Diabetes and hypertension are associated with lowered cognitive performance among middle-aged Brazilian adults: cross-sectional analyses nested in the longitudinal Pró-Saúde study

ABSTRACT BACKGROUND: Cardiovascular risk factors are frequently associated with lowered cognitive performance among elderly people, but rarely among middle-aged adults. OBJECTIVES: To investigate associations between cardiovascular risk factors (age, physical inactivity, smoking, alcohol use, hypertension and diabetes) and lower cognitive performance among middle-aged (45-64 years) Brazilian adults. DESIGN AND SETTING: Cross-sectional study nested within the Pró-Saúde cohort. From 2,876 baseline study participants (1999), we randomly selected 488 participants and gave them validated and standardized cognitive tests (2012). METHODS: We used multiple linear and logistic regression analyses to detect associations of cardiovascular risk factors with crude scores in cognitive tests on memory (word test) and executive function (verbal fluency tests), and with overall cognitive performance scores, respectively. RESULTS: All cognitive test scores presented statistically significant inverse associations with age and direct associations with education. There was no association between lower cognitive performance and smoking or alcohol use. In both 1999 and 2012, after adjusting for sex, age and schooling, being physically active was inversely associated with lower performance regarding late memory. For individuals with diabetes in 1999, there was an association with lower performance regarding executive function, while there was a borderline association for those reporting it only in 2012. Having a diagnosis of hypertension since 1999 was associated with lower performance regarding both memory and executive functions, while reporting hypertension in 2012 was associated with lower performance regarding executive function. CONCLUSIONS: Aging, low schooling and cardiovascular risk factors may represent life course disadvantages associated with cognitive decline even among middle-aged Brazilian adults.

marked social inequalities and social and healthcare services that are insufficient to address the problems associated with an aging population. 9 It is likely that poor social conditions may be responsible for the premature onset and higher burden of dementia in Brazil. 9 Brazil has high incidence of diabetes, hypertension and dementia. We hypothesized that cardiovascular risk factors, including hypertension and diabetes, might be associated with lowered cognitive performance, not only in the elderly but also among middle-aged adults in Brazil.

OBJECTIVE
We investigated the hypothesis that cardiovascular risk factors might be associated with lowered cognitive performance among middle-aged adults in Brazil. Our subjects were Brazilian civil servants participating in the Pró-Saúde study, a cohort study in Rio de Janeiro, Brazil, that had the goal of evaluating several dimensions of health-related determinants. 10

METHODS
We conducted this cross-sectional study nested within the Pró-Saúde cohort. 10 From 2,876 participants involved in the baseline studies in 1999, a subsample was randomly selected from all strata: both sexes, two age groups (less than 50 years versus 50 years or over) and two educational levels (less than high school versus high school or more). There was no difference in clinical and sociodemographic characteristics between the participants in this study sample and those of the baseline population (data not shown).
In 2012, cognitive tests were given to 488 participants aged 35-64 years. We analyzed their cognitive performance according to sociodemographic characteristics (sex, age and schooling) and cardiovascular risk factors (smoking, physical inactivity, alcohol use and presence of diabetes and hypertension), which were investigated both in 1999 and in 2012. Since exposures can change over time, the presence of each risk factor was described as present only in 1999 or only in 2012, or present both in 1999 and in 2012. All participants with diabetes in 1999 confirmed that they still had the disease in 2012, and this analysis was therefore presented as cases described just in 2012 or in both 1999 and 2012. The other risk factors presented changes over the period. Thus, these were described as reported just in 1999 or just in 2012, or both in 1999 and in 2012.
We excluded 40 participants (8.8%) who reported having previously had a stroke or who reported use of drugs that can interfere with cognition: antipsychotics, anticonvulsants, antiparkinsonian drugs and anticholinesterase drugs. Cases of previous stroke were excluded based on self-reported morbidity within a medical diagnosis. All participants were told to bring their prescriptions and the packages of any drugs that they had been using during the preceding two weeks. 10 A standardized questionnaire was applied to investigate whether the participants had smoked at least one hundred cigarettes during their lifetime and how many alcoholic beverages they had consumed over the preceding two weeks. Leisure-time physical activity was assessed via a dichotomous response (yes, no) regarding whether this had been practiced over the previous two weeks. 10 Hypertension was defined based on reported use of antihypertensive drugs and on systolic blood pressure ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg. 11 Diabetes was defined through reported use of antidiabetic drugs and through fasting glucose level ≥ 126 mg/dl or glycated hemoglobin (HbA1C) level ≥ 6.5%. 12 Age was classified into a reference category (35-44 years old) and two groups of middle-aged adults (45-54 and 55-64 years old).
We used the international standard classification to stratify education levels: incomplete elementary education (< 8 years); completed elementary education and incomplete secondary education (8-10 years); completed secondary education and incomplete tertiary education (11-14 years); or completed tertiary education (> 14 years). 13 Validated and standardized cognitive tests to assess cognitive performance were administered in the same order, in a quiet room, by a trained interviewer. 14 These field researchers applied the cognitive tests after undergoing training consisting of eight hours of theory classes and another eight-hour training period with volunteers, followed by certification.
A word memory test was used to assess the participants' immediate memory (learning) by asking them to repeat ten unrelated words presented to them three times. Each word was presented for two seconds, in different orders. Memory recall was assessed by asking the participants to say the same ten words after they had completed other tests. Executive function was assessed by means of verbal fluency tests: 1) a semantic test in which the participants were asked to say as many names of animals as possible; and 2) a phonemic test, in which the participants were asked to say words starting with the letter F (phonemic test), for one minute. This battery of cognitive tests has been found to present moderate reliability for memory and good reliability for executive function. 15,16 The participants were stratified into two age groups (< 40 and 40+) and four levels of education (< 8, 8-10, 11-14 and > 14 years of schooling), with the aim of enabling adjustments to the scores for associations with age and education. The mean and standard deviation (SD) of the scores from each cognitive test were calculated for each stratum. A composite z-score was calculated by adding the z-scores of the cognitive tests. A composite cognitive test z-score that was one or more SD below the average was considered to represent low cognitive performance. 17 The statistical analyses were conducted using the STATA software, version 14.0 (Statacorp, College Station, Texas 77845, United States). 18 Multiple linear regression analysis was used to determine associations between sociodemographic variables and the crude scores of tests. Multiple logistic regression was used to examine associations between cardiovascular risk factors and lowered cognitive performance. Both the multivariate logistic model and the linear regression model were adjusted for sex, age and education. 19 The Research Ethics Committee of the Institute of Social Medicine, Rio de Janeiro State University, approved the study protocol in October 2011, under the number 0041.0.259.000-11.
All participants signed a written informed consent statement.

RESULTS
Among the participants, women were slightly predominant (51.8%). Most of the participants (56.9%) had had more than 14 years of schooling, and none of them had had less than 8 years of schooling. During the 13 years of follow-up, 18.6% and 2.1% of the participants developed hypertension and diabetes, respectively.
In univariate analysis, women performed better than men regarding learning (immediate memory). Older age, lower schooling level and presence of hypertension were significantly associated with worse performance in all cognitive tests, while diabetes was associated with worse performance in the phonemic verbal fluency test ( Table 1).
Further analysis using linear regression showed that there were no or minimal differences in cognitive performance between the The observed beneficial effect of physical activity on cognition was expected, given that several previous studies have documented its beneficial influence on brain function. 21 The small number of current smokers (10%) in this study may explain the absence of any association between smoking and cognitive performance.