Methods and measures
During both phases of the Lieto study all participants were interviewed about their socioeconomic background, physical and psychosocial factors, functional abilities, use of medications and health behaviour. They were also clinically examined by a health centre physician who was part of the research team (RI or ML). Clinical tests were performed by a trained research nurse. Medical records in the Härkätie Health Centre, Lieto, were used in recording previous diagnoses. Similar measures were used in both phases.
Cognitive functioning was measured during both phases with an MMSE performed by a trained nurse. The MMSE scale consists of 23 items, and the sum score ranges from 0 to 30, higher scores indicating better cognitive performance . The mean change in MMSE sum scores during the follow-up was used as an outcome variable.
Information about the use of all medications prior to seven days before the interview was collected in a personal interview conducted by a trained nurse at the baseline and follow-up interviews to describe the total medication at both data collection phases. The participants had been informed to bring along their prescription forms and medications in order to confirm their current use of medication. A health centre physician from the research team (RI or ML) verified the medications from medical records. In cases where the person interviewed was unable to answer questions adequately, a close relative or caregiver provided the relevant information. If the participant was unable to visit the health centre, a trained nurse made a home visit to check the medications. All the prescribed medications (both regular and irregular) and non-prescribed medications (vitamins etc.) were taken into account.
Medications were defined by using the Anatomical Therapeutic Chemical (ATC) Classification (1996) . The groups of medications defined as those with an effect on the central nervous system (CNS) and used in the analyses of this study were as follows: benzodiazepines and related drugs (BZDs) (ATC codes N05BA, N05CD, N03AE01, N05CF, A03CA, C01DA70, M05AA51, N06CA01, N02BA71), antipsychotics (APs) (ATC codes N05A, N06CA01), antidepressants (ADs) (ATC codes N06A, N06CA), opioids (Ops) (ATC codes N01AH, N02A, N02BE51, R05DA, R05FA), anticholinergic medications (AChs) (ATC codes N04A, N05AA01, N05AA02, N05AB01, N05AB02, N05AB03, N05AB04, N05AC01, N05AC02, N05AF01, N05AF03, N05AF05, N05BB01, N06AA04, N06AA06, N06AA09, N06AA12, 102AG, A03AA, A03AB, A03AX03, A03B, A03CA, A03CB31, A03DA, A03FA01, A04AD01, A04AD12, C01BA01, C01BA03, C01BA51, C01BA71, R03BB, M03B, G04BD, S01FA, R01BA01, R01BA51, R06AB01, R06AE03, R06AE53) and antiepileptic medications (AEs) (ATC code N03A) [18, 19].
The usages of these groups of CNS medications were first dichotomized (regular or irregular use vs. no use). The doses of the medications were not taken into account. Eight variables describing the use of CNS medications were then formed: BZDs, APs, ADs, psychotropics (including BZDs, APs or ADs), Ops, AChs, AEs and any CNS medications (including BZDs, APs, ADs, Ops, AChs or AEs). Finally, 21 variables describing all combinations of the CNS medications were formed.
Previously known risk factors of cognitive decline  such as age, sex, basic education, hypertension, atrial fibrillation or flutter, diabetes mellitus, congestive heart disease and smoking at the baseline were used as control factors. Interviews were used in collecting data about basic education and current smoking. Hypertension, diabetes and congestive heart disease were defined according to clinical examination, medical history or previous diagnoses in the medical records. The diagnosis of atrial fibrillation or flutter was based on a diagnosis in the medical records or on electrocardiograms (ECG) recorded during the baseline examination.
Informed consent was obtained from all participants or their caregivers in both phases of the study. The study plans of the first and second Lieto studies were approved by the Ethical Committee of the Hospital District of Southwest Finland.
The analyses were performed for the total population and separately for men, women and younger (65-74 yrs) and older (≥ 75 yrs) age groups. Participants who at the beginning of the follow-up were using one type of the CNS medications described in the methods and measures section were first compared with participants who did not use any of these medications at baseline and then with participants not using the medication concerned. Participants using a combination of two or more of the CNS medications were compared with participants using none of the CNS medications or their combinations.
Chi-square and Fisher exact tests were used to test differences in categorical variables between sexes, age groups and, diagnoses as well as between medication users and control groups at baseline and during the follow-up examination. The significances of changes in MMSE sum scores during the follow-up in the total population and in all subgroups were tested using the Wilcoxon signed rank test. The differences of the mean MMSE sum scores, the mean ages and the mean number of medications and the changes of mean MMSE sum scores between the groups were tested with the Mann-Whitney U test. Associations between the use of a certain group of CNS medications or the use of a combination of CNS medications and the risk of cognitive decline were first analyzed by the Mann-Whitney U test. The significances of the differences in the changes of cognitive functioning during the follow-up between the users of a certain group of CNS medications or the users of a combination of CNS medications and the corresponding control group of nonusers were first tested with the Mann-Whitney U test. After these analyses, adjusted analyses using the analysis of covariance were performed for those groups in which the associations between the use of a certain CNS medication or the concomitant use of certain CNS medications and the risk for cognitive decline were significant (p < 0.05) or tended to be significant (0.05 < p < 0.10) in these bivariate analyses. The associations of these variables (age, sex, education, hypertension, atrial fibrillation or flutter, diabetes mellitus, congestive heart disease and smoking at baseline) with the decline in MMSE sum scores were first analyzed in the total population, and only the variables that were significantly associated with decline (higher age, p < 0.001, and congestive heart disease, p = 0.002) were adjusted in the analyses of covariance.