1 INTRODUCTION

According to the 2015 World Alzheimer Report, as many as 46.8 million people worldwide were suffering from dementia in 2015,1 resulting in impaired quality of life and high costs in the health care system. The prevalence of dementia is growing rapidly, almost doubling every 20 years, and the prognosis for 2030 is 75 million cases and for 2050 as many as 132 million worldwide.1

In population-based samples from Sweden, investigated in 1976–2006, approximately 2% were estimated to have dementia at age 70% and 5% at age 75,2 with the risk doubling every 5 years, resulting in a 45% rate among people 90 years and older. However, aggregated data from Europe and North America show that the incidence rate of dementia has declined by 13% per decade over the last 25 years. Although the causes of this decline remain unknown, the findings strongly suggest that the risk of dementia is modifiable, i.e. some risk factors are preventable.3 The contradictory facts of declining incidence in high-income countries and increasing global prevalence may be explained by less risk factor prevention in low-income countries and increasing life expectancy and longer survival with dementia in general.1

Cognitive impairment (CI) can manifest both as severe cognitive impairment (dementia) and as mild cognitive impairment (MCI). MCI is defined as a decline in cognitive functions, which goes beyond what is expected in normal ageing, but does not yet meet the criteria for dementia.4 MCI is a heterogeneous condition and can involve multiple cognitive domains such as language, memory, learning, attention, visuospatial functions and executive functions.5 Just like dementia, the risk of MCI increases with age. However, prevalence estimates in previous studies differ due to differences in MCI criteria, populations studied and methodology. Prevalence in the population aged over 65 years ranges between 10%–20%.6-8 Among persons 70–75 years old, the estimated prevalence is 20%–25%.7, 9

The single most important risk factor for developing CI is age.7, 10, 11 Several other factors are associated with an increased risk. Non-modifiable factors are sex,1 the ApoE ε4 allele for Alzheimer's disease,7, 8, 11 and mutations in genes coding for proteins such as Tau, TDP-43, α-synuclein, APP or Presenilin 1 or 2.10 Modifiable risk factors associated with CI are lower level of education,7, 8, 11 mid-life high BMI/abdominal obesity,8, 10, 11 living alone,12 hypertension,8, 11, 13 high intake of alcohol,10, 11 sleep disturbances,6, 11 physical inactivity,6, 8, 10, 11, 14 mental stress,15 and depression.6, 8, 10, 11 Moreover, cardiovascular diseases,10, 13 diabetes,6, 8, 11, 13 hyperlipidaemia,10, 11 and smoking,8, 10, 11, 16 are suggested risk factors although study results diverge. In addition, many studies conclude that frequent, mentally and socially stimulating activities, both in mid-life and later in life, are protective factors for CI.8, 10, 11, 14

Most of the prior studies of risk factors for cognitive impairment are retrospective and based on selected patients already being diagnosed with MCI or dementia. Prospective population-based cohort studies are scarce.14, 15, 17 Consequently, there is a need for more research on the prevalence of cognitive decline in the general population, and on early risk factors for CI.

‘The Study of Men Born in 1943’ is a prospective cohort study in Gothenburg, Sweden, following a sample of men from the age of 50, where cognitive testing was performed at age 73. The aim of this study was to investigate whether factors measured at the examination at age 50 predicted cognitive impairment 23 years later.

2 MATERIALS & METHODS 2.1 Study population

In 1993 a random population sample of 50% of all men born in 1943 and residing in Gothenburg, Sweden, were invited to participate in a longitudinal, prospective study, i.e. ‘The Study of Men Born in 1943’. The aim of the study was to investigate the prevalence of risk factors for cardiovascular disease. Of 1462 invited men, 798 (55%) agreed to participate in the study. The participants were followed prospectively and were invited to a re-examination at age 60 (2003) and at age 71 (2014). The study has previously been described in detail elsewhere.18, 19

In April 2016, we invited all the men who participated in the 2014 examination to take part in a cognitive study. Of the 530 men still alive and residing in Sweden, 342 agreed to participate (65%) and were examined in the cognitive study (Figure 1). The cognitive examination was performed in 2016 and early 2017 when all the participants were 73 years of age. Eight participants were excluded after the cognitive examination because of poor Swedish language skills, which significantly affected their test results, and one participant due to impaired eyesight. As a result, the final cognitive study population totalled 333 (63% of those invited and still alive).

Flowchart of study

All the participants have given their informed consent and the study has been approved by the Gothenburg Regional Research Ethics Board (No: 886–13 and T187-16) and complies with the Declaration of Helsinki. The study is registered at ClinicalTrials.gov, with id-number: NCT03138122.

2.2 Examination at age 50

The physical examinations at baseline in 1993 were performed in the morning after an overnight fast and included height, weight, waist circumference, blood pressure, and venous blood sampling. Details of the examination are given in supplementary material A.

Examinations of extrapyramidal symptoms were performed according to the Simpson-Angus Scale.20 The selected tests were: Gait, Shoulder Shaking, Elbow Rigidity, Wrist Rigidity, Leg Pendulousness, Tremor and Cogwheel Rigidity. The participants’ performances were graded from 0 to 4. For further description of the method see Supplementary Material A. The participants were video-filmed while performing these tests and assessment of Gait was retrospectively done by two physicians, one specialist in psychiatry and one specialist in internal medicine, by analysing the video-film.

Prior to the examination, the participants completed a questionnaire addressing education, marital status, smoking habits, alcohol use, coffee consumption, physical activity during leisure time, cardiovascular diseases diagnosed by a physician, sleeping habits, stress and psychosocial factors (Supplementary Material A).

2.3 Cognitive testing at age 73

The test battery consisted of 11 tests (Table 1), out of which six constitute the Cognitive Assessment Battery (CAB), an established test battery for distinguishing between persons with no CI, MCI and dementia.5 The tests followed a strict manual that described precisely in which order and how each test should be introduced, performed and graded. Test leaders, trained by an experienced psychologist specializing in cognitive decline, administered the tests. A test session lasted 40–60 min. The participants signed an informed consent before the session. For a detailed description of the tests, see Supplementary Material B.

TABLE 1. Cognitive tests included in ‘The Study of Men Born in 1943’, and the results from each test Tests (in order of performance) Cognitive domain Min-max Cut-off Mean Median Fail †, n (%)† Text recall (immediate vs delayed) Memory & Learning (−21)–21 <0 1 1 69 (21) SDMT Speed & Attention 0–110 <25 p 35 35 36 (11) Naming test Language 0–30 <25 p 27 27 56 (17) Clox & Cube Visuospatial 0–12 <11 p 11 12 81 (25) Token test Language 0–6 <5 p 5 5 71 (21) Stroop III Executive functions 0-no max >49 sec 34 31 26 (8) ROCF Copying Visuospatial 0–36 <27 p 30 29 87 (26) FAS Language 0-no max <24 p 34 33 78 (23) ROCF Recall Memory & Learning 0–36 <7.5 p 15.5‡ 14.5‡ 38 (11) PaSMO A−1 Executive functions 0-no max >93 sec 81‡ 75‡ 155 (47) Trailmaking A Speed & Attention 0-no max >55 sec 46 44 74 (22) Abbreviations: CAB, Cognitive Assessment Battery; F-A-S, Verbal fluency test letters F, A and S; PaSMO, Parallel Serial Mental Operations; ROCF, Ray-Osterrieth Complex Figure; SDMT, Symbol Digit Modalities Test. †Participants with a test result indicating cognitive impairment. ‡Includes only participants able to finish the test. 2.4 Definitions of mild and severe cognitive impairment

For the complete test battery (11 tests); fewer than four failed tests were considered normal, four to seven failed tests were defined as MCI, and more than seven failed tests were regarded as severe cognitive impairment. These cut-off levels correspond well to the CAB cut-offs, where MCI is defined as two to three failed tests (of six), and severe cognitive impairment/dementia as more than three failed tests.5

2.5 Statistical methods

Descriptive statistics were used to present the main characteristics of the study population. To measure the association of the risk factors with the outcome variable, cognitive impairment (mild or severe), odds ratios were calculated by logistic regression.

Since education had a strong association with CI, an adjustment for education was made in the subsequent analyses. Thus, multiple logistic regression models were used with an adjustment for education as a potential confounder and for stepwise multivariable analysis.

The final multiple regression model built for the multivariable analysis included variables according to the following two steps: Step 1 included variables significantly related to the outcome, after adjustment for education. The variables were analysed in separate groups according to the type of variable (i.e. anthropometrics, co-morbidity, lifestyle, sleep, extrapyramidal symptoms). Seven variables did not belong to a specific group and were excluded from the Step 1 analysis. Step 2 included all the variables still significantly associated with the outcome after Step 1, as well as the seven variables excluded from Step 1. A p-value of ≤.05. was considered statistically significant.

The statistical analyses were conducted using the SAS statistical software package, version 9.4; SAS Institute Inc.

3 RESULTS

Of the 333 men participating in the cognitive study, 80 (24.0%) performed at a level corresponding to MCI and four (1.2%) at a level consistent with a severe cognitive impairment. The results from each test are shown in Table 1.

The prevalence of CI was significantly lower among participants with a higher level of education (Table 2). Men with more than 12 years of education (university studies) had an odds ratio of 0.23 (95% CI 0.12–0.46, p < .001) for CI, when compared with participants with fewer than 10 years of study. Due to this strong association with CI, adjustment for education was made in the subsequent analyses.

TABLE 2. Comparison between participants with Cognitive impairment (CI) at age 73 and controls with regard to demographics, anthropometrics, co-morbidity and extrapyramidal remarks at baseline (age 50) Variable No CI CI Univariate Adjusted for education n = 249 n = 84 OR (95% CI) p OR (95% CI) p Education, n % <10 years 77 (30.9) 50 (59.5) 1.00 – – 10–12 years 86 (34.5) 21 (25.0) 0.38 (0.21–0.68) .001 >12 years 86 (34.5) 13 (15.5) 0.23 (0.12–0.46) <.0001 Living alone, n % 24 (9.6) 19 (22.6) 2.74 (1.41–5.31) .002 2.80 (1.40–5.63) .004 Height (m), mean (SD) 1.79 (0.07) 1.80 (0.07) n.s. n.s. Weight (kg), mean (SD) 82.5 (10.5) 85.7 (11.8) 1.03 (1.00–1.05) .03 1.02 (1.00–1.05) .04 Waist (cm), mean (SD) 93.5 (8.3) 96.9 (8.7) 1.05 (1.02–1.08) .05 1.04 (1.01–1.08) .006 BMI (kg/m2), mean (SD) 25.7 (3.0) 26.5 (3.3) n.s. n.s. BMI ≥30, n % 17 (6.8) 12 (14.3) 2.27 (1.04–4.99) .04 n.s. Myocardial infarction, n % 4 (1.6) 0 (0.0) n.s. n.s. Heart failure, n % 0 (0.0) 2 (2.4) 15.12 (0.71–318.19) .01 n.s. Atrial fibrillation, n % 1 (0.4) 1 (1.2) n.s. n.s. Stroke, n % 1 (0.4) 0 (0.0) n.s. n.s. Hypertension, n % 44 (17.7) 25 (29.8) 1.97 (1.12–3.50) .02 1.84 (1.02–3.33) .04 Diabetes: prev. diag. or Fp-gluc ≥6, n % 3 (1.2) 4 (4.8) 4.16 (0.91–18.95) .05 5.09 (1.04–24.96) .04 Fp-glucose (mmol/L), mean (SD) 4.47 (0.60) 4.63 (0.88) n.s. 1.44 (1.01–2.04) .04 S-cholesterol (mmol/L), mean (SD) 5.78 (1.05) 5.82 (1.00) 1.04 (0.82–1.32) .04 n.s. S-triglycerides (mmol/L), mean (SD) 1.51 (0.84) 1.69 (1.14) n.s. n.s. Shoulder shaking remark, n % 7 (3.0) 11 (13.6) 5.16 (1.93–13.82) .0004 5.70 (2.01–16.19) .001 Elbow rigidity remark, n % 15 (6.3) 11 (13.6) 2.33 (1.02–5.30) .04 2.61 (1.10–6.23) .03 Leg pendulousness remark, n % 3 (1.3) 11 (13.6) 12.26 (3.37–45.17) <.0001 13.84 (3.56–53.80) .0001 Cogwheel remark, n % 1 (0.4) 4 (4.9) 12.26 (1.35–111.35) .005 17.18 (1.72–171.88) .02 EPS remarks (tot. remarks >1) 6 (2.5) 12 (15.0) 6.79 (2.46–18.78) <.0001 8.20 (2.76–24.33) .0001 Note n.s.=not significant, used for p values >.05. Abbreviation: EPS, Extrapyramidal symptoms.

A number of variables, at baseline, such as; living alone, higher body weight, larger waist circumference, BMI ≥30, hypertension, higher fasting p-glucose, and diabetes were associated with having CI at age 73, after an adjustment for education (Table 2). In addition, heavy smoking, high alcohol consumption, poorer quality of sleep, difficulty falling asleep, financial stress, bad family situation, hidden irritability, worse mood, less endurance, rarely being busy, and poorer quick-thinking ability were also more frequently associated to CI at age 73, after an adjustment for education (Table 3).

TABLE 3. Comparison between participants with Cognitive impairment (CI) at age 73 and controls with regard to lifestyle, sleeping habits, psychological variables and perceived stress at baseline (age 50) Variable No CI CI Univariate Adjusted for education n = 249 n = 84 OR (95% CI) p OR (95% CI) p Smoking, n % n.s. n.s. Never smoked 93 (37.4) 30 (35.7) Ex-smoker 106 (42.6) 28 (33.3) Current smoker 50 (20.1) 26 (31.0) Heavy smoker (>14 cig/day), n % 18 (7.2) 15 (17.9) 2.79 (1.34–5.82) .005 2.94 (1.35–6.37) .006 Need morning ‘eye-opener’, n % 15 (6.0) 15 (17.9) 3.39 (1.58–7.28) .001 3.31 (1.49–7.33) .003 Current drinking problem (alcohol), n % 3 (1.2) 3 (3.6) n.s. 5.54 (1.03–29.76) .05 Previous drinking problem (alcohol), n % 12 (4.8) 8 (9.5) n.s. n.s. Coffee (cups per day), mean (SD) 4.4 (2.4) 4.5 (2.9) n.s. n.s. Sedentary lifestyle, n % 31 (12.4) 9 (10.7) n.s.