Participants of the Cardiovascular Risk Factors, Aging and Dementia (CAIDE) study in Finland were first examined at midlife (baseline) in the North Karelia Project and the FINMONICA study, where individuals were assessed in one of the following years for the baseline assessment: 1972, 1977, 1982 or 1987 . Baseline participation rates ranged between 82 and 90%. In 1998, a random sample of 2000 survivors living in the cities of Kuopio and Joensuu, aged 65–79, were invited for a first re-examination (Fig. 1). A total of 1449 (72.5%) individuals participated and 1409 completed the cognitive assessments. The mean follow-up time was 21 years (SD = 4.9). Participants returned for a second re-examination between 2005 and 2008. In 2005, of the 2000 original sample, 1426 were still alive and were still living in the same region. When invited, 909 (63.7%) of them participated and 852 completed the cognitive assessment. A total of 1511 individuals participated in at least one re-examination, and 750 participated in both. Mean ages at each time point were: at baseline, 50 years (SD = 6.0, age range: 39-64); at the first re-examination, 71.3 years (SD = 4.0, age range 65–80); at the second re-examination, 78.6 years (SD = 3.7, age range 72–90). Local ethics committees approved the CAIDE study and participants provided written informed consent. The study complies with the Declaration of Helsinki.
Measurement of work-related stress
Perceived work-related stress was measured in midlife using two questions focusing on job demands. These questions were adapted from the questions validated by Karasek et al.  and have been used reliably by various research groups [3, 20, 31]. Both questions have the same 5-point likert scale. The questions were, “How often do you struggle to cope with the amount of work?” and “How often are you bothered by constant hurry at work?”. After reverse coding to facilitate the interpretation of the results, the response options were: 1 = never, 2 = rarely, 3 = sometimes, 4 = often, 5 = always. Data on work-related stress were available for 1273 participants. Both questions were highly correlated (Spearmans ρ = 0.623, p < 0.001) and were summed to produce a composite measure of work-related stress.
At both re-examinations, a comprehensive battery of neuropsychological tests was administered to assess multiple cognitive domains. For the current study, we used the following tests that were administered at both re-examinations: (1) global cognition measured by the mini mental state examination (MMSE) ; (2) episodic memory assessed by an immediate word recall test (10-word list); (3) executive functioning measured by the Stroop test (time difference between the task of naming the color of the ink used to write the name of another color, and the task of naming colors of dots); (4) verbal fluency tested by category fluency test (number of correct animal names generated in 60 s); (5) psychomotor speed assessed by the letter digit substitution test; (6) manual dexterity measured by the bimanual Purdue Pegboard test. Dementia diagnosis was carried out using a three-step protocol previously described .
At baseline (midlife), assessments and survey methods were standardized and adhered to international guidelines and the World Health Organization (WHO) (Multinational MONItoring of trends and determinants in CArdiovascular disease) MONICA protocol . Re-examination surveys were similar and comparable to those at baseline. Baseline surveys involved self-administered questionnaires on medical history, sociodemographic factors, health status, health-related behaviors and psychological-related factors. We selected the following covariates previously shown to be associated with worse cognition and/or high levels of stress: age, sex, education, APOE ε4, respiratory, cardio/cerebrovascular and musculoskeletal conditions, and type of occupation (white collar vs. blue collar). Occupation type was measured by asking individuals to select their longest-held occupation among the following categories: office/service, farming/forestry, mining/industrial/construction work, housewives, or other. Hopelessness was measured using the following two questions described previously : “I feel that it is impossible to reach the goals I would like to strive for” and “The future seems to me to be hopeless, and I cannot believe that things are changing for the better”. A five-point Likert scale was used, coded as 0 = absolutely agree; 1 = somewhat agree; 2 = cannot say; 3 = somewhat disagree; or 4 = absolutely disagree. A trained nurse verified the answers and addressed participants’ questions. The nurse also measured height, weight and blood pressure. A venous blood sample was obtained, and allowed for measures of biomarkers, including cholesterol and APOE genotype from blood leucocytes, for which HHaI digestion and polymerase chain reaction were used . The Hospital Discharge Register was used for information on respiratory and cardio/cerebrovascular conditions (chronic obstructive pulmonary disease, asthma, coronary artery disease, stroke, myocardial infarction, atrial fibrillation, cardiovascular surgery, heart failure or diabetes). These conditions were combined into a dichotomous variable (yes/no) reflecting the presence of any midlife respiratory or cardio/cerebrovascular conditions. All covariates were measured at baseline.
We conducted analyses using Stata 13.0 (Stata Corp, College Station, TX, USA). We analyzed participant baseline characteristics using Chi-square (χ
2) tests for categorical variables (data reported as percentages), and Student t tests for continuous variables (data reported as means (standard deviations [SD])). The significance level for all analyses was set at p < 0.05 (Table 1). Zero-skewness log-transformations were applied to cognitive test scores (Stata command lnskew0). Results were standardized to have SD = 1.
Participants with dementia at the first re-examination were excluded from analyses at the first re-examination. Participants with dementia at the second re-examination were excluded from analyses at the second re-examination. To maximize sample size, all subjects with cognitive assessments in at least one re-examination (n = 1332) were considered in analyses. This means that the analyses are based on both the first and second re-examination combined. For subjects with cognitive assessments in both re-examinations (n = 685), two observations were included (i.e., one for test results at the first re-examination, and one for test results at the second re-examination). Data were organized in what is often referred to as long format.
To investigate the associations between midlife work-related stress and cognition, we performed linear regression analyses for each of the cognitive domains. We reported results as β-coefficient and 95% confidence intervals (CI). All analyses were adjusted for a basic set of confounders: age, sex, years of education and follow-up time (Model 1). Model 2 additionally adjusted for the type of occupation. Model 3 additionally for APOE ε4 genotype, hopelessness and midlife respiratory, cardio/cerebrovascular conditions.