Data for these analyses were derived from the NEDICES study, a longitudinal, population-based survey of the prevalence, incidence, and determinants of primary age-associated conditions of the elderly.[18–20] Detailed accounts of the study population and sampling methods have been published.[18–20]
The survey area consisted of three communities: (1) Las Margaritas (approximately 14,800 inhabitants), a working-class neighborhood in Getafe (Greater Madrid); (2) Lista (approximately 150,000 inhabitants), a professional-class neighborhood in Central Madrid, and (3) Arévalo (approximately 9,000 inhabitants), the agricultural zone of Arévalo County (125 km northwest of Madrid).[18–20] Up-to-date residents list was generated from population registers. In each community, survey eligibility was restricted to residents aged 65 years or older who were present there on December 31, 1993, or during six or more months of 1993.[18–20] Eligible persons who had moved away from the survey area were not traced. In Margaritas and Arévalo, every eligible subject was to be screened.[18–20] However, because of the large number of elderly residents in Lista, proportionate stratified random sampling was used to select subjects for screening.[18–20]
Briefly, at the time of their baseline assessment (1994–1995), 5,278 elderly subjects were interviewed using a 500-item screening questionnaire that assessed demographic factors and medical conditions.[18–20] Data were collected on demographics, current medications, and medical conditions during the face-to-face interview.[18–20] Subjects were asked to bring all medications taken in the past one week to the clinic, where the interviewer recorded the name and dose.[18–20] We assessed depressive symptoms by self-report, using a single screening question (‘Do you suffer from depression?’). The same approach has been used in previous population-based studies of depression.[21, 22] A comorbidity index was calculated based on the presence of the following conditions, according to a published comorbidity score developed in ambulatory care settings: atrial fibrillation, cancer, chronic obstructive pulmonary disease, depression, dementia, diabetes, epilepsy (treated), heart failure, myocardial infarction, psychiatric disorders, renal disease, and stroke. To assess sleep duration, each participant was asked to indicate their "total hours of actual sleep in 24 hours ".[24, 25] Participants indicated their typical total daily sleep duration as the sum of nighttime sleep and daytime napping.[24, 25] Physical activity was classified as sedentary, light, moderate, and high. Details of the measurement of physical activity in the NEDICES study have been reported elsewhere.
A Spanish adaptation of the Pfeffer Functional Activities Questionnaire was administered to participants. This version includes 11 items; the total score for the 11 items ranges from 0 (completely independent) to 33 (completely dependent). Participants were asked to rate their current health using a 5-point scale using the question "In general terms, how would you describe your health: excellent, good, fair, poor, or very poor?".[27, 28] There were a small number of subjects in several categories. Hence, as in several previous studies, we collapsed response options into three categories.[28, 29] These three categories were very good/good, fair, and poor/very poor.[28, 29]
The same methods were used during the second (i.e., follow-up) evaluation (1997–1998).
As described,[30, 31] a 37-item Mini-Mental State Examination (37-MMSE) was administered at both the baseline assessment (1994–1995) and the follow-up assessment (1997–1998). This was a Spanish adaptation of the standard MMSE. [30, 31] It included all of the standard MMSE items and three additional items: (1) an attention task, i.e., "say 1, 3, 5, 7, 9 backward", (2) a visual order, i.e., a man raising his arms, and (3) a simple construction task, i.e., copying two overlapping circles.[30, 31]
The diagnosis of dementia was assigned using the Diagnostic and Statistical Manual of Mental Disorders (DSM)–IV criteria and required evidence of cognitive impairment and impairment in social or occupational function.
Final selection of participants
Of the 5,278 participants evaluated at baseline, we excluded 467 participants with dementia, including 306 with dementia diagnosed at baseline evaluation (1994–1995) (i.e., prevalent cases), and 161 who developed dementia by the follow-up evaluation (1997–1998) (i.e., incident cases). We further excluded 2,187 participants who were evaluated at baseline because they declined a follow-up assessment or had incomplete follow-up assessments, had died or were unreachable (N = 1,246) or with incomplete 37-MMSE examinations (N = 938) or incomplete smoking status (N = 3).
The final sample of 2,624 was similar to the base sample of 5,278 participants in terms of sex (1,485 [56.6%] vs. 3,040 [57.6%] women, chi-square = 0.72, p = 0.395). However, they were more educated (268 [10.1%] vs. 711 [13.6%] were illiterate, chi-square = 18.58, p < 0.001) and, on average, 1.6 years younger (72.7 ± 5.9 vs. 74.3 ± 7.0 years, t = 11.0, p < 0.001).
Statistical analyses were performed in SPSS Version 25.0 (SPSS, Inc., Chicago, IL). None of the continuous variables (age, years of education, drink-years, comorbidity index, body mass index, and Pfeffer Functional Activities Questionnaire total score) was normally distributed (Kolmogorov-Smirnov, p < 0.001), even after log-transformation. Therefore, baseline characteristics scores were compared using Mann–Whitney and Kruskal–Wallis tests. The Chi-square test was applied to determine associations between categorical variables.
For all participants, we calculated baseline pack-years (i.e., packs of cigarettes smoked per day multiplied by years smoked). Pack-years were categorized into tertiles (< 19.0, 19.0–47.0, > 47.0). Participants who had never smoked were coded as having zero pack-years, and they served as the reference group in those analyses, which included all participants. In addition, for other analyses, participants were classified into three groups: never smokers, ex-smokers, and current smokers.
Data on ethanol consumption (average number of drinks consumed per day and duration of consumption in years) were used to compute drink-years.[35, 36] One "drink-year" was defined as the intake of one ethanol drink per day for one year. One drink consisted of one can of beer (360 mL), one glass of wine (120 mL), or 30 mL of liquor.[35, 36]
Change in 37-MMSE (baseline 37-MMSE score − follow-up 37-MMSE score) was divided into tertiles (lower tertile ≥ 2 points improvement in score, higher tertile ≥ 2 point decline in score). We dichotomized this variable into higher (cognitive decline) vs. middle and lower tertiles for the current analyses.
We used Cox proportional-hazards models to estimate hazard ratios (HRs) for the risk of cognitive decline, generating 95% confidence intervals (CIs). The time variable was the years from the first evaluation date (1994 to 1995) to the date of the second evaluation (1997 to 1998). The dependent (outcome) variable was a higher tertile of 37-MMSE change (cognitive decline), with the other two tertiles as the reference group. We considered baseline variables associated with either higher tertile of 37-MMSE change or long-life cumulative smoking exposure. Variables assessed at baseline that we considered included age in years, sex, years of education, sleep duration (≤ 5, 6–8 ≥ 9 hours), drink-years, use of medications that potentially affect cognitive function, comorbidity index, body mass index, depressive symptoms, subjective wellbeing (good/excellent, average and bad/terrible), physical activity (sedentary lifestyle, mild, moderate, and high physical activity), and Pfeffer Functional Activities Questionnaire total score. The exposure variable was pack-years tertiles (< 19.2, 19.2–49.0, > 49.0). The reference group was composed of those who have never smoked (zero pack-years). In other analyses, the exposure variable was the smoking status (ex-smoker and current smoker); the reference group was those who have never smoked.