly related to the extent of cognitive impairment while the reading habit appeared as protective; age maintained a positive correlation with cognitive impairment. Physical impairment (SPPB score < 5) tended, though weakly, to play as risk factor for cognitive decline.

The effects of BMI should be also read along with the interaction that this variable had with age. The interaction of BMI with age was significant and negative. This means that the positive effect of the BMI decreased as the age of the individual increased.

Discussion

In the current study age is a risk factor for cognitive impairment, according to what is already known and reported (35., 36., 37., 38., 39., 40., 41., 42., 43., 44.). In fact, aging is generally accompanied by modification of the cognitive profile (35). Many older people may experience difficulties in various cognitive domains such as memory (36), attention (37), executive functions (38, 39), and processing speed (40) and consequently, a loss of autonomy in activities of daily living (41., 42., 43.).

There have been several prospective or nested epidemiologic reports (3-6, 8, 10, 11, 45-48) evaluating BMI in relationship to dementia. Most (3-6, 45, 46, 48) have shown a high BMI to be a risk factor for dementia when measured at least a decade prior to cognitive decline onset. Other longitudinal studies describe that a high BMI has a protective effect when measured within five years before the onset of dementia: Luchsinger at al. showed that dementia risk decreased with increasing BMI in older people over 76 years (49) and Buchman at al. reported that declining body mass index is associated with increased risk of incident Alzheimer disease (AD) (50). Many cross-sectional reports underlined that low BMI or underweight are related to dementia (51., 52., 53., 54., 55.). In our analysis higher BMI score is a risk factor for the occurrence of cognitive decline over seven years. This relationship remained after potential confounders were considered.

These findings demonstrate that the relationship between overweight and cognition is complex, dynamic and a function of time. Obesity accompanies and even promotes survival (56., 57., 58.), but then becomes a risk factor for dementia among those who are susceptible due to survivorship to high age (59). The interplay of adiposity and the brain occurring over the course of the lifespan is complex and in relationship to developmental origins, midlife comorbidities, disruptions in brain structure and cognition (59, 60). The association between obesity and dementia is suggested by the central role of obesity in the development of hypertension, dyslipidemia, diabetes and metabolic syndrome. In the Honolulu-Asia Aging Study a higher cardiovascular risk factor burden in middle age increased the risk of vascular dementia 25 years later (61). Brain atrophy, an early marker of cognitive decline, is a manifestation of neuronal degeneration (62, 63) and has been related, using computed tomography (CT), to higher BMI levels 24 years before (64). Elevated BMI is associated with reduced global brain volumes, in a cross-sectional study of women and men aged 40-66 years (65). A number of studies demonstrate that persons with AD have lower weight and BMI (54, 66., 67., 68., 69.) and lose body mass (70, 71). This may result from a variety of factors including changes in eating habits and access to adequate nutrition due to memory impairment (72., 73., 74.), changes in taste and smell, solitude and depression. Evidences in literature suggest that involuntary weight loss (IWL) is more a consequence than a cause of dementia, in particular of AD, but chronic diseases and disability, inadequately treated from the nutritional standpoint, may also worsen and accelerate cognitive impairment (75). Obesity and high BMI values dynamically play the role of long-term risk factors and protective factors in the short-term, when the degenerative or vascular brain lesions cause cognitive decline and also induce loss of memory and appetite, apathy, disinterest, depression and thus weight loss. Being overweight is a risk factor for cardiovascular disease in adults, along with hypertension, metabolic syndrome and diabetes. In people who survive beyond 70 years of age, overweight plays a protective role with respect to mortality, but contributes to the onset of cognitive impairment.

As we know, there are not many studies that have evaluated the effect of the habit of reading in older people towards the changes in the cognitive profile. Uchida et al. introduced a new cognitive intervention program for normal aged people, and demonstrated that daily training program involving reading and arithmetic problems was effective for the improvement of cognitive functions of community-dwelling elderly (76).

Nouchi et al. used simple training tasks (reading aloud and simple calculation) to reveal the beneficial effects of learning therapy on a wide range of cognitive functions in elderly people (77). Jefferson et al. showed cross-sectional findings suggesting that education and reading ability are the most-robust proxy measures of cognitive reserve in relation to late-life cognition (78). Our previous cross-sectional work showed that reading habit is associated with better cognitive performance; this correlation remained significant after adjusting for confounding factors, such as comorbidity and hearing function (30). In the current study reading habit shows itself as a protective factor against cognitive decline, confirming the correlation highlighted in our previous cross-sectional work (30). Public health programs that promote the diffusion of habit to read, even in old age, could reduce the incidence of cognitive impairment.

The ability to maintain independence and quality of life in older people depends significantly on the maintenance of cognitive and physical performance but the association between cognitive and physical performance and their rate of change with age is unclear. Some evidences indicate that motor and cognitive systems are both likely to be influenced by processes, both developmental and degenerative, which regulate CNS function; the suggestion that there are developmental and ageing processes which influence both cognitive and physical performance has been termed the “common cause” hypothesis (79, 80). According to this hypothesis is the evidence that both cognitive and physical performance can be negatively influenced in parallel by factors such as chronic disease, sedentary lifestyles and poor socioeconomic conditions (81., 82., 83., 84., 85.). Atkinson et al. found that baseline global cognitive function and global cognitive decline independently predicted declines in specific physical measures but that none of their physical measures predicted cognitive change; these findings support their hypothesis that cognitive function decrements in older adults on average precede or co-occur with physical performance declines (86). On the contrary many studies have demonstrated that baseline physical performance predicts declines in cognitive function and/or incident cognitive impairment or dementia (13-16, 87-98). The current study is in line with these findings, in fact good physical performance aims to show itself, though weakly, as a protective factor against cognitive decline. In particular the SPPB, as well as being a predictor of the degree of functional autonomy and disability (99), might be a marker of cognitive status.

Conclusions

Age affects the extent of cognitive decline after seven years. A high BMI, at advanced age, may play a protective role regarding the mortality, but is a risk factor for cognitive decline. The reading habit shows itself as a protective factor against cognitive decline. Good physical performance aims to show itself, though weakly, as a protective factor against cognitive decline.

This study has some limitations. Firsty, the magnitude of the decline occurred in the population observed from 2003 to 2010 is contained in 2.86 points of MMSE score. It is possible that a greater decline would have allowed the identification of greater linkages between potential risk factors and loss of cognition. Such a low cognitive decline is possibly due also to a re-test effect: it has been showed that people who are asked to compile any cognitive test have a better performance if the same test is submitted even long after the first one (100, 101). In addition, the MMSE test could have been submitted to respondents also in other settings not depending on the survey in the years between cross-sectional and follow up. Secondly, the sample size is limited and this makes the statistical tests less powerful. The third limitation is that some risk factors for cognitive decline, as diabetes (102) and hypertension (103), are not statistically significant in our analysis. Perhaps this is again due to the low sample size.

However, there are also several strengths. Firstly, this is a systematic sample of adults aged 70 years and older from one of the longest lived communities in Italy and the sample is highly representative of the community. Secondly, the sample observed is residing at home and was drawn at random from the list of residents of the municipality of Treviso: the study sample is therefore representative of the real world. In fact many studies are being conducted in selected samples: patients, only men or only women, etc. The third strength is that the study sample was well characterized by a large amount of data that was in part set out in previous published works (18, 22, 30). Furthermore measurement of cognitive impairment using the MMSE score represents a simple and homogeneous tool; often the diagnosis of dementia is underestimated and does not take into account the presence of mild cognitive impairment. Finally, some studies involving cognition and BMI do not consider physical performance; we have also considered this

parameter among the potential factors related to cognitive impairment. Additional years of observation will further confirm these findings.

Conflict of interest statement

None.

‘No Disclosures to Report’