Benzodiazepine, psychotropic medication, and ...

Alzheimer’s & Dementia - (2015) 1-10

Benzodiazepine, psychotropic medication, and dementia: A population-based cohort study Dalia Shasha,b,1, Tobias Kurtha,b,1, Marion Bertranda, Carole Dufouila,b, Pascale Barberger-Gateaub,c, Claudine Berrd, Karen Ritchied, Jean-Francois Dartiguesb,c, Bernard Begaudb,e, Annick Alperovitcha,b, Christophe Tzourioa,b,* a

Inserm Research Center for Epidemiology and Biostatistics (U897) – Team Neuroepidemiology, Bordeaux, France b University of Bordeaux, Bordeaux France c Inserm Unit 897—Epidemiology and Biostatistics Research Center (U897) – Team Epidemiology and Neuropsychology of Cerebral Aging, Bordeaux, France d Inserm Unit 1061—Neuropsychiatry: Epidemiological and Clinical Research, University of Montpellier, Montpellier, France e Inserm Unit 657—Pharmacoepidemiology and Evaluation of the Impact of Health Products on Populations, Bordeaux, France

Abstract

Introduction: Benzodiazepine use has been associated with increased risk of dementia. However, it remains unclear whether the risk relates to short or long half-life benzodiazepines and whether it extends to other psychotropic drugs. Methods: Prospective cohort study among 8240 individuals 65, interviewed on medication use. Incident dementia confirmed by an end point committee after a multistep procedure. Results: During a mean of 8 years of follow-up, 830 incident dementia cases were observed. Users of benzodiazepines at baseline had a 10% increased risk of dementia (adjusted hazard ratio [HR], 1.10; 95% confidence interval, 0.90–1.34). However, long half-life (.20 hours) benzodiazepine users had a marked increased risk of dementia (HR 5 1.62; 1.11–2.37) compared with short half-life users (HR 5 1.05; 0.85–1.30). Users of psychotropics had an increased risk of dementia (HR 5 1.47; 1.16–1.86). Discussion: Results of this large, prospective study show increased risk of dementia for long half-life benzodiazepine and psychotropic use. Ó 2015 The Alzheimer’s Association. Published by Elsevier Inc. All rights reserved.

Keywords:

Cohort studies; Alzheimer’s disease; Dementia; Benzodiazepine; Psychotropic medication

The authors report a full disclosure for the last 3 years for each author: D.S. is a full-time employee at Boehringer Ingelheim, Germany since January 1, 2013. Her work on this article has been completed before starting this job. T.K. has received investigator-initiated research funding from the French National Research Agency, the US National Institutes of Health, and the Parkinson’s Disease Foundation. He has received honoraria from the BMJ and from Cephalalgia for editorial services and from the American Academy of Neurology for educational lectures. M.B. has no disclosures. C.D. has received honoraria from the American Academy of Neurology for educational lectures and from Eisai, Inc for providing methodological expertise. P.B.-G. has received funding for travel or speaker honoraria from Lesieur, Bausch & Lomb, Aprifel, Danone Institute, Canadian Association of Gerontology, the Jean Mayer Human Nutrition Research Center on Aging, Tufts University, Alzheimer’s Association, Groupe Lipides et Nutrition, Institut Pasteur, Conseil Regional d’Aquitaine; serves on the editorial boards of Disability and Rehabilitation and the Journal of Alzheimer’s disease; has received consultancy fees from Vifor Pharma; and receives research support from Danone Research, Institut Carnot LISA, and Groupe Lipides et Nutrition. C.B. has received investigator-initiated research funding from the French National Research Agency. K.R. has received research grants from the Alzheimer’s Association

UK, the French National Research Agency, and the Montpellier University Hospital. Honoraria have also been received from the University of Gothenburg for external scientific evaluation. J.-F.D. has received investigatorinitiated research funding from the Alzheimer Plan Foundation, AGRICA, IPSEN Pharma, and Novartis Pharma. He has received honoraria from IPSEN and Novartis Pharma for contributing to a scientific advisory panel. B.B. has received investigator-initiated research funding from the French Health Ministry; he is chair of the scientific committee for two pharmacoepidemiologic studies conducted by the contract research organization LA-SER (London): one on medicines used in osteoarthritis, the other on the use of homeopathic remedies by French practitioners. A.A. has received honoraria from the Fondation Plan Alzheimer and the Fondation Bettencourt-Schueller for contributing to scientific advisory committees. C.T. has received fees from the Fondation Plan Alzheimer for participating in scientific committees. He has also received investigator-initiated research funding from the French National Research Agency (ANR) and the Fondation Plan Alzheimer for the 3C study. 1 These authors contributed equally to the work. *Corresponding author. Tel./Fax: 133-5-57-57-16-59. E-mail address: [email protected]

http://dx.doi.org/10.1016/j.jalz.2015.10.006 1552-5260/Ó 2015 The Alzheimer’s Association. Published by Elsevier Inc. All rights reserved.

2

D. Shash et al. / Alzheimer’s & Dementia - (2015) 1-10

1. Introduction Benzodiazepines and psychotropic medications are among the most widely used drugs in developed countries [1]. The prevalence of use for benzodiazepines alone among patients aged 65 years is estimated to be 30% in France, 20% in Canada and Spain, and approximately 15% in Australia [2–5]. Sleep disorders, anxiety, and depressive symptoms, all of which are frequent in elderly individuals, are the main indications for the use of benzodiazepines and other psychotropic medications [6,7]. Because they are often used in the frail elderly population, the safety and effects of benzodiazepines and psychotropic drugs on cognition and dementia are of great public health interest. Both benzodiazepines and psychotropic medications interact with the neurotransmitter system [8,9], and previous studies have found that benzodiazepine use is associated with an increased risk of dementia [10–13]. Although the results of the small case-control studies that evaluated this association were inconclusive [14–16], the findings of two prospective cohort studies have suggested that the use of benzodiazepines is associated with a 1.5 to threefold increased risk of dementia [12,13]. However, several important questions remain regarding the differences between the use of benzodiazepines with long versus short half-lives [17] as well as whether the effects are limited to benzodiazepine or extend to other psychotropic drugs. We, therefore, investigated the association between benzodiazepine medication use and the risk of dementia in the Three-City study, a large, population-based cohort study, with a focus on drugs with short versus long half-lives. It also examined the effects of other psychotropic medications on dementia. 2. Methods Details about the Three-City study were published previously [18]. Briefly, the Three-City study is a populationbased longitudinal study that investigated the association between vascular factors and dementia in noninstitutionalized individuals aged 65 years starting in 1999. A total of 9294 participants were recruited at baseline (T0) from the electoral rolls of the cities of Bordeaux, Dijon, and Montpellier in France. Data were collected using a standardized questionnaire and face-to-face interviews as well as by medical examinations conducted at a medical center or hospital. Data were collected at follow-up at T2 (year 2), T4 (year 4), T6 (year 6), T7 (year 7), and T10 (year 10). The study protocol of the Three-City study was approved by the Ethics Committee of the University-Hospital of Kremlin-Bic^etre, France, and written informed consent was obtained from each participant. 2.1. Exposure, outcome, and covariates At baseline and at follow-up time points T2, T4, and T7, the participants were asked by trained interviewers whether

they had used any medication more than once a week during the last month. The interviewers followed a standardized protocol, and the prescriptions and medication packages were systematically checked during a home visit. The drugs were coded based on the World Health Organization Anatomical Therapeutic Chemical (ATC) classification system. All classes of benzodiazepines and their derivative drugs were noted, including anxiolytic (ATC code: N05BA), hypnotic and sedative (N05CD and N05CF), antiepileptic (N03AE), and myorelaxing (M03BX07) drugs. A previous population-based study in France suggested no substantially different associations between prevalent versus incident users of benzodiazepines and the risk of dementia [13]. Here, we used the prevalent users at baseline as the main exposure variable to increase the power and provide an estimation of the association at the population level. Prevalent use was defined as any report of benzodiazepine use at baseline (T0). In a secondary analysis, we used the incident users of these medications. Incident users (or initiators) were those participants who did not report any benzodiazepine or psychotropic use at baseline but who reported benzodiazepine use at T2. In addition, the benzodiazepine users were split into two groups, i.e., users of long-acting benzodiazepines (half-life .20 hours) and users of short-acting ones (half-life 20 hours). The half-life cutoff time was based on a definition proposed by the French National Agency for Drug Safety (Agence nationale de securite du medicament et des produits de sante [ANSM]) and adopted in a previous publication about medication use in the ThreeCity study [19]. The definition of the use of psychotropic medications (ATC Code: N05, N06) excluded benzodiazepine users to differentiate the two groups and also excluded users of anti-dementia medication (N06D). The medications in this definition include antipsychotics, anxiolytics, hypnotics and sedatives, antidepressants, psychostimulants, and nootropics. Prevalent users of psychotropics reported taking the medication at baseline (T0), whereas initiators were those who began taking psychotropics at T2 but who did not report any benzodiazepine or psychotropic use at T0. 2.2. Outcome assessment The diagnosis of dementia was based on a three-step procedure at each follow-up as described elsewhere [18,20]. First, trained psychologists administered a battery of neuropsychological tests. Second, a neurologist examined all the participants in Bordeaux and Montpellier. In Dijon, owing to the large number of participants, only those who screened positive for dementia using the mini-mental state examination (MMSE) [21] and the Isaacs set test [22] underwent further clinical examination. For subjects suspected of having dementia, further data on cognitive disorders and their consequences on daily activities were collected using a standardized protocol, and the study neurologist or


geriatrician established a provisional diagnosis. Third, an independent committee of neurologists and geriatricians reviewed all potential cases of dementia and reached a consensus on the diagnosis of dementia according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria [23]. Dementia classification was based on the criteria of the National Institute of Neurological and Communication Disorders and Stroke and on the criteria of the Alzheimer’s Disease and Related Disorders Association [24].

2.3. Covariates We considered covariates that were measured at inclusion (T0) for our analyses. The covariates that were considered in this study were based on an a priori basis and included age, sex, measured body mass index (,25, 25–30, and 30 kg/ m2), living status (living alone), education (four categories), self-perceived health (very good, good, medium, bad, and very bad), alcohol consumption (non-, ex-, or current drinkers), smoking (current, ever, or never), diabetes, hypertension (systolic blood pressure 140 mm Hg, diastolic blood pressure 90 mm Hg, or use of antihypertensive medication), previous history of cardiovascular or cerebrovascular disease, hypercholesterolemia (lipid-lowering treatment or total cholesterol level 7.25 mmol/L), cognitive status (MMSE ,24; 24–27; and 28), center (Bordeaux, Dijon, or Montpellier), cranial trauma (with loss of consciousness, yes or no), depressive symptoms (based on the Center for Epidemiologic Studies Depression Scale (CES-D) [25] and defined as a CES-D score 17 for men and 23 for women [26]), anxiety (75th percentile of Spielberger’s state-trait anxiety inventory), and insomnia.

2.4. Statistical analysis 2.4.1. Main analyses As illustrated in Fig. 1, we excluded participants with prevalent dementia (n 5 214) and those with no information about dementia during follow-up (n 5 840). Thus, 8240 participants of 9294 were included in our main analyses. In the main analyses, first, we investigated the risk of dementia associated with prevalent benzodiazepine use at T0; second, we investigated the use of short or long half-life benzodiazepine drugs; and third, we investigated the use of psychotropic drugs. For all analyses, participants who did not use benzodiazepine or psychotropic drugs (n 5 5884) served as the reference group. For the analyses that evaluated the association between benzodiazepine use and dementia, we excluded 617 participants who used psychotropic drugs from the models, leaving 7623 participants for these analyses. For the models that evaluated the association between psychotropic drugs and the risk of dementia, 1246 participants who also used benzo-

3

diazepines were not included, leaving 6994 participants for these analyses. We ran age-, sex-, and center-adjusted models as well as two multivariable-adjusted models. The first included the a priori selected potential confounding factors i.e., body mass index, living status, education, self-perceived health, alcohol consumption, smoking, diabetes, history of hypertension, previous history of cardiovascular or cerebrovascular problems, hypercholesterolemia, cranial trauma, and cognitive status at baseline. As depressive symptoms, anxiety, and insomnia can be prodromal symptoms for dementia, we added them to a second multivariable model that allowed us to evaluate their roles separately. To further evaluate the effect of depressive symptoms, anxiety, and insomnia on the association of benzodiazepine or psychotropic medication use with dementia, we ran stratified analyses using these covariates. We also estimated hazard ratios in nondrinkers (excluding past drinkers) and current drinkers separately. For the main approach, we used a multivariable Cox proportional hazards model to estimate the relative hazard of dementia over the follow-up period. Person-time was calculated from the date of inclusion to the first of the following events: incident dementia, death, loss to follow-up, or the date of the 10-year follow-up visit, whichever came first. We tested the proportionality assumption of the Cox proportional hazard models using an interaction term between the logarithm of the follow-up time and main exposures and found no significant violation. Furthermore, the association between the use of benzodiazepines according to half-life type and dementia was assessed using separate Cox proportional hazard models that controlled for the same set of confounders as described previously. Participants who reported the use of both a shortacting and a long-acting benzodiazepine (n 5 78) were excluded from this analysis. To evaluate whether depression, insomnia, or anxiety modified the association of benzodiazepines or other psychotropics with dementia, we contrasted the main model to a model that included the appropriate interaction terms using the likelihood ratio test. We also performed an analysis of concomitant users of benzodiazepines and psychotropic medications, excluding anyone who only took psychotropic medication or only benzodiazepines (n 5 1860). Missing values were recoded for all dichotomous and categorical variables to constitute their own category and included them in the multivariable models. The variables with the most missing values were hypercholesterolemia (nmiss 5 228, 3.2%), anxiety (nmiss 5 487, 6.8%), and insomnia (nmiss 5 632, 8.9%). All other variables had fewer than 150 missing values (,2%). 2.4.2. Secondary analysis In the secondary analyses, we evaluated the associations of initiators of benzodiazepines or psychotropic medications and the risk of subsequent dementia. For these analyses, we first excluded 1212 subjects who died, developed dementia,

4


Fig. 1. Exclusion and inclusion of participants from the 3C study. Abbreviation: 3C, three-city.

or were lost to follow-up at the time point of the first followup 2 years after baseline, T2. For the analysis of both benzodiazepine and psychotropic initiators and their risk of dementia, all remaining psychotropic or benzodiazepine users at baseline (T0) were excluded (n 5 1924), as were all initiators of both psychotropics and benzodiazepines. This left 5035 participants for these analyses. The first part of the analysis focused on the association of benzodiazepine initiation at T2 with the risk of subsequent dementia. We excluded 232 participants who used psychotropics at T2 to create a reference group composed of individuals who reported no use of benzodiazepines or other psychotic drugs at T2. Next, we ran the same Cox proportional hazard models and controlled for the same set of covariates as in the main analysis. The second part of the analysis assessed the association of psychotropic medication with the risk of dementia. To create the reference group, we excluded the 467 participants who reported the use of benzodiazepine at T2. For the analysis itself, we used the Cox proportional hazards model as described for benzodiazepine users. All data were analyzed using SAS version 9.3 (SAS Institute Inc, Cary, NC, USA). All reported P values are two-tailed, and we considered P , .05 to be statistically significant.

3. Results 3.1. Main analysis The 8240 subjects were followed up for up to 11 years (median follow-up, 8 years). During this time, 830 cases of

dementia were confirmed. Table 1 summarizes the characteristics of benzodiazepine users, psychotropic medication users, and users of both at inclusion and compares them with the 5884 nonbenzodiazepine and nonpsychotropic users. Benzodiazepine and psychotropic users were older, more likely to be female, lived alone more frequently, were more likely to have hypertension or hypercholesterolemia, were less likely to currently drink alcohol or smoke, had more depressive symptoms, insomnia, and anxiety, had lower MMSE scores, and had a poorer perception of their own health compared with nonusers. Those who reported using both benzodiazepines and psychotropics were younger, more likely to be nondrinkers, have depressive symptoms, have lower MMSE scores, and have poorer self-perceived health compared with users of just benzodiazepines or just psychotropics. Table 2 summarizes the association of benzodiazepine or psychotropic medication use with incident dementia. Prevalent users of benzodiazepines had a hazard ratio of 1.23 (95% confidence interval, 1.02–1.48) for increased risk of dementia in the age-, sex-, and center-adjusted model. However, after adjusting for confounding factors, including anxiety, depression, and insomnia, this association showed a multivariable adjusted hazard ratio (HRa) of 1.10 (0.90–1.34). When we made a distinction between users of short (,20 hours) and long half-life benzodiazepines, we found no association between the use of short half-life benzodiazepines and dementia (HRa 5 1.05, 0.85–1.30 in the multivariable model described in Table 2 and taking into account several variables including depression, anxiety, and insomnia). In contrast, we found with the same modeling


5

Table 1 Characteristics of benzodiazepine prevalent users compared with nonusers as well as initiators compared with noninitiators in the Three-City study (n 5 8240)

Characteristics Age Female Body mass index, n (%) ,25 kg/m2 25 kg/m2–,30 kg/m2 30 kg/m2 Education, n (%) No or primary studies Primary studies with diploma or higher primary studies Higher technical certificate Academic qualification certificate or higher Living alone, n (%) Alcohol consumption, n (%) Drinkers Ex-drinkers Nondrinkers Hypertension, n (%) Systolic 160 mm Hg, diastolic 95 mm Hg, or use of antihypertensive medication Depressive symptoms, n (%) CES-D 23 women, CES-D 17 men Diabetes, n (%) Hypercholesterolemia, n (%) Treatment or level 7.25 mmol/L History of cardiovascular or cerebrovascular disease, n (%) Smoking status, n (%) Never Past Current Self-perceived health, n (%) Very good Good Middle Poor Very poor MMSE, n (%) Low score (,24) Medium score (24–28) High score (28) Insomnia, n (%) Cranial trauma, n (%) Anxiety, n (%) 75th percentile of Spielberger inventory trait

Nonusers at inclusion, n 5 5884

Benzodiazepine user, n 5 1246

Psychotropic medication user, n 5 617

Combined benzodiazepine and psychotropic users, n 5 493

73.6 6 5.3 3314 (56.3)

75.2 6 5.7 909 (73.0)

75.1 6 5.6 435 (70.5)

74.7 6 5.3 395 (80.1)

2741 (46.6) 2366 (40.2) 746 (12.7)

609 (48.9) 456 (36.6) 165 (13.2)

279 (45.2) 225 (36.5) 104 (16.9)

247 (50.1) 173 (35.1) 66 (13.4)

421 (7.2) 2228 (37.9) 1358 (23.1) 1865 (31.3) 1893 (32.2)

127 (10.2) 524 (42.1) 296 (23.8) 294 (23.6) 570 (45.8)

47 (7.6) 215 (34.9) 139 (22.5) 94 (15.2) 263 (42.6)

54 (11.0) 216 (43.8) 111 (22.5) 111 (22.5) 223 (45.2)

4816 (81.9) 923 (15.7) 132 (2.2)

966 (77.5) 235 (18.9) 42 (3.4)

472 (76.5) 127 (20.6) 18 (2.9)

338 (68.6) 123 (25.0) 30 (6.1)

3481 (59.2)

835 (67.0)

395 (64.0)

280 (56.8)

532 (9.0) 452 (7.7)

247 (19.8) 96 (7.7)

117 (19.0) 54 (8.8)

181 (36.7) 37 (7.5)

2084 (35.4) 494 (8.4)

502 (40.3) 149 (12.0)

27 (40.0) 59 (9.7)

189 (38.3) 47 (9.5)

3462 (58.8) 2089 (35.5) 331 (5.6)

820 (65.8) 361 (29.0) 65 (5.2)

424 (68.7) 165 (26.7) 27 (4.4)

364 (73.8) 107 (21.7) 22 (4.5)

484 (8.2) 3400 (57.8) 1809 (30.7) 161 (2.7) 7 (0.1)

42 (3.4) 529 (42.5) 558 (44.8) 105 (8.4) 6 (0.5)

16 (2.6) 306 (49.6) 255 (41.3) 33 (5.4) 4 (0.7)

9 (1.8) 160 (32.5) 251 (50.9) 63 (12.8) 5 (1.0)

346 (5.9) 544 (9.3) 4964 (84.4) 798 (13.6) 133 (7.7)

840 (6.7) 158 (12.7) 994 (79.8) 374 (30.0) 97 (7.8)

48 (7.8) 63 (10.2) 497 (80.6) 129 (20.9) 50 (8.1)

46 (9.3) 62 (12.6) 382 (77.5) 138 (28.0) 36 (7.3)

1185 (20.1)

416 (33.4)

210 (21.5)

235 (47.7)

Abbreviations: CES-D, Center for Epidemiologic Studies Depression Scale; MMSE, mini-mental state examination. NOTE. Percentages may not add up to 100% because of rounding or missing values.

an HRa of 1.62 (1.11–2.37) for dementia with the use of long half-life benzodiazepines (Table 2). With regard to psychotropic medication use, we found an increased risk of dementia in an age-, sex-, and center-adjusted model, which decreased after adjustment but which remained statistically significant (HRa 5 1.47, 1.16–1.86). Concomitant use of psychotropic medication and benzodiazepines was associated with the highest effect estimate (HRa 5 1.70, 1.32–2.19). We performed stratified analyses to investigate potential confounding by indication (i.e., anxiety, depressive

symptoms, and insomnia; Table 3). We found no significantly different association patterns for benzodiazepine and psychotropic medication with dementia according to the presence or absence of these three factors (the smallest P value for interactions using the likelihood ratio test was 0.15). In general, the effect estimates were either very similar or even higher among participants who did not have anxiety, depression, or insomnia. For example, the HRa of dementia associated with the use of long halflife benzodiazepines was 2.02 (1.28–3.17) for participants who had no depressive symptoms and 1.08 (0.36–3.22)


6

Table 2 Association between benzodiazepine and psychotropic medication use at baseline T0 and dementia in the 3C study adjusted for different confounders Benzodiazepine and psychotropic drugs use Nonusers (n 5 5884) Cases 5 496 Benzodiazepines (n 5 1246) Cases 5 151 Long half-life (n 5 209) Cases 5 30 Short half-life (n 5 987) Cases 5 119 Psychotropic medication (n 5 617) Cases 5 91 Benzodiazepine and psychotropic users (n 5 493) Cases 5 92

Age, sex, and center adjusted

Multivariable I*

Multivariable IIy

1.00 (Reference)

1.00 (Reference)

1.00 (Reference)

1.23 (1.02–1.48)

1.14 (0.94–1.38)

1.10 (0.90–1.34)

1.83 (1.26–2.65)

1.68 (1.16–2.46)

1.62 (1.11–2.37)

1.17 (0.95–1.44)

1.09 (0.88–1.35)

1.05 (0.85–1.30)

1.62 (1.29–2.03)

1.55 (1.23–1.95)

1.47 (1.16–1.86)

2.18 (1.74–2.74)

1.96 (1.55–2.49)

1.70 (1.32–2.19)

Abbreviation: 3C, three-city. *Adjusting for age, sex, center, factors body mass index, living status, education, self-perceived health, alcohol consumption, smoking, diabetes, history of hypertension, previous history of cardiovascular or cerebrovascular problems, hypercholesterolemia, cranial trauma, and cognitive status at baseline. y Adjusting for all variables in model I plus depression, anxiety, and insomnia.

for those with depressive symptoms. With regard to alcohol consumption, the HRa of dementia associated with the use of benzodiazepines was 1.22 (0.75–2.01) in nondrinkers and 1.15 (0.92–1.44) in current drinkers; for other classes of psychotropics, HRsa were 2.17 (1.34–3.52) and 1.33 (0.99–1.79), respectively.

tropic medication users with the risk of subsequent dementia was very similar to the pattern for prevalent users (Table 4). The multivariable HRsa for dementia were 1.12 (0.80–1.58) for incident benzodiazepine users and 1.90 (1.29–2.81) for incident psychotropic users. There were too few outcome events in the subgroups to allow further stratification according to the use of short or long half-life benzodiazepines.

3.2. Secondary analysis At the first follow-up visit (T2), a total of 5035 subjects who did not report any benzodiazepine or other psychotropic medication use at baseline (T0) and who were not diagnosed with dementia at T2 were eligible for the incident user analysis. These participants were followed up for a maximum of 9 years (median of 7 years). The pattern of association of incident benzodiazepine or incident psycho-

4. Discussion The results of this large population-based, prospective cohort study of individuals aged 65 years showed no overall association between prevalent benzodiazepine use and the risk of developing dementia during a median follow-up of 8 years. When distinguishing between short and long

Table 3 Association between benzodiazepine use at baseline T0 and hazard of dementia in the 3C study stratified on anxiety, depression, insomnia, and psychotropic medication use Stratification variable Benzodiazepine and psychotropic drugs use Nonusers (n 5 5884) Cases 5 496 Benzodiazepines (n 5 1246) Cases 5 151 Long half-life (n 5 209) Cases 5 30 Short half-life (n 5 987) Cases 5 119 Psychotropic medication (n 5 617) Cases 5 91

No anxiety

Anxiety

No depressive symptoms

Depressive symptoms

No insomnia

Insomnia

1.00 (Reference)

1.00 (Reference)

1.00 (Reference)

1.00 (Reference)

1.00 (Reference)

1.00 (Reference)

1.20 (0.93–1.56)

0.89 (0.59–1.34)

1.26 (1.01–1.57)

0.67 (0.41–1.10)

1.17 (0.90–1.52)

1.16 (0.77–1.75)

2.10 (1.28–3.45)

1.15 (0.51–2.60)

2.02 (1.28–3.17)

1.08 (0.36–3.22)

1.64 (1.01–2.69)

2.03 (0.91–4.56)

1.15 (0.86–1.53)

0.86 (0.56–1.34)

1.18 (0.93–1.50)

0.69 (0.40–1.20)

1.09 (0.81–1.46)

1.11 (0.72–1.73)

1.37 (0.98–1.92)

1.44 (0.93–2.23)

1.60 (1.22–2.09)

1.08 (0.62–1.87)

1.59 (1.17–2.16)

1.02 (0.59–1.77)

Abbreviation: 3C, three-city. NOTE. All models adjusted for age, sex, center, factors body mass index, living status, education, self-perceived health, alcohol consumption, smoking, diabetes, history of hypertension, previous history of cardiovascular or cerebrovascular problems, hypercholesterolemia, cranial trauma, and cognitive status at baseline.


7

Table 4 Association between benzodiazepine and psychotropic medication initiation at T2 and dementia in the 3C study (n 5 5035) Benzodiazepine and psychotropic drugs use Nonusers (n 5 4336) Cases 5 307 Benzodiazepine initiators (n 5 467) Cases 5 42 Psychotropic medication (n 5 232) Cases 5 30

Age, sex, and center adjusted

Multivariable I*

Multivariable IIy

1.00 (Reference)

1.00 (Reference)

1.00 (Reference)

1.25 (0.90–1.74)

1.15 (0.82–1.61)

1.12 (0.80–1.58)

2.08 (1.43–3.03)

1.92 (1.30–2.83)

1.90 (1.29–2.81)

Abbreviation: 3C, three-city. *Adjusting for age, sex, center, factors body mass index, living status, education, self-perceived health, alcohol consumption, smoking, diabetes, history of hypertension, previous history of cardiovascular or cerebrovascular problems, hypercholesterolemia, cranial trauma, and cognitive status at baseline. y Adjusting for all variables in model 1 plus depression, anxiety, and insomnia.

half-life benzodiazepines, we found no association between short half-life benzodiazepines and the risk of dementia. In contrast, users of long half-life benzodiazepines had about a 60% increased risk of developing dementia, an association that was not explained by major confounding factors. In particular, it was not explained by depression, anxiety, or insomnia. Four previous studies showed an increased risk of dementia in benzodiazepine users. Two case-control studies were conducted in Taiwan using the health insurance data of people aged 45years; both found an increased risk of dementia in prevalent and chronic users over a maximum follow-up of 8 years [10,11]. Another nested case-control study in France showed an increased risk only in former benzodiazepine users in elderly people aged 65 years over a maximum follow-up of 8 years [16]. The PAQUID study provides the most recent population-based data. This cohort of elderly French participants (70 years) in Bordeaux had follow-up of up to 15 years [13] and reported an HRa of 1.60 (1.08–2.38) for dementia in initiators of benzodiazepine. The results of the Caerphilly cohort study also showed an increased risk of dementia for benzodiazepine users during a 22year follow-up period (adjusted odds ratio [ORa], 2.94; 1.16–7.46) [12]. A recently published nested casecontrol study of about 2000 older members of a public drug plan in the province of Quebec, Canada, reported an ORa of 1.51 (1.36–1.69) for the risk of Alzheimer’s disease associated with benzodiazepine use overall and 1.70 (1.46–1.98) for long acting ones [17]. Other previous studies found either no association between benzodiazepine use and dementia [15] or even a protective association [14], but they were based on small samples and had a retrospective design, making direct comparisons to our data difficult. It is well known that benzodiazepine drugs induce both amnestic and nonamnestic cognitive impairments, with evidence of a dose-response relationship [27], and memory impairment has been shown to be one of the stron-

gest predictors for dementia [28]. However, there is currently no demonstrated mechanism linking benzodiazepine use to dementia. Furthermore, our finding of an about 50% increase in the risk of dementia for prevalent users of nonbenzodiazepine psychotropic medications, if confirmed, raises the question of a mechanism that would not be specific to benzodiazepines. The risk of dementia was higher for those who concomitantly took psychotropic medications and benzodiazepine (an increased risk of about 70%). Owing to the implications of this association and the possible additive effects of using benzodiazepines and other psychotropic medications, this merits further examination in future studies. Psychotropic medications, and particularly antidepressant drugs, are sometimes given in the early phase of dementia to distinguish between depression and dementia, so reverse causation might have a strong impact. However, we found that the association between psychotropic medication and dementia was stronger in participants without depressive symptoms, and we did not observe any significant modifying effect. These stratification results, therefore, reduce (but do not exclude) the likelihood that the increased risk of dementia in benzodiazepine and other psychotropic users is due to its indication being a prodromal symptom of dementia. Anxiety is another possible confounder as it has been associated with the risk of dementia, and anxiety symptoms may be masked because they have been effectively treated by benzodiazepines. In the present study, we found no significant association between benzodiazepine use and dementia among anxious individuals (stratified analysis presented in Table 3). In the same stratified analysis, the risk of dementia was increased in benzodiazepine users compared with that in nonusers in the nonanxious group. These results do not negate the hypothesis that the association between benzodiazepines and dementia might be partly confounded by the effect of benzodiazepines on anxiety symptoms. However, in the nonanxious group,

8


the proportion of false negative individuals (i.e., classified as nonanxious because their symptoms are effectively treated by benzodiazepines) is likely to be very small compared with the true negative ones (i.e., correctly classified as nonanxious). Therefore, although this masking/ confounding effect could have some impact, it is unlikely that it would create by itself the relationship observed between benzodiazepine use and dementia. Furthermore, this effect would be even less plausible for the association observed between nonbenzodiazepine psychotropic drugs and dementia. A cohort of incident users is a methodologically sound way to address issues of confounder temporality and depletion of susceptibles. In the secondary analysis, which used this design, the results did not change remarkably for users of psychotropics medication and benzodiazepines. Furthermore, as some users may have started treatment between study inclusion and the first follow-up 2 years later, depletion of susceptibles would still be possible in an incident user design, as is the case with most cohort studies that do not account for medication duration. However, the incident user design avoids confounder temporality issues. Our study has several strengths: a population-based prospective design; a large number of participants and outcome events; a large amount of information about potential confounding factors; standardized, neurologistvalidated assessment of dementia; and information about many covariates that could potentially interfere with the investigated associations. Several limitations should be taken into consideration when evaluating our results. First, although our study had a mean follow-up time of 8 years, this may not have been long enough to pick up an association between overall benzodiazepine use and the risk of dementia. In two other cohort studies, such an association was apparent after longer follow-up times [12,13]. Second, dementia develops over years, and its early symptoms may lead to prescriptions for benzodiazepines or psychotropic drugs. However, in our stratified analyses, we found no evidence that depression, anxiety, or insomnia were factors that strongly influenced the association between the use of benzodiazepines or other psychotropic drug use on the risk of dementia. Third, despite our efforts to adequately control for confounding by many factors, residual and nonmeasurable confounding remains possible, as our data are observational. Nevertheless, to address this possible association at the population level, an observational study is the only feasible study design. Fourth, although our assessment of medication use and outcome events was standardized, misclassification is possible. However, owing to the prospective design, such misclassification would likely result in a random bias (bias toward a null

association). Fifth, although we had a large number of outcome events, a limited number of participants developed dementia in subgroups, particularly among users of long half-life benzodiazepines (n 5 30). Finally, despite the population-based setting of our study, participants in cohort studies of the elderly differ from the general elderly population [29], so that generalizability to other settings or age groups may be limited. These findings provide important information for general practitioners and health authorities as they provide evidence that the association between benzodiazepines and dementia is limited to long half-life benzodiazepines and is not observed in users of short half-life benzodiazepines. Our data further extend the increased risk of dementia to users of other psychotropic medications. Therefore, physicians should carefully weigh the expected benefits of the use of benzodiazepines or other psychotropic medications with the potential adverse effects, avoid long half-life molecules whenever possible, and limit the prescription time to a few weeks, as recommended by current guidelines. The French National Agency for Drug Safety, ANSM, recommends the use of only short half-life benzodiazepines (,20-hour half-life) in the elderly because of the risk of accumulation of the drugs or their metabolites and also recommends their use for a limited period of time. Our results strongly support adherence to these guidelines. The potential impact of the association between psychotropic drugs and the risk of dementia suggests that additional studies are needed to determine whether these drugs are markers of an underlying process that leads to dementia or part of a causal pathway and evaluate potential additive effects with benzodiazepines. Acknowledgments The Three-City Study is conducted under a partnership agreement between the Institut National de la Sante et de la Recherche Medicale, the Victor Segalen–Bordeaux II University, and Sanofi-Aventis. The Fondation pour la Recherche Medicale funded the preparation and initiation of the study. The Fondation Plan Alzheimer partly funded the follow-up of the study. The 3C study is also supported by the Caisse Nationale Maladie des Travailleurs Salaries, Direction Generale de la Sante, MGEN, Institut de la Longevite, Conseils Regionaux of Aquitaine and Bourgogne, Fondation de France, and Ministry of Research–INSERM Programme “Cohortes et collections de donnees biologiques.” The funding organizations played no role in the design and conduct of the study, in the collection, management, analysis, and interpretation of the data or in preparation, review, or approval of the article.


RESEARCH IN CONTEXT

1. Systematic review: In January 2014, we performed a MEDLINE search with the words “benzodiazepine, dementia or Alzheimer’s disease, and epidemiological studies.” We found 53 articles, and among them, we selected the original articles studying the role of benzodiazepine on the risk of dementia. 2. Interpretation: Results of this large, prospective cohort show that the association between benzodiazepine and dementia differs according to medication half-life. An association between benzodiazepine and risk dementia was limited to long half-life benzodiazepine and not observable among short half-life benzodiazepine users. Our data further extend the increased risk of dementia to users of other psychotropic medications. 3. Future directions: These findings provide important information for prescribers and health authorities. When prescribing benzodiazepines, physicians should avoid long half-life molecules. With regard to other psychotropics, a re-analysis of existing cohorts would help confirming or not this finding which could have a major impact on drug prescription in the elderly.

References [1] Lagnaoui R, Depont F, Fourrier A, Abouelfath A, Begaud B, Verdoux H, et al. Patterns and correlates of benzodiazepine use in the French general population. Eur J Clin Pharmacol 2004; 60:523–9. [2] Fourrier A, Letenneur L, Dartigues JF, Moore N, Begaud B. Benzodiazepine use in an elderly community-dwelling population. Characteristics of users and factors associated with subsequent use. Eur J Clin Pharmacol 2001;57:419–25. [3] Hogan DB, Maxwell CJ, Fung TS, Ebly EM, Canadian Study of Health and Aging. Prevalence and potential consequences of benzodiazepine use in senior citizens: results from the Canadian Study of Health and Aging. Can J Clin Pharmacol 2003;10:72–7. [4] Carrasco-Garrido P, Jimenez-Garcia R, Astasio-Arbiza P, OrtegaMolina P, de Miguel AG. Psychotropics use in the Spanish elderly: Predictors and evolution between years 1993 and 2003. Pharmacoepidemiol Drug Saf 2007;16:449–57. [5] Windle A, Elliot E, Duszynski K, Moore V. Benzodiazepine prescribing in elderly Australian general practice patients. Aust N Z J Public Health 2007;31:379–81. [6] European Medicines Agency. Summary of product characteristics for benzodiazepines as anxiolytics or hypnotics. Available from: www. ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/ 2009/09/WC500003774.pdf 1994 [cited 2012]. [7] Gorevski E, Bian B, Kelton CM, Martin Boone JE, Guo JJ. Utilization, spending, and price trends for benzodiazepines in the US Medicaid

9

program: 1991-2009. Ann Pharmacother 2012;46:503–12. [8] Li M, Szabo A, Rosenberg HC. Down-regulation of benzodiazepine binding to alpha 5 subunit-containing gamma-aminobutyric acid(A) receptors in tolerant rat brain indicates particular involvement of the hippocampal CA1 region. J Pharmacol Exp Ther 2000; 295:689–96. [9] Greenshaw AJ. Neurotransmitter interactions in psychotropic drug action: Beyond dopamine and serotonin. J Psychiatry Neurosci 2003; 28:247–50. [10] Wu CS, Wang SC, Chang IS, Lin KM. The association between dementia and long-term use of benzodiazepine in the elderly: Nested case-control study using claims data. Am J Geriatr Psychiatry 2009; 17:614–20. [11] Wu CS, Ting TT, Wang SC, Chang IS, Lin KM. Effect of benzodiazepine discontinuation on dementia risk. Am J Geriatr Psychiatry 2011; 19:151–9. [12] Gallacher J, Elwood P, Pickering J, Bayer A, Fish M, Ben-Shlomo Y. Benzodiazepine use and risk of dementia: Evidence from the Caerphilly Prospective Study (CaPS). J Epidemiol Community Health 2012;66:869–73. [13] Billioti de Gage S, Begaud B, Bazin F, Verdoux H, Dartigues JF, Peres K, et al. Benzodiazepine use and risk of dementia: Prospective population based study. BMJ 2012;345:e6231. [14] Fastbom J, Forsell Y, Winblad B. Benzodiazepines may have protective effects against Alzheimer disease. Alzheimer Dis Assoc Disord 1998;12:14–7. [15] Lagnaoui R, Tournier M, Moride Y, Wolfson C, Ducruet T, Begaud B, et al. The risk of cognitive impairment in older community-dwelling women after benzodiazepine use. Age Ageing 2009;38:226–8. [16] Lagnaoui R, Begaud B, Moore N, Chaslerie A, Fourrier A, Letenneur L, et al. Benzodiazepine use and risk of dementia: A nested case-control study. J Clin Epidemiol 2002;55:314–8. [17] Billioti de Gage S, Moride Y, Ducruet T, Kurth T, Verdoux H, Tournier M, et al. Benzodiazepine use and risk of Alzheimer’s disease: Case-control study. BMJ 2014;349:g5205. [18] Alperovitch A. Vascular factors and risk of dementia: Design of the three-city study and baseline characteristics of the study population. Neuroepidemiology 2003;22:316–25. [19] Lechevallier-Michel N, Gautier-Bertrand M, Alperovitch A, Berr C, Belmin J, Legrain S, et al. Frequency and risk factors of potentially inappropriate medication use in a community-dwelling elderly population: Results from the 3C study. Eur J Clin Pharmacol 2005; 60:813–9. [20] Alperovitch A, Blachier M, Soumare A, Ritchie K, Dartigues JF, Richard-Harston S, et al. Blood pressure variability and risk of dementia in an elderly cohort, the Three-City Study. Alzheimers Dement 2014; 10:S330–7. [21] Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12:189–98. [22] Isaacs B, Kennie AT. The Set test as an aid to the detection of dementia in old people. Br J Psychiatry 1973;123:467–70. [23] Association AP. Diagnostic and statistical manual of mental disorders: DSM-IV. 4th ed. Washington, DC: American Psychiatric Association; 1994. [24] McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA work group under the auspices of department of health and human services task force on Alzheimer’s disease. Neurology 1984;34:939–44. [25] Radloff L. The CES-D scale: A self-report depression scale for research in the general population. Appl Psych Meas 1977; 1:385–401. [26] Fuhrer R, Rouillon F, La version Franc¸aise de l’echelle CES-D (Center for Epidemiologic Studies-Depression Scale). Description et

10


traduction de l’echelle d’auto-evaluation. Pyschiatr Psychobiol 1989; 4:181–90. [27] Tannenbaum C, Paquette A, Hilmer S, Holroyd-Leduc J, Carnahan R. A systematic review of amnestic and non-amnestic mild cognitive impairment induced by anticholinergic, antihistamine, GABAergic and opioid drugs. Drugs Aging 2012; 29:639–58.

[28] Grober E, Lipton RB, Hall C, Crystal H. Memory impairment on free and cued selective reminding predicts dementia. Neurology 2000; 54:827–32. [29] Dufouil C, Ducimetiere P, Alperovitch A. Sex differences in the association between alcohol consumption and cognitive performance. EVA Study Group. Epidemiology of Vascular Aging. Am J Epidemiol 1997; 146:405–12.