Potentially inappropriate medication: Association between the use of antidepressant drugs and the subsequent risk for dementia

Author’s Accepted Manuscript Potentially inappropriate medication: Association between the use of antidepressant drugs and the subsequent risk for dementia Kathrin Heser, Tobias Luck, Susanne Röhr, Birgitt Wiese, Hanna Kaduszkiewicz, Anke Oey, Horst Bickel, Edelgard Mösch, Siegfried Weyerer, Jochen Werle, Christian Brettschneider, HansHelmut König, Angela Fuchs, Michael Pentzek, Hendrik van den Bussche, Martin Scherer, Wolfgang Maier, Steffi G. Riedel-Heller, Michael Wagner

PII: DOI: Reference:

www.elsevier.com/locate/jad

S0165-0327(17)31523-9 http://dx.doi.org/10.1016/j.jad.2017.09.016 JAD9224

To appear in: Journal of Affective Disorders Received date: 31 July 2017 Revised date: 11 September 2017 Accepted date: 13 September 2017 Cite this article as: Kathrin Heser, Tobias Luck, Susanne Röhr, Birgitt Wiese, Hanna Kaduszkiewicz, Anke Oey, Horst Bickel, Edelgard Mösch, Siegfried Weyerer, Jochen Werle, Christian Brettschneider, Hans-Helmut König, Angela Fuchs, Michael Pentzek, Hendrik van den Bussche, Martin Scherer, Wolfgang Maier, Steffi G. Riedel-Heller and Michael Wagner, Potentially inappropriate medication: Association between the use of antidepressant drugs and the subsequent risk for dementia, Journal of Affective Disorders, http://dx.doi.org/10.1016/j.jad.2017.09.016 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting galley proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Title: Potentially inappropriate medication: Association between the use of antidepressant drugs and the subsequent risk for dementia Author information1a: Kathrin Heser1, Tobias Luck2, Susanne Röhr2,3, Birgitt Wiese4, Hanna Kaduszkiewicz5, Anke Oey4, Horst Bickel6, Edelgard Mösch6, Siegfried Weyerer7, Jochen Werle7, Christian Brettschneider8, Hans-Helmut König8, Angela Fuchs9, Michael Pentzek9, Hendrik van den Bussche10, Martin Scherer10, Wolfgang Maier1, 11, Steffi G. Riedel-Heller2 */ Michael Wagner1, 11* (shared last authorship) for the AgeCoDe & AgeQualiDe study groups 1

Department of Psychiatry and Psychotherapy, University of Bonn, Bonn, Germany

2

Institute of Social Medicine, Occupational Health and Public Health, University of Leipzig, Leipzig, Germany 3

LIFE – Leipzig Research Center for Civilization Diseases, University of Leipzig, Leipzig, Germany

4

Working Group Medical Statistics and IT Infrastructure, Institute of General Practice, Hannover Medical School, Hannover, Germany 5

Institute of General Practice, Medical Faculty, Kiel University, Kiel, Germany

6

Department of Psychiatry, Technical University Munich, Munich, Germany

7

Central Institute of Mental Health, Medical Faculty Mannheim/Heidelberg University, Mannheim, Germany 8

Department of Health Economics and Health Services Research, Hamburg Center for Health Economics, University Medical Center Hamburg-Eppendorf, Hamburg, Germany 9

Institute of General Practice (ifam), Medical Faculty, Heinrich-Heine University Düsseldorf, Düsseldorf, Germany 10

Department of Primary Medical Care, University Medical Center Hamburg-Eppendorf, Hamburg, Germany 11

DZNE, Center for Neurodegenerative Diseases, Bonn, Germany

1a

Corresponding author: Dr. Kathrin Heser, postal address: Department of Psychiatry and Psychotherapy, University of Bonn, Sigmund-Freud-Straße 25, 53105 Bonn, Germany; e-mail: [email protected]; telephone number: ++49-228-287-19827, fax: ++49-228-287-90-19827

1

ABSTRACT Background: Potentially inappropriate medication (PIM)2 is associated with an increased risk for detrimental health outcomes in elderly patients. Some antidepressant drugs are considered as PIM, but previous research on the association between antidepressants and subsequent dementia has been inconclusive. Therefore, we investigated whether the intake of antidepressants, particularly of those considered as PIM according to the Priscus list, would predict incident dementia. Methods: We used data of a prospective cohort study of non-demented primary care patients (n = 3239, mean age = 79.62) to compute Cox proportional hazards models. The risk for subsequent dementia was estimated over eight follow-ups up to 12 years depending on antidepressant intake and covariates. Results: The intake of antidepressants was associated with an increased risk for subsequent dementia (HR3 = 1.53, 95% CI4: 1.16-2.02, p = .003; age-, sex-, education-adjusted). PIM antidepressants (HR = 1.49, 95% CI: 1.06-2.10, p = .021), but not other antidepressants (HR = 1.04, 95% CI: 0.66-1.66, p = .863), were associated with an increased risk for subsequent dementia (in age-, sex-, education-, and depressive symptoms adjusted models). Significant associations disappeared after global cognition at baseline was controlled for. Limitations: Methodological limitations such as selection biases and self-reported drug assessments might have influenced the results. Conclusions: Only antidepressants considered as PIM were associated with an increased subsequent dementia risk. Anticholinergic effects might explain this relationship. The association disappeared after the statistical control for global cognition at baseline. Nonetheless, physicians should avoid the prescription of PIM antidepressants in elderly patients whenever possible. Keywords: late-life depression, antidepressant drugs, antidepressants, potentially inappropriate medication, dementia, oldest-old

2

PIM = potentially inappropriate medication HR = hazard ratio 4 CI = confidence interval 3

2

INTRODUCTION Potentially inappropriate medication (PIM) is associated with an increased risk for adverse drug effects in elderly patients, with an increased risk for detrimental outcomes such as hospitalization and mortality (e.g., Lau et al., 2005) and with higher monetary costs due to hospital admissions caused by adverse drug events (e.g., Schneeweiss et al., 2002). The intake of anticholinergic drugs considered as PIM can also increase the risk for subsequent dementia (e.g., Carrière et al., 2009; Jessen et al., 2010; Gray et al., 2015). Risacher et al. (2016) found that anticholinergic drug intake was associated with poorer cognitive performance, reduced brain metabolism, higher levels of brain atrophy, and an increased risk for clinical conversion to mild cognitive impairment or dementia in cognitively normal participants. Several different drug groups were summarized in the Priscus list by Holt et al. (2010) as potentially inappropriate for elderly patients in the German health care system. Besides other drugs like analgesics or antibiotics, several antidepressants are included in this list, because their intake is associated with an increased risk for delirious states or cognitive deficits due to anticholinergic effects (e.g., amitriptyline, doxepine, and imipramine (all tricyclic antidepressants (TCA)); cf. Holt et al., 2010). Nonetheless, these drugs are frequently prescribed even in patients with cognitive impairment. A German study found that antidepressant treatment in older patients aged 65 years and older was often initiated with TCA (Jobski et al., 2017). Wucherer et al. (2017) found that more than every fifth in a sample of community-dwelling primary care patients with dementia received at least one PIM, most frequently anticholinergic antidepressants and long-acting benzodiazepines. Some findings on the association between the intake of antidepressant drugs and the subsequent risk for dementia exist, but results are rather inconsistent. In a population-based study of participants aged 65 and older, antidepressant drug use accompanied by persistent depressive symptoms was associated with a highly increased risk for dementia, but only in men (Fuhrer et al., 2003). In a nationally representative study of participants with an average age of 72 years, the use of antidepressant drugs did not modify cognitive decline over six years after the adjustment for depressive symptoms, medical comorbidity, and anticholinergic burden (Saczynski et al., 2015). However, Gray et al. (2015) found a dose-response relationship between anticholinergic properties exhibited by drugs like TCA and the risk of subsequent dementia over ten years. Interestingly, both selective serotonin reuptake inhibitor (SSRI) and non-SSRI antidepressants were associated with an increased risk for subsequent dementia over up to 18 years of follow-up in another study (Wang et al., 2016). Antidepressant drug intake was also associated cross-sectionally with an increased brain atrophy in an elderly sample without dementia (Geerlings et al., 2012). In a retrospective casecontrol study, Lee et al. (2016) found that most antidepressant drugs (e.g., SSRI, monoamine oxidase inhibitors (MAOI), heterocyclic antidepressants) were associated with a 1.5 to 2.5 fold increased risk for dementia, whereas TCA, surprisingly, were associated with a decreased risk for subsequent dementia. Kessing et al. (2009) showed in a register-based study that the intake of antidepressants (especially SSRI but also other classes) was associated with an increased risk for dementia, but the risk decreased with an increasing number of prescriptions albeit it remained to be higher compared to non-users. On the other hand, some authors hypothesized that antidepressants might contribute to the dementia risk reduction (e.g., Bali et al., 2016) by increasing brain-derived neurotrophic factor (BDNF) levels (e.g., Caraci et al., 2010) or hippocampal neurogenesis (e.g., Dranovsky & Hen, 2006). Hence, the results of the existing studies on the association between the intake of antidepressant drugs and subsequent risk for dementia are inconclusive. Few studies found no association or a 3

reduced risk for subsequent dementia in case of antidepressant use. An association between depression and an increased risk for subsequent dementia was found in several studies (e.g., Ownby et al., 2006; Diniz et al., 2013) and depression itself was considered as a risk factor for or a prodrome of dementia. One might argue that a successful treatment of depression for example by the means of antidepressants might reduce the risk for subsequent dementia, if depression would be considered as a true risk factor for dementia. Several studies found an association between anticholinergic antidepressants (mainly TCA) and an increased risk for subsequent dementia, but there are also findings suggesting that SSRI might be associated with an increased subsequent dementia risk. The age of participants, the limitations of study designs, the length of follow-ups, and the classification of antidepressants might account for different results. Studies that included old and very old participants are rather scarce. Additionally, studies on the association between antidepressant drug intake and risk for subsequent dementia sometimes did not control for the severity of depressive symptoms and, even more frequently, did not control for the cognitive status. However, assessments of both depressive symptoms and cognitive status are necessary from our point of view. If an association between the intake of antidepressant drugs and the risk for subsequent dementia would disappear after the statistical control for depressive symptoms, this might indicate that patients with more depressive symptoms receive antidepressants more frequently and depression severity, but not antidepressants might explain a possible association. Additionally, several studies found (also in the present study cohort) that depression can have a prodromal association with subsequent dementia (e.g., Li et al., 2011; in AgeCoDe cohort: Heser et al., 2013). Thus, an association between the intake of antidepressant drugs and subsequent dementia might also be explained by mild cognitive deficits due to an early dementia process but not due to antidepressants. Therefore, the aim of our study was to investigate the association between the use of antidepressant drugs and subsequent dementia also after adjusting for depressive symptoms and global cognition besides other relevant covariates in an elderly study cohort using a longitudinal study design with long-term follow-up periods. We hypothesized that the intake of antidepressant drugs, primarily those considered as PIM by the Priscus list (Holt et al., 2010), would increase the risk for subsequent dementia. The results of our study might partly contribute to the validation of the Priscus list suggested by others (e.g., Amann et al., 2012). METHODS Sample We analyzed data from a German prospective multi-center study (the German Study on Ageing, Cognition, and Dementia in Primary Care Patients (AgeCoDe) and the Study on needs, health service use, costs and health-related quality of life in a large sample of oldest-old primary care patients (85+) (AgeQualiDe), which is a continuation (follow-up 7-9) and extension of AgeCoDe). Participants at least 75 years of age were recruited at baseline in 2003/2004 via general practitioners (GPs) in six German cities (Bonn, Düsseldorf, Hamburg, Leipzig, Mannheim, and Munich). Written informed consent was provided prior to participation. The study has been approved by the local ethics committees of all participating study centers and complied with the ethical standards of the Declaration of Helsinki as revised 1989. The interviews were conducted in person by trained research assistants. The sample initially consisted of 3327 participants. Follow-up assessments after baseline were conducted every 18 months until follow-up 6 and every 10 months afterwards until follow-up 8. Data of covariates and predictors assessed at baseline were used to predict the dementia status until follow-up 8 in survival analyses. Follow-up 8 was conducted approximately 12 years after baseline. 4

Subjects who received a dementia diagnosis at baseline (n = 70) and subjects without information about their dementia status at any assessment (n = 18) were excluded. Sample characteristics are given in table 1. -- Table 1 here -Covariates In the adjusted analyses, we controlled for age at baseline (in years), sex (male and female), and education according to Casmin (König et al., 1988) as demographical covariates. Depressive symptoms at baseline were entered in a second step as a measure of depression severity. They were assessed using the German short version of the Geriatric Depression Scale (GDS; Sheikh & Yesavage, 1986; Gauggel & Birkner, 1999) containing 15 items that can be affirmed or negated. The MiniMental State Examination (MMSE; Folstein et al., 1975) at baseline was used to statistically control for global cognitive functioning in a final step, due to existing evidence that late-life depression can be a prodromal state of dementia. Antidepressant drugs and PIM The documentation of drug intake at baseline was used in the present study. Antidepressant drugs were defined by Anatomical Therapeutic Chemical Classification System (ATC) codes that started with N06A (N06AA = non-selective monoamine reuptake inhibitors, N06AB = SSRI, N06AF = MAOI (non-selective), N06AG = monoamine oxidase A inhibitors, N06AX = other antidepressants; see table 2). Imipramine (N06AA02), clomipramine (N06AA04), trimipramine (N06AA06), amitriptyline (N06AA09), doxepine (N06AA12), maprotiline (N06AA21), fluoxetine (N06AB03), and tranylcypromine (N06AF04) were considered as PIM antidepressant drugs according to the Priscus list by Holt et al. (2010). The other antidepressants listed in table 2 were considered as non-PIM. The six non-selective monoamine reuptake inhibitors (N06AA) constituted almost all cases of PIM antidepressants in our sample. Two (no antidepressants and antidepressants) and three (no antidepressants, only non-PIM antidepressants, and at least one PIM antidepressant) groups of antidepressant intake were generated. -- Table 2 here -Dementia diagnoses The dementia assessment was based on the Structured Interview for Diagnosis of Dementia of Alzheimer type, Multi-infarct Dementia and Dementia of other Aetiology according to DSM-III-R, DSM-IV, and ICD-10 (SIDAM; Zaudig & Hiller, 1996) implemented by a trained research assistant. When SIDAM could not be assessed, a Global Deterioration Scale (Reisberg et al., 1982) score of at least 4 and/or a Blessed Dementia Rating Scale (Blessed et al., 1968) score were used. Cases of incident dementia were validated by a geriatric expert. Alzheimer´s disease dementia (AD) was diagnosed according to the DSM-IV criteria (APA, 1994). Vascular dementia (VD) was diagnosed according to the NINDS-AIREN criteria (Román et al., 1993). All-cause dementia included AD, vascular dementia, mixed dementia, other specific forms of dementia (such as Lewy body dementia or dementia caused by substance abuse), and not specified cases. The presence or absence of dementia was diagnosed at every assessment. Incident dementia cases from follow-up 1 until follow-up 8 were considered, whereas prevalent cases at baseline were excluded. Analyses were confined to all-cause 5

dementia as outcome variable due to the rather small number of antidepressant drug intake cases and due to the rather small group sample sizes of specific dementia aetiologies. Statistical analyses Chi-square tests of independence were performed to examine the distribution of the three antidepressant groups (i.e., participants without antidepressant drugs, with non-PIM antidepressants only, and with at least one PIM antidepressant drug) relating to the categorical covariates (i.e., sex and education). Group differences of continuous covariates (i.e., age, GDS, and MMSE at baseline) between participants of the three antidepressant drug groups were tested with an analysis of variance (ANOVA) using Scheffé post hoc criterion in case of significant group differences. KaplanMeier survival curves were compared for participants with and without antidepressant drug intake to determine whether there were differences in the survival distributions regarding the development of all-cause dementia; log-rank tests were computed to test for statistical significance between the groups (2 groups: antidepressants yes vs. no; 3 groups: no antidepressants, non-PIM antidepressants only, PIM antidepressants). Cox proportional hazards models were used to investigate the association between antidepressant drug intake and covariates assessed at baseline and the time until all-cause dementia was diagnosed (from follow-up 1 as the earliest data point to follow-up 8 as the latest data point) in crude and adjusted models. Hazard ratios (HR) and 95% confidence intervals (CI) were computed. Adjusted hazard ratios were computed in a three-stage approach. Demographical covariates (i.e., age, sex, and education) were entered to the crude model at first (model 2), followed by depressive symptoms (model 3) in a second and global cognitive function (model 4) in a third step. All-cause dementia was used as outcome variable in proportional hazards models. Level of significance was set to < .05. RESULTS The three groups of antidepressant drug intake did not differ in their educational level (χ2 (4, N = 3127) = 7.37, p = .118) and did not differ in age (F(2,3124) = 2.41, p = .090). Sex showed a significant association (χ2 (2, N = 3127) = 9.58, p = .008). More specifically, PIM antidepressant intake was more frequent in females (p = .002) and no antidepressant intake was more frequent in males (p = .016). Depressive symptoms (F(2,3120) = 37.03, p < .001) and MMSE (F(2,3122) = 6.10, p = .002) were significantly different in the three groups. Post hoc analyses using the Scheffé post hoc criterion for significance indicated that depressive symptoms were higher in participants with antidepressant drug intake (non-PIM: M = 3.90, p < .001; PIM: M = 3.48, p < .001) than in participants without antidepressant drug intake (M = 2.18), but participants with non-PIM and with PIM drug intake did not differ in their depressive symptoms score (p = .462). Post hoc analyses also indicated that MMSE baseline scores were lower in participants with PIM antidepressant drug intake compared to participants without antidepressant drug intake (PIM: M = 26.86, no antidepressants: M = 27.49; p = .003; non-PIM: M = 27.27), whereas the other group comparisons of MMSE scores did not differ significantly. Of 3111 cases available for further analyses, 186 participants were taking antidepressant drugs at baseline. 79 individuals were taking non-PIM antidepressants and 107 were taking PIM antidepressants. Figure 1 shows Kaplan-Meier survival curves for all-cause dementia by antidepressant drug intake using two and three groups. The log-rank tests indicated significant differences in the survival distribution between participants with and without antidepressant drug 6

intake. The survival distributions for two (χ2(1) = 6.367, p = .012) and three antidepressant drug groups were statistically different (χ2(2) = 9.482, p = .009) indicating that the intake of antidepressants reduced the duration of the dementia-free status. -- Figure 1 here -In the crude proportional hazards model (see table 3, model 1), the intake of antidepressant drugs irrespective of their PIM status was associated with a significantly increased risk for subsequent allcause dementia (by about 42%). After adjusting for demographical covariates (i.e., age, sex, and education) in a first step, the intake of antidepressant drugs was still associated with a significantly increased risk for subsequent all-cause dementia (by about 53%; see table 3, model 2), but this association disappeared after depressive symptoms and global cognitive status were successively controlled for (see table 3, models 3 and 4). -- Table 3 here -The PIM status of antidepressants was considered in the following proportional hazards models. Three groups were generated (group 1: subjects without antidepressant drug intake (= reference group), group 2: subjects who took only antidepressants not considered as PIM, group 3: subjects who took antidepressants considered as PIM). The intake of non-PIM antidepressants was not associated with an increased risk for all-cause dementia, whereas the intake of PIM antidepressants was associated with an increased risk for all-cause dementia by a factor of about 1.7 in crude analyses (see table 4, model 1) and also after demographical covariates were controlled for (see table 4, model 2). After the inclusion of depressive symptoms (model 3), the association between the intake of PIM antidepressants and all-cause dementia decreased and disappeared after the inclusion of the global cognitive status (model 4). -- Table 4 here -DISCUSSION Overall, the intake of antidepressant drugs was associated with an increased risk for subsequent allcause dementia. However, more detailed analyses revealed that this was only true for antidepressants defined as potentially inappropriate by the German Priscus list (Holt et al., 2010), with almost all PIM antidepressants being non-selective monoamine reuptake inhibitors. This finding is in accordance with earlier studies on the association between the intake of drugs with anticholinergic effects and the subsequent risk for dementia (e.g., Carrière et al., 2009; Jessen et al., 2010; Gray et al., 2015). Controlling for demographical covariates did not markedly alter associations, but after the inclusion of depressive symptoms and especially after the inclusion of global cognition, the association between antidepressant PIM intake and subsequent risk for all-cause dementia disappeared. Unfortunately, we could not determine whether a long-term intake of PIM antidepressants before baseline might have elicited mild cognitive deficits at baseline as the former duration of drug intake was not documented in our study. Other studies found that especially a continuous intake of anticholinergic drugs was associated with subsequent cognitive decline and dementia (cf. Carrière et al., 2009; also cf. Papenberg et al., 2017). Although a decrease of general PIM prescriptions was observed in the AgeCoDe cohort over 4.5 years by Zimmermann et al. (2013), a significant decrease 7

of PIM antidepressants was not found. Qualitative studies also reported that patients express reservation against the cessation of antidepressant prescription (Verbeek-Heida & Mathot, 2006; Leydon et al., 2007). Hence, one might speculate about previous long-term PIM antidepressant intake in case of reservation against cessation before the baseline assessment of our study that could have contributed to the development of mild global cognitive deficits. In accordance with this hypothesis, other studies found that the intake of (anticholinergic) antidepressant drugs was associated with higher levels of brain atrophy and worse cognitive performance in elderly cognitively normal participants or participants without dementia (e.g., Geerlings et al., 2012; Risacher et al., 2016). Other studies found dose-response relationships indicating an association between an increased dementia risk in the presence of higher anticholinergic levels (e.g., Jessen et al., 2010; Gray et al., 2015). The study of Jessen et al. (2010) was conducted in the same study cohort as the present study and our results partially replicate and extend their findings as longer follow-up periods were available by now and we here studied a subset of anticholinergic drugs (i.e., antidepressants) controlling for baseline cognition. Our results are not in accordance with other studies that found an association between an increased risk of subsequent dementia and non-TCA antidepressants, for example SSRI (e.g., Kessing et al., 2009; Wang et al., 2016). Our results also contradict with a methodologically weaker retrospective case-control study that found a decreased dementia risk by (anticholinergic) TCA, whereas other antidepressant substance groups increased the dementia risk (Lee et al., 2016). It might be that physicians adapt the prescription of antidepressants to the cognitive status of elderly patients due to anticholinergic effects of TCA. Therefore, to find a dementia risk decrease by TCA might reflect that patients with cognitive impairment or dementia would receive TCA prescriptions less frequently. Following this hypothesis of conservative TCA prescribing in the presence of cognitive deficits might strengthen the reliance of our finding that anticholinergic PIM drugs were associated with an increased dementia risk. According to Mulsant et al. (2014), experienced psychiatrists often believe that TCA possess higher efficacy than SSRI, which might influence the practice of prescription as well despite cognitive concerns. Associations between depression, especially in late-life, and an increased risk for subsequent dementia were found in several studies (e.g., Li et al., 2011; Diniz et al., 2013; Heser et al., 2013) and depression itself was considered as a risk factor or a prodrome of dementia. In line with these results, higher levels of depressive symptoms were also associated with an increased risk for subsequent dementia in the present analyses. Studies regarding modifiable risk factors of dementia that estimate population attributable risk and the effects of risk factor reduction should keep potential costs of depression treatment in mind. Our study has several strengths. We provide clinical evidence that contributes to the validation of the Priscus list by Holt et al. (2010), which was derived from an expert consultation. Our study sample was rather large and longitudinal assessments also including depressive symptoms and cognitive tests were conducted over up to 12 years. Our participants were very old so that we can add valuable and additional evidence to the field as former studies often included younger participants. As reported in other studies, we also found that the intake of PIM antidepressants was higher in women (Amann et al., 2012). Beside these strengths, there are also limitations of our study. Our epidemiological cohort sample was rather small compared to register-based studies and we did not include the risk for other detrimental health outcomes such as falls or glaucoma aggravation in the present study that are also part of the expert justification of the Priscus list (Holt et al., 2010) and might also be due to 8

anticholinergic drug effects. Furthermore, drugs are considered as PIM due to associations that were found in group studies, but there might be individuals that tolerate particular PIM drugs well and might grow old with them although others might dispose the respective drug due to side effects or might even drop out of the study sample due to severe adverse drug effects. To avoid this potential bias, future studies should record the history of drug intake over longer periods of observation. A qualitative study in a subsample of the present cohort provided patient- and physician-related factors that might contribute to the chronical use of PIM in general (cf. Pohontsch et al., 2017). About 6% of our sample took antidepressant drugs at baseline. This proportion was very similar to another study that reported an antidepressant ratio of 6.2% (Kaup et al., 2016). Other authors reported ratios of about 14% (Mulsant et al., 2014), but this estimation also included younger persons of at least 60 years in the USA. Although we considered antidepressant PIM drug intake, the effects of other PIM drugs were not examined. The assessment of drugs was based on conscientious inquiries, but the quality of the data also depended on the ability and the motivation of the interviewees to give a complete overview of their medication. Participants of older age cohorts might be more embarrassed about the intake of antidepressants, but this should result in a lower general report frequency without a bias towards PIM or non-PIM antidepressants. Dementia-free participants with worse baseline global cognition should have had an increased risk for subsequent dementia and might have had more incomplete reports on their medications. However, this bias should have led to an underestimation of the association that we expected. The duration of drug intake and the drug dosages were not considered in the present study. Antidepressant drugs are also prescribed due to other reasons than depression (i.e., insomnia or pain); unfortunately, this information was not assessed in our study, but it was statistically controlled for depressive symptoms. However, current antidepressant treatment, response or non-response to it, or the efficacy of treatment might have influenced depressive symptoms, leading to difficulties with the interpretation of this score. Further studies with larger sample sizes might be helpful to investigate the association between different antidepressant drug groups (e.g., TCA) or antidepressant drug agents (e.g., amitriptyline) to derive potential mechanisms that mediate the association between PIM antidepressant drug intake and the subsequent risk for dementia. We cannot completely rule out that the association between PIM antidepressant intake and subsequent dementia was influenced by the type of depression, although we statistically controlled for depressive symptoms. To summarize our results, we found that not antidepressants per se, but only PIM antidepressant drugs were associated with an increased risk for subsequent all-cause dementia. This association disappeared after global cognitive abilities at baseline were statistically controlled for. This might be a result of deleterious effects of pre-existing long-term anticholinergic antidepressant drug intake on cognitive baseline performance. Our results can provide valuable information for health services and physicians who prescribe antidepressant drugs to elderly and very old patients. As only PIM antidepressants and more specifically mainly TCA with anticholinergic properties increased the risk for subsequent dementia, physicians should avoid their initial prescription or discontinue already existing PIM antidepressants. Non-PIM antidepressants were not associated with a higher risk for subsequent dementia and should, therefore, be selected instead under the consideration of risks and benefits.

FUNDING 9

This publication is part of the German Research Network on Dementia (KND), the German Research Network on Degenerative Dementia (KNDD; German Study on Ageing, Cognition and Dementia in Primary Care Patients; AgeCoDe), and the Health Service Research Initiative (Study on Needs, health service use, costs and health-related quality of life in a large sample of oldest-old primary care patients (85+; AgeQualiDe)) and was funded by the German Federal Ministry of Education and Research (grants KND: 01GI0102, 01GI0420, 01GI0422, 01GI0423, 01GI0429, 01GI0431, 01GI0433, 01GI0434; grants KNDD: 01GI0710, 01GI0711, 01GI0712, 01GI0713, 01GI0714, 01GI0715, 01GI0716; grants Health Service Research Initiative: 01GY1322A, 01GY1322B, 01GY1322C, 01GY1322D, 01GY1322E, 01GY1322F, 01GY1322G). The funding source was not involved in the conduct of the research or the preparation of the article. ACKNOWLEDGMENTS We want to thank both all participating patients and their general practitioners for their good collaboration. Members of the AgeCoDe & AgeQualiDe Study Group: Wolfgang Maier (Principal Investigator), Martin Scherer (Principal Investigator), Steffi G. Riedel-Heller (Principal Investigator), Heinz-Harald Abholz, Christian Brettschneider, Cadja Bachmann, Horst Bickel, Wolfgang Blank, Sandra EifflaenderGorfer, Marion Eisele, Annette Ernst, Angela Fuchs, André Hajek, Kathrin Heser, Frank Jessen, Hanna Kaduszkiewicz, Teresa Kaufeler, Mirjam Köhler, Hans-Helmut König, Alexander Koppara, Diana Lubisch, Tobias Luck, Dagmar Lühmann, Melanie Luppa, Tina Mallon, Manfred Mayer, Edelgard Mösch, Michael Pentzek, Jana Prokein, Alfredo Ramirez, Susanne Roehr, Anna Schumacher, Janine Stein, Susanne Steinmann, Franziska Tebarth, Hendrik van den Bussche (Principal Investigator 20022011), Carolin van der Leeden, Michael Wagner, Klaus Weckbecker, Dagmar Weeg, Jochen Werle, Siegfried Weyerer, Birgitt Wiese, Steffen Wolfsgruber, Thomas Zimmermann. AUTHOR CONTRIBUTION Kathrin Heser was involved in the data acquisition, undertook the statistical analysis and wrote the first draft of the manuscript. Michael Wagner contributed to the conception of the study, the data interpretation and revisited the work critically for important intellectual content. Tobias Luck was involved in the study organisation, the data management and revisited the work critically for important intellectual content. Susanne Röhr, Edelgard Mösch, Jochen Werle, Christian Brettschneider, and Angela Fuchs were involved in the data acquisition and revisited the work critically for important intellectual content. Birgitt Wiese managed the data, was involved in the data analysis, and revisited the work critically for important intellectual content. Hanna Kaduszkiewicz was involved in the earlier study organisation and revisited the work critically for important intellectual content. Anke Oey was involved in the data management and revisited the work critically for important intellectual content. Wolfgang Maier, Martin Scherer, Hendrik van den Bussche, and Steffi G. Riedel-Heller designed the study as Principal Investigators and revisited the work critically for important intellectual content. Horst Bickel, Siegfried Weyerer, Hans-Helmut König, and Michael Pentzek contributed to the design of the study and revisited the work critically for important intellectual content. All authors contributed to and have approved the final manuscript. 10

CONFLICT OF INTEREST Horst Bickel has received research support from Organon and Dr. Wilmar Schwabe and fees for lectures from Bayer, Merz, Lundbeck, Dr. Wilmar Schwabe, and Wyeth. Siegfried Weyerer has received research grants or draws a fee for speech from the following companies: AstraZeneca, Bristol-Myers Squibb, Eli Lilly, Janssen Cilag, Novartis, Pfizer, and Wyeth. Wolfgang Maier has received research grants respectively from, is member of the Advisory Boards, or draws a fee for speech from the following companies: AstraZeneca, Bristol-Myers Squibb, Eli Lilly, Janssen Cilag, Lundbeck, Merck, Pfizer, Sanofi Aventis, Schering, and Böhringer. The other authors declare that they do not have any competing interests.

REFERENCES Amann, U., Schmedt, N., Garbe, E., 2012. Prescribing of potentially inappropriate medications for the elderly: an analysis based on the PRISCUS list. Dtsch. Artzebl. Int. 109, 69-75. APA, 1994. DSM-IV: Diagnostic and Statistical Manual of Mental Disorders, fourth ed., American Psychological Association, Washington DC. Bali, V., Holmes, H.M., Johnson, M.L., Chen, H., Fleming, M.L., Aparasu, R.R., 2016. Comparative effectiveness of second-generation antidepressants in reducing the risk of dementia in elderly nursing home residents with depression. Pharmacotherapy. 36, 38-48. Blessed, G., Tomlinson, B.E., Roth, M., 1968. The association between quantitative measures of dementia and of senile change in the cerebral grey matter of elderly subjects. Br. J. Psychiatry. 114, 797-811. Caraci, F., Copani, A., Nicoletti, F., Drago, F., 2010. Depression and Alzheimer´s disease: neurobiological links and common pharmacological targets. Eur. J. Pharmacol. 626, 64-71. Carrière, I., Fourrier-Reglat, A., Dartigues, J.F., Rouaud, O., Pasquier, F., Ritchie, K., Ancelin, M.L., 2009. Drugs with anticholinergic properties, cognitive decline, and dementia in an elderly general population: the 3-city study. Arch. Intern. Med. 169, 1317-1324. Diniz, B.S., Butters, M.A., Albert, S.M., Dew, M.A., Reynolds, C.F.3rd, 2013. Late-life depression and risk of vascular dementia and Alzheimer´s disease: systematic review and meta-analysis of community-based cohort studies. Br. J. Psychiatry. 202, 329-335. Dranovsky, A., Hen, R., 2006. Hippocampal neurogenesis: regulation by stress and antidepressants. Biol. Psychiatry. 59, 1136-1143. Folstein, M.F., Folstein, S.E., McHugh, P.R., 1975. “Mini-Mental State“. A practical method for grading the cognitive state of patients for the clinician. J. Psychiatr. Res. 12, 189-198. 11

Fuhrer, R., Dufouil, C., Dartigues, J.F., 2003. Exploring sex differences in the relationship between depressive symptoms and dementia incidence: prospective results from the PAQUID Study. J. Am. Geriatr. Soc. 51, 1055-1063. Gauggel, S., Birkner, B., 1999. Validity and reliability of a German version of the Geriatric Depression Scale (GDS). Z. Klin. Psychol. Forsch. Praxis. 28, 18-27. Geerlings, M.I., Brickman, A.M., Schupf, N., Devanand, D.P., Luchsinger, J.A., Mayeux, R., Small, S.A., 2012. Depressive symptoms, antidepressant use, and brain volumes on MRI in a population-based cohort of old persons without dementia. J. Alzheimers. Dis. 30, 75-82. Gray, S.L., Anderson, M.L., Dublin, S., Hanlon, J.T., Hubbard, R., Walker, R., Yu, O., Crane, P.K., Larson, E.B., 2015. Cumulative use of strong anticholinergics and incident dementia: a prospective cohort study. JAMA Intern. Med. 175, 401-407. Heser, K., Tebarth, F., Wiese, B., Eisele, M., Bickel, H., Köhler, M., Mösch, E., Weyerer, S., Werle, J., König, H.H., Leicht, H., Pentzek, M., Fuchs, A., Riedel-Heller, S.G., Luppa, M., Prokein, J., Scherer, M., Maier, W., Wagner, M., AgeCoDe Study Group, 2013. Age of major depression onset, depressive symptoms, and risk for subsequent dementia: results of the German study on Ageing, Cognition, and Dementia in Primary Care Patients (AgeCoDe). Psychol. Med. 43, 1597-1610. Holt, S., Schmiedl, S., Thürmann, P.A., 2010. Potentially inappropriate medications in the elderly: the PRISCUS list. Dtsch. Arztebl. Int. 107, 543-551. Jessen, F., Kaduszkiewicz, H., Daerr, M., Bickel, H., Pentzek, M., Riedel-Heller, S., Wagner, M., Weyerer, S., Wiese, B., van den Bussche, H., Broich, H., Maier, W., 2010. Anticholinergic drug use and risk for dementia: target for dementia prevention. Eur. Arch. Psychiatry. Clin. Neurosci. 260(Suppl 2), S111-115. Jobski, K., Schmedt, N., Kollhorst, B., Krappweis, J., Schink, T., Garbe, E., 2017. Characteristics and drug use patterns of older antidepressant initiators in Germany. Eur. J. Clin. Pharmacol. 73, 105-113. Kaup, A.R., Byers, A.L., Falvey, C., Simonsick, E.M., Satterfield, S., Ayonayon, H.N., Smagula, S.F., Rubin, S.M., Yaffe, K., 2016. Trajectories of depressive symtpoms in older adults and risk of dementia. JAMA Psychiatry. 73, 525-531. Kessing, L.V., Søndergård, L., Forman, J.L., Andersen, P.K., 2009. Antidepressants and dementia. J. Affect. Disord. 117, 24-29. König, W., Lüttinger, P., Müller, W., 1988. A comparative analysis of the development and structure of educational systems: Methodological foundations and the construction of a comparative education scale, CASMIN Working Paper 12. University Mannheim, Mannheim, Germany. Lau, D.T., Kasper, J.D., Potter, D.E., Lyles, A., Bennett, R.G., 2005. Hospitalization and death associated with potentially inappropriate medication prescriptions among elderly nursing home residents. Arch. Intern. Med. 165, 68-74. Lee, C.W., Lin, C.L., Sung, F.C., Liang, J.A., Kao, C.H., 2016. Antidepressant treatment and risk of dementia: a population-based, retrospective case-control study. J. Clin. Psychiatry. 77, 117-122. 12

Leydon, G.M., Rodgers, L., Kendrick, T., 2007. A qualitative study of patient views on discontinuing long-term selective serotonin reuptake inhibitors. Fam. Pract. 24, 570-575. Li, G., Wang, L.Y., Shofer, J.B., Thompson, M.L., Peskind, E.R., McCormick, W., Bowen, J.D., Crane, P.K., Larson, E.B., 2011. Temporal relationship between depression and dementia: findings from a large community-based 15-year follow-up study. Arch. Gen. Psychiatry. 68, 970-977. Mulsant, B.H., Blumberger, D.M., Ismail, Z., Rabheru, K., Rapoport, M.J., 2014. A systematic approach to pharmacotherapy for geriatric major depression. Clin. Geriatr. Med. 30, 517-534. Ownby, R.L., Crocco, E., Acevedo, A., John, V., Loewenstein, D., 2006. Depression and risk for Alzheimer disease: systematic review, meta-analysis, and metaregression analysis. Arch. Gen. Psychiatry. 63, 530-538. Papenberg, G., Bäckman, L., Fratiglioni, L., Laukka, E.J., Fastbom, J., Johnell, K., 2017. Anticholinergic drug use is associated with episodic memory decline in older adults without dementia. Neurobiol. Aging. 55, 27-32. Pohontsch, N.J., Heser, K., Löffler, A., Haenisch, B., Parker, D., Luck, T., Riedel-Heller, S.G., Maier, W., Jessen, F., Scherer, M., 2017. General practitioners´ views on (long-term) prescription and use of problematic and potentially inappropriate medication for oldest-old patients – A qualitative interview study with GPs (CIM-TRIAD study). BMC Fam. Pract. 18:22. Reisberg, B., Ferris, S.H., de Leon, M.J., Crook, T., 1982. The Global Deterioration Scale for assessment of primary degenerative dementia. Am. J. Psychiatry. 139, 1136-1139. Risacher, S.L., McDonald, B.C., Tallman, E.F., West, J.D., Farlow, M.R., Unverzagt, F.W., Gao, S., Boustani, M., Crane, P.K., Petersen, R.C., Jack, C.R. Jr., Jagust, W.J., Aisen, P.S., Weiner, M.W., Saykin, A.J., Alzheimer´s Disease Neuroimaging Initiative, 2016. Association between anticholinergic medication use and cognition, brain metabolism, and brain atrophy in cognitively normal older adults. JAMA Neurol. 73, 721-732. Román, G.C., Tatechimi, T.K., Erkinjuntti, T., Cummings, J.L., Masdeu, J.C., Garcia, J.H., Amaducci, L., Orgogozo, J.M., Brun, A., Hofman, A., Moody, D.M., O´Brien, M.D., Yamaguchi, T., Grafman, J., Drayer, B.P., Bennett, D.A., Fisher, M., Ogata, J., Kokmen, E., Bermejo, F., Wolf, P.A., Gorelick, P.B., Bick, K.L., Pajeau, A.K., Bell, M.A., DeCarli, C., Culebras, A., Korczyn, A.D., Bogousslavsky, J., Hartmann, A., Scheinberg, P., 1993. Vascular dementia: diagnostic criteria for research studies. Report of the NINDS-AIREN International Workshop. Neurology. 43, 250-260. Saczynski, J.S., Rosen, A.B., McCammon, R.J., Zivin, K., Andrade, S.E., Langa, K.M., Vijan, S., Pirraglia, P.A., Briesacher, B.A., 2015. Antidepressant use and cognitive decline: The Health and Retirement Study. Am. J. Med. 128, 739-746. Schneeweiss, S., Hasford, J., Göttler, M., Hoffmann, A., Riethling, A.K., Avorn, J., 2002. Admissions caused by adverse drug events to internal medicine and emergency departments in hospitals: a longitudinal population-based study. Eur. J. Clin. Pharmacol. 58, 285-291.

13

Sheikh, J.I., Yesavage, J.A., 1986. Geriatric Depression Scale (GDS): Recent evidence and development of a shorter version, in: Brink, T.L. (Ed.), Clinical Gerontology: A Guide to Assessment and Intervention. Haworth, New York, pp. 165-173. Verbeek-Heida, P.M., Mathot, E.F., 2006. Better safe than sorry – why patients prefer to stop using selective serotonin reuptake inhibitor (SSRI) antidepressants but are afraid to do so: results of a qualitative study. Chronic Illn. 2, 133-42. Wang, C., Gao, S., Hendrie, H.C., Kesterson, J., Campbell, N.L., Shekhar, A., Callahan, C.M., 2016. Antidepressant use in the elderly is associated with an increased risk of dementia. Alzheimer. Dis. Assoc. Disord. 30, 99-104. Wucherer, D., Eichler, T., Hertel, J., Kilimann, I., Richter, S., Michalowsky, B., Thyrian, J.R., Teipel, S., Hoffmann, W., 2017. Potentially inappropriate medication in community-dwelling primary care patients who were screened positive for dementia. J. Alzheimers. Dis. 55, 691-701. Zaudig, M., Hiller, W., 1996. SIDAM Handbook. Structured Interview for Diagnosis of Dementia of Alzheimer type, Multi-infarct Dementia and Dementia of other Aetiology according to DSM-III-R, DSM-IV, and ICD-10 [German version], Huber, Bern. Zimmermann, T., Kaduszkiewicz, H., van den Bussche, H., Schön, G., Brettschneider, C., König, H.H., Wiese, B., Bickel, H., Mösch, E., Luppa, M., Riedel-Heller, S., Werle, J., Weyerer, S., Fuchs, A., Pentzek, M., Hänisch, B., Maier, W., Scherer, M., Jessen, F., AgeCoDe-Study Group, 2013. [Potentially inappropriate medication in elderly primary care patients: a retrospective, longitudinal analysis]. Bundesgesundheitsblatt Gesundheitsforschung Gesundheitsschutz. 56, 941-949.

(A)

(B)

Figure 1. Kaplan-Meier survival curves of time to incidence of all-cause dementia from baseline to follow-up 8 by antidepressant drug intake with (A) two groups (antidepressant drug yes vs. no) and (B) three groups (no antidepressant drug, non-PIM antidepressant drug, PIM antidepressant drug).

14

Table 1: Sample characteristics with status of dementia diagnosis until follow-up 8 (up to 12 years after baseline) and status of predictors at baseline. No dementia until FUP8 Dementia until FUP8 (n = 2525) (n = 714) Age, M (SD) 79.38 (3.61) 80.49 (3.58) Sex, n (%) Female 1615 (64) 502 (70.3) Male 910 (36) 212 (29.7) Education, n (%) Low 1555 (61.6) 448 (62.7) Medium 696 (27.6) 192 (26.9) High 274 (10.8) 74 (10.4) GDS score , M (SD) 2.13 (2.25) 2.59 (2.44) Missing 2 2 MMSE, M (SD) 27.65 (1.79) 26.83 (2.11) missing 2 0 Antidepressant drug, n (%) No 2294 (90.8) 631 (88.4) Yes 131 (5.2) 55 (7.7) missing 100 (4.0) 28 (3.9) Antidepressant drug (3 groups) No 2294 (90.8) 631 (88.4) Non-PIM antidepressant 60 (2.4) 19 (2.7) PIM antidepressant 71 (2.8) 36 (5.0) missing 100 (4.0) 28 (3.9) FUP8 = follow-up 8, Classification of education according to Casmin (König et al., 1988), GDS = short version of the Geriatric Depression Scale (Gauggel & Birkner, 1999), MMSE = Mini-Mental State Examination (Folstein et al., 1975), PIM = potentially inappropriate medication.

15

Table 2: ATC codes of antidepressant drugs (N06A) including the active substances and their PIM status according to the German Priscus list (Holt et al., 2010). ATC code Active substance PIM = yes N06AA Non-selective monoamine reuptake inhibitors N06AA01 Desipramine N06AA02 Imipramine yes N06AA03 Imipramine oxide N06AA04 Clomipramine yes N06AA05 Opipramol N06AA06 Trimipramine yes N06AA07 Lofepramine N06AA08 Dibenzepine N06AA09 Amitriptyline yes N06AA10 Nortriptyline N06AA11 Protriptyline N06AA12 Doxepine yes N06AA13 Iprindole N06AA14 Melitracen N06AA15 Butriptyline N06AA16 Dusolepin N06AA17 Amoxapine N06AA18 Dimetacrine N06AA19 Amineptine N06AA21 Maprotiline yes N06AA23 Quinupramine N06AB Selective serotonin reuptake inhibitors N06AB02 Zimelidine N06AB03 Fluoxetine yes N06AB04 Citalopram N06AB05 Paroxetine N06AB06 Sertraline N06AB07 Alaproclate N06AB08 Fluvoxamine N06AB09 Etoperidone N06AB10 Escitalopram N06AF Monoamine oxidase inhibitors, non-selective N06AF01 Isocarboxazid N06AF02 Nialamide N06AF03 Phenelzine N06AF04 Tranylcypromine yes N06AF05 Iproniazide N06AF06 Iproclozide N06AG Monoamine oxidase A inhibitors N06AG02 Moclobemide N06AG03 Toloxatone N06AX other antidepressants N06AX01 Oxitriptan N06AX02 Tryptophan N06AX03 Mianserin N06AX04 Nomifensine N06AX05 Trazodone N06AX06 Nefazodone 16

N06AX07 Minaprine N06AX08 Bifemelane N06AX09 Viloxazine N06AX10 Oxaflozane N06AX11 Mirtazapine N06AX12 Bupropion N06AX13 Medifoxamine N06AX14 Tianeptine N06AX15 Pivagabine N06AX16 Venlafaxine N06AX17 Milnacipran N06AX18 Reboxetine N06AX19 Gepirone N06AX21 Duloxetine N06AX22 Agomelatine N06AX23 Desvenlafaxine N06AX24 Vilazodone N06AX25 Hyperici herba N06AX26 Vortioxetine QN06AX90 Selegiline ATC = Anatomical Therapeutic Chemical Classification System.

17

Table 3: Prediction of subsequent all-cause dementia until follow-up 8 by antidepressant drug intake in two groups (yes vs. no) and covariates at baseline in proportional hazards models. All-cause dementia Model 1 Model 2 Model 3 Model 4 n = 3093 n = 3093 n = 3089 n = 3087 HR (95% CI) p HR (95% CI) p HR (95% CI) p HR (95% CI) p Antidepressant 1.42 .012 1.53 .003 1.30 .068 1.17 .272 intake (1.08-1.88) (1.16-2.02) (0.98-1.72) (0.88-1.55) Age 1.13 .000 1.13 .000 1.11 .000 (1.11-1.15) (1.10-1.15) (1.09-1.13) Sex 0.93 .375 0.97 .727 0.97 .753 (0.78-1.10) (0.82-1.15) (0.82-1.15) Education .060 .116 .275 Education (1) 0.82 .026 0.84 .045 1.03 .708 (0.69-0.98) (0.71-1.00) (0.87-1.24) Education (2) 0.85 .214 0.88 .348 1.25 .108 (0.65-1.10) (0.68-1.15) (0.95-1.64) Depressive 1.11 .000 1.09 .000 symptoms (1.08-1.15) (1.06-1.13) Global 0.77 .000 cognition (0.74-0.80) HR = hazard ratio, CI = confidence interval, antidepressant drug intake = yes or no, classification of education according to Casmin (König et al., 1988), low education as reference (ref.), education (1) = medium education (vs. ref.), education (2) = high education (vs. ref.), depressive symptoms = Geriatric Depression Scale (Gauggel & Birkner, 1999), global cognition = Mini-Mental State Examination (MMSE; Folstein et al., 1975).

18

Table 4: Prediction of subsequent all-cause dementia until follow-up 8 by antidepressant drug intake (3 groups) and covariates at baseline in proportional hazards models. All-cause dementia Model 1 Model 2 Model 3 Model 4 n = 3093 n = 3093 n = 3089 n = 3087 HR (95% CI) p HR (95% CI) p HR (95% CI) p HR (95% CI) p Antidepressant .010 .006 .070 .356 intake non-PIM vs. 1.10 .674 1.29 .277 1.04 .863 1.00 .989 none (0.70-1.74) (0.82-2.04) (0.66-1.66) (0.63-1.59) PIM vs. 1.68 .002 1.70 .002 1.49 .021 1.28 .151 none (1.20-2.35) (1.21-2.38) (1.06-2.10) (0.91-1.81) Age 1.13 .000 1.12 .000 1.11 .000 (1.11-1.15) (1.10-1.15) (1.09-1.13) Sex 0.93 .399 0.98 .779 0.98 .777 (0.79-1.10) (0.82-1.16) (0.82-1.16) Education .065 .130 .282 Education (1) 0.82 .029 0.84 .054 1.04 .689 (0.69-0.98) (0.71-1.00) (0.87-1.24) Education (2) 0.85 .207 0.88 .335 1.25 .111 (0.65-1.10) (0.67-1.14) (0.95-1.63) Depressive 1.11 .000 1.09 .000 symptoms (1.08-1.15) (1.06-1.13) Global 0.77 .000 cognition (0.74-0.80) HR = hazard ratio, CI = confidence interval, antidepressant drug intake: 3 groups = no intake, non-PIM antidepressant, PIM antidepressant, PIM = potentially inappropriate medication, classification of education according to Casmin (König et al., 1988), low education as reference (ref.), education (1) = medium education (vs. ref.), education (2) = high education (vs. ref.), depressive symptoms = Geriatric Depression Scale (Gauggel & Birkner, 1999), global cognition = Mini-Mental State Examination (MMSE; Folstein et al., 1975).

Potentially inappropriate antidepressants were associated with an increased subsequent dementia risk Other antidepressants were not associated with an increased subsequent dementia risk The association between antidepressants and dementia disappeared after cognition was controlled for

19