Uncertain Association Between Benzodiazepine Use and the Risk of Dementia: A Cohort Study

JAMDA xxx (2019) 1e11

JAMDA journal homepage: www.jamda.com

Original Study

Uncertain Association Between Benzodiazepine Use and the Risk of Dementia: A Cohort Study Yeon-Hee Baek BA a, Hyesung Lee MSc a, Woo Jung Kim MD, PhD b, Jee-Eun Chung PhD c, Nicole Pratt PhD d, Lisa Kalisch Ellett PhD d, Ju-Young Shin PhD a, * a

School of Pharmacy, Sungkyunkwan University, Suwon, Gyeonggi-do, South Korea Department of Psychiatry, Yongin Severance Hospital, Yonsei University College of Medicine, Yongin, Gyeonggi-do, South Korea c College of Pharmacy and Institute of Pharmaceutical Science and Technology, Hanyang University, Ansan, Gyeonggi-do, South Korea d Quality Use of Medicines and Pharmacy Research Centre, School of Pharmacy and Medical Sciences, University of South Australia, Adelaide, Australia b

a b s t r a c t Keywords: Dementia benzodiazepine cohort studies propensity score

Objective: To examine the association between benzodiazepine use and the risk of dementia. Design, setting, and participants: We conducted a retrospective cohort study, using a nationwide healthcare database of South Korea (2002e2016). The participants included new users of benzodiazepines aged
50 years, with no prior prescription record of benzodiazepines or a history of dementia within the previous 5 years (2002e2006). Methods: Outcome was defined as an incident dementia with specified algorithms using diagnosis and prescription records, with the application of a 5-year lag-time following the index date during which outcomes were censored. We used a multivariable Cox proportional hazard model to estimate hazard ratio (HR) and the 95% confidence interval (CI). Comorbidities and comedications were treated as timevarying covariates in 90-day windows, and an active comparator was used to reduce potential bias from confounding by indication. Active comparators were defined as new-users of antidepressants. Results: Our final participants included 616,256 patients, after propensity score estimation and matching on a 1:1 ratio. We observed a 23% increase in the risk of dementia in benzodiazepine users, compared with that in nonusers, over a mean follow-up period of 5.5 years (HR 1.23, 95% CI 1.14e1.32). A consistent finding was observed when the lag-time duration was extended to 7 years, revealing a close to null association (HR 1.17, 95% CI 1.04e1.30). When new-users of antidepressants were used as the active comparator, no increase in the risk of dementia with benzodiazepines was observed over 7 years (HR 1.01, 95% CI 0.81e1.27). Conclusions and implications: A significant association was observed between benzodiazepine use and the risk of dementia, compared with nonusers. However, a null or negative association was observed with the use of the active comparator, suggesting the absence of a causal association between dementia and benzodiazepine use. Ó 2019 AMDA e The Society for Post-Acute and Long-Term Care Medicine.

This research was supported by a grant of the Korean Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health and Welfare, Republic of Korea (Grant number: NHCRHC17C0020). This study was conducted independently of the funder. This study was approved by the institutional review board of Sungkyunkwan University (No. 2017-08-008), which waived the informed consent, as only deidentified data were used in this study. The findings of this study were presented at the 2018 Fall International Convention of The Pharmaceutical Society of Korea in Jeju Island, South Korea, on October 19, 2018. The authors declare no conflicts of interest. Y.-H.B. and H.L. contributed equally to this work as co-first authors. * Address correspondence to Ju-Young Shin, PhD, School of Pharmacy, Sungkyunkwan University, 2066, Seobu-ro, Jangan-gu, Suwon, Gyeonggi-do, South Korea. E-mail address: [email protected] (J.-Y. Shin). https://doi.org/10.1016/j.jamda.2019.08.017 1525-8610/Ó 2019 AMDA e The Society for Post-Acute and Long-Term Care Medicine.

Benzodiazepines are widely prescribed drugs for the treatment of psychiatric and neurologic disorders, which are highly prevalent (>13%) in older patients.1,2 However, concerns have been raised regarding the potential risk of cognitive decline and dementia associated with benzodiazepine use. It has been hypothesized that chronic benzodiazepine use may lead to a decrease in cognitive reserve,3 downregulation of gamma-aminobutyric acid receptors,4 or degradation in the capability to utilize alternative neural networks because of lowered brain activation levels.5 The Asian population may be at a higher risk of benzodiazepine-induced dementia because of genetic variations in cytochrome P450 2C19 (CYP2C19); diazepam, for example, is metabolized by CYP2C19 via a demethylation process.6 Higher prevalence of poor metabolism because of the genetic

2

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

variation of CYP2C197 and slower elimination of diazepam in Asians8 than in Caucasians may result in drug overexposure, leading to a potential increased risk of dementia in this population. However, the association between benzodiazepine use and dementia is still controversial because of contradictory study findings. A recent systematic review, including 12 observational studies, reported a 1.38-fold increase in the risk of dementia in benzodiazepine users compared with nonusers.9 However, high heterogeneity was observed among these studies, with 5 studies reporting a positive association,10e14 3 reporting no association,15e17 and 1 reporting a nonsignificant dose-response relationship.18 The inconsistencies among the findings of these studies may be due to methodological limitations such as potential exposure misclassification due to 2- to 3year follow-up visit in prospective studies,13,19 no consideration of disease latency,11,19 or confounding by indication resulting from comparison with nonusers only11,13,18,19 as benzodiazepines are often used in the management of the prodromal symptoms of dementia. Given the inconsistent results on the association between benzodiazepine use and dementia and the possible genetic variations associated with the risk, further studies are required to evaluate the association between this drug class and dementia. The objective of this nationwide population-based study was to assess the association between benzodiazepine use and the risk of dementia in South Korea.

December 31, 2016 (Supplementary Table 1). Nonusers were those who had no prescription records of either benzodiazepines or antidepressants during the follow-up period. Cohort entry date was set as January 1, 2007 for all eligible patients. We excluded the following patients: (1) those with a history of benzodiazepine or antidepressant use or dementia within the previous 5 years (2002e2006), to include new users and identify incident dementia only; (2) those who died before cohort entry date; (3) those who were diagnosed with dementia between cohort entry date and exposure initiation; and (4) those who had the same incident prescription date for both benzodiazepines and antidepressants, to avoid inappropriate exposure group assignment. Follow-up started from the index date, defined as the first prescription of benzodiazepine or antidepressant or, for nonusers, a randomly selected medical visit date after cohort entry date. Each benzodiazepine user was matched to both a nonuser and an antidepressant user (1:1) by age (in 1-year band), sex, and period between cohort entry date and index date. By matching on period, time-lag bias, a bias that occurs when the person-time of a comparator is exaggerated, was eliminated (Figure 2).26 Patients matched in the agesex exact matched cohort were further matched using propensity scores (PS).

Methods

Exposure was defined using an intention-to-treat approach to avoid bias from informative censoring, as treatment discontinuation or switching can be associated with patients’ prognosis. The risk of dementia could either be overestimated, if patients who discontinued benzodiazepines were at a lower risk than persistent benzodiazepine users, or underestimated, if patients who discontinued benzodiazepines were at a higher risk than persistent benzodiazepine users.27 The duration-response relationship was assessed in terms of the cumulative exposure duration, using the proportion of days covered (PDC), calculated by summing the durations of all the prescriptions divided by the person-years of follow-up for each participant. For the comparison with antidepressant users, we estimated hazard ratios (HRs) in the corresponding PDC strata: short-term (PDC < 0.15), medium-term (0.15  PDC <0.3), and long-term (PDC
0.3). The risk difference was examined according to the duration of action of benzodiazepines; short-acting users: <24 hours half-life; long acting users:
24 hours half-life; concurrent users: prescribed both shortacting and long-acting benzodiazepines on the index date (Supplementary Table 2).

Data Sources We retrieved health insurance claims data from the National Health Insurance Service (NHIS) database of South Korea, which covers the entire 50 million population of South Korea, from January 1, 2002 to December 31, 2016. In this population, there were 13 million people aged over 50 years in 2007, and we obtained the data of 1.6 million patients who met the eligibility criteria from 30% randomly sampled data of 4 million patients. The NHIS is the universal singlepayer national healthcare system of South Korea,20 and the database contains the records of all demographic and clinical information on insurance-covered inpatient and outpatient visits, procedures, prescriptions, and diagnosis records (coded using International Classification of Diseases, 10th Revision (ICD-10)).21 In addition, this database contains death records, obtained from data linked to the national statistics of South Korea. The overall positive predictive value of the diagnoses in the South Korean healthcare database is 82%.22 This study was approved by the institutional review board, which waived the informed consent, as only de-identified data were used in this study. Study Design and Population We conducted a nationwide retrospective cohort study using a study population that comprised benzodiazepine users, nonusers, and antidepressant users (Figure 1). Antidepressants [selective serotonin reuptake inhibitors (SSRIs) and serotonin-norepinephrine reuptake inhibitors (SNRIs)] were selected as the active comparator because they are also psychotropic drugs, and have overlapping indications with benzodiazepines, such as depression and anxiety. In Korean clinical practice, 40.7% of benzodiazepine prescriptions are for anxiety, followed by depression (accounts for 27.7% of benzodiazepine prescriptions).23 Moreover, clinical guidelines recommend antidepressants as first-line treatment for both depression and anxiety.24 SSRI and SNRI antidepressants were selected because these drugs have less anticholinergic effects, which are known to be associated with increased dementia risk.25 Benzodiazepine or antidepressant users included in the study were those who received
1 prescription of benzodiazepine or antidepressant, between January 1, 2007 and

Exposure Definition

Outcome Definition The primary outcome was defined as an incident dementia event with prespecified algorithms using diagnosis and prescription records to reduce outcome misclassification (Supplementary Table 1). A dementia event was defined as the presence of
1 inpatient or 2 outpatient dementia diagnoses from the neurology or psychiatry department, using up to the fifth diagnosis record, and a prescription of dementia medication. All patients were followed up from their index date until death because of any cause or the end of the study period (December 31, 2016), whichever occurred first. Specifically, we performed these analyses in which patients who developed dementia within the predefined lag-times from their index dates were censored. Predefined lag-times were (1) 7 years; (2) 5 years; (3) 3 years; and (4) 1 year. Lag-times were applied to avoid protopathic bias because of pre-existing undiagnosed dementia and to address disease latency for drug-induced dementia. In secondary analyses, we redefined the outcome definition, given the uncertainties regarding the validity of the outcome of dementia, with the application of a 5-year lag time. We applied

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

Fig. 1. Selection of study participants and construction of exact matched cohort and propensity score matched cohort.

3

4

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

Benzodiazepine users vs. non-users Index date : The first date of benzodiazepine use

Patient A : Benzodiazepine user Duration matching Patient B : Non-user Cohort entry date 2007.01.01.

Index date : Randomly selected medical visit on the same date

Benzodiazepine users vs. antidepressant users Patient C : Benzodiazepine user Duration matching Patient D : Antidepressant user

Cohort entry date 2007.01.01.

Index date : The first date of exposure

: Benzodiazepine prescription records : Antidepressant prescription records : Medical visit

Fig. 2. Selection of duration-matched index date in benzodiazepine users vs nonusers cohort, and benzodiazepine users vs antidepressant users cohort to avoid time-lag bias.

stricter definitions for the outcome, as follows: (1) presence of
1 inpatient or 2 outpatient dementia diagnoses, restricted to primary and secondary diagnoses, and a prescription of dementia medication; (2) presence of dementia diagnosis from a tertiary hospital and a prescription of dementia medication. We applied other instrumental definitions: (3) presence of
1 record of a dementia diagnosis in either an inpatient or outpatient setting and a prescription of dementia medication; (4) presence of either
1 inpatient or 2 outpatient dementia diagnoses or a prescription of dementia medication; and (5) presence of either a dementia diagnosis or a prescription of dementia medication.

Statistical Analyses Patient characteristics were measured in the year preceding the index date and included demographic information (age, sex, and insurance type), medical institution type, polypharmacy (defined as
5 medications in a single prescription), the number of inpatient visits, comorbidities, comedications, Charlson comorbidity index (CCI), and index year. PS was estimated for 3 mutually exclusive groups: (1) benzodiazepine users; (2) nonusers; and (3) antidepressant users, by fitting a multivariable logistic regression model using all predefined covariates (listed in Table 1). The model discrimination was assessed with C-statistics, and it was considered successful if the estimate was between 0.6 and 0.8. Benzodiazepine users were matched to both nonusers and antidepressant users (1:1) with PS, using the Greedy 8/1 digit matching macro within the age-sex exact matched cohort.28 For comparison of baseline characteristics, we calculated absolute standardized difference; an imbalance between groups was defined as absolute standardized difference value of >0.1. The incidence of dementia per 1000 person-years was calculated by dividing the number of dementia events by the sum of person-years at risk, and then multiplying the estimates by 1000. We used a multivariable Cox proportional hazard model to estimate HR and 95% confidence intervals (CI). The models were adjusted for time dependent confounders including cerebrovascular disease and depression, which have been previously reported as risk factors for dementia.29 Comorbidities and comedications were treated as time-varying covariates in 90-day increments. In the benzodiazepine users vs nonusers cohort, the age-sex exact-matched cohort was

adjusted for CCI, the number of inpatient visits, depression, hypertension, diabetes mellitus, hyperlipidemia, cerebrovascular disease, stroke, antipsychotics, antiepileptics, tricyclic antidepressants, platelet aggregation inhibitors, anticoagulants, statins, antihypertensives, and antidiabetes drugs; the PS-matched cohort was adjusted for CCI, polypharmacy, depression, stroke, antipsychotics, narcotic analgesics, tricyclic antidepressants, and anticoagulants. In benzodiazepine users vs antidepressant users cohort, the age-sex exact-matched cohort was adjusted for CCI, polypharmacy, the number of inpatient visits, and all the comorbidities and comedications and the PSmatched cohort was adjusted for CCI, depression, and cerebrovascular disease. We conducted stratified analyses with the application of a 5-year lag-time to confirm whether there was additional confounding, using a single model with interaction terms according to age group (in 5year band), sex, subtype of dementia, insurance type, and medical institution type. All statistical analyses were performed using SAS Enterprise Guide 7.1 for Windows (SAS Institute Inc, Cary, NC). A 2tailed value of P < .05 indicated statistical significance. Results Among 1,576,452 eligible patients, we identified 1,125,346 benzodiazepine users, 433,429 nonusers, and 17,677 antidepressant users (Figure 1). The benzodiazepine users vs nonuser cohort included 775,606 patients after exact matching and 616,256 patients after PS matching (C-statistic ¼ 0.69). The benzodiazepine users vs antidepressant users cohort included 25,508 patients after exact matching and 18,272 patients after PS matching (C-statistic ¼ 0.65). Table 1 shows the baseline characteristics of benzodiazepine users, nonusers, and antidepressant users in age-sex exact-matched cohorts and PS-matched cohorts. After PS matching, all the variables were wellbalanced between the 2 groups, other than medical institution type, in both cohorts. Benzodiazepine Users vs Nonusers Table 2 shows the HRs for dementia associated with benzodiazepine use compared with nonuse. The risk of dementia decreased with the increased duration of lag-time in the PS-matched cohort over a

Table 1 Baseline Characteristics of Benzodiazepines Users, Nonusers, and Antidepressant Users in Age-Sex Exact-Matched and PS-Matched Cohort, 2007e2016 Benzodiazepine Users vs Nonusers

Benzodiazepine Users vs Antidepressant Users PS Matched Cohort (C-statistics ¼ 0.69)

Age-Sex Exact-Matched Cohort Nonusers (n ¼ 387,803)

aSD

67,808 106,099 79,778 53,274 33,594 20,461 13,358 13,431

67,808 106,099 79,778 53,274 33,594 20,461 13,358 13,431

0.00

(17.5) (27.4) (20.6) (13.7) (8.7) (5.3) (3.4) (3.5)

(17.5) (27.4) (20.6) (13.7) (8.7) (5.3) (3.4) (3.5)

Benzodiazepine Users (n ¼ 308,128) 54,449 85,133 63,693 41,762 26,223 15,868 10,444 10,556

(17.7) (27.6) (20.7) (13.6) (8.5) (5.1) (3.4) (3.4)

Nonusers (n ¼ 308,128)

53,242 84,759 63,802 42,226 26,495 16,033 10,545 11,026

(17.3) (27.5) (20.7) (13.7) (8.6) (5.2) (3.4) (3.6)

0.00 137,209 (35.4) 250,594 (64.6)

137,209 (35.4) 250,594 (64.6)

365,383 (94.2) 22,420 (5.8)

370,576 (95.6) 17,227 (4.4)

36,854 61,697 44,054 2204 241,059 1935

21,401 35,837 28,069 3438 270,521 25,837

108,442 (35.2) 199,686 (64.8)

293,434 (95.2) 14,694 (4.8)

293,083 (95.1) 15,045 (4.9)

26,222 45,864 34,051 1779 198,784 1428

19,572 34,625 24,478 2,513 209,631 17,309

235,651 (60.8) 121,081 (31.2) 31,071 (8.0)

(6.4) (11.2) (7.9) (0.8) (68.0) (5.6)

180,559 (46.6) 207,244 (53.4)

4276 (1.1) 3344 (0.9) 88,598 70,203 79,841 25,928

(22.8) (18.1) (20.6) (6.7)

1987 (0.5) 2641 (0.7) 80,278 58,741 61,581 17,847

(20.7) (15.1) (15.9) (4.6)

11,661 (91.4) 1093 (8.6)

1921 (0.6) 2310 (0.7)

0.05 0.08 0.12 0.09

65,581 49,782 54,200 16,303

(21.3) (16.2) (17.6) (5.3)

1950 (0.6) 2370 (0.8) 65,563 50,727 54,959 16,466

(21.3) (16.3) (17.8) (5.3)

0.00 0.00 0.01 0.00

aSD

0.00

1037 2042 2000 1543 1160 810 371 173

1042 2012 2000 1564 1157 808 385 168

0.01

(11.4) (22.4) (21.9) (16.9) (12.7) (8.9) (4.1) (1.9)

(11.4) (22.0) (21.9) (17.1) (12.7) (8.8) (4.2) (1.8)

0.00 4063 (44.5) 5073 (55.5)

4102 (44.9) 5034 (55.1)

8482 (92.8) 654 (7.2)

8477 (92.8) 659 (7.2)

1260 2072 1422 56 4258 68

998 2479 1368 167 4081 43

0.00

2234 4103 1779 226 4294 118

(17.5) (32.2) (13.9) (1.8) (33.7) (0.9)

0.17 (13.8) (22.7) (15.6) (0.6) (46.6) (0.7)

(10.9) (27.1) (15.0) (1.8) (44.7) (0.5)

0.28 4659 (36.5) 5135 (40.3) 2960 (23.2)

0.03 3886 (42.5) 3827 (41.9) 1423 (15.6)

3929 (43.0) 3725 (40.8) 1482 (16.2)

0.07 8498 (66.6) 4256 (33.4)

0.01 5708 (62.5) 3428 (37.5)

5766 (63.1) 3370 (36.9)

0.18 287 (2.3) 12,467 (97.7)

0.00 0.00

Antidepressant Users (n ¼ 9,136)

0.59 (9.9) (16.3) (11.7) (0.5) (61.0) (0.5)

8054 (63.1) 4700 (36.9)

4607 (1.5) 303,521 (98.5)

Benzodiazepine Users (n ¼ 9136)

0.09

0.00 4540 (1.5) 303,588 (98.5)

0.07 0.02

11,954 (93.7) 800 (6.3)

5748 (45.1) 5333 (41.8) 1673 (13.1)

169,092 (54.9) 139,036 (45.1)

aSD

0.00 5707 (44.7) 7047 (55.3)

0.01 170,428 (55.3) 137,700 (44.7)

5353 (1.4) 382,450 (98.6)

(11.1) (21.5) (21.8) (17.6) (13.1) (8.9) (4.2) (1.9)

5707 (44.7) 7047 (55.3)

1260 2085 1494 69 7778 68

170,007 (55.2) 108,899 (35.3) 29,222 (9.5)

0.02 6283 (1.6) 381,520 (98.4)

1410 2736 2786 2241 1669 1136 530 246

0.03 168,449 (54.7) 112,397 (36.5) 27,282 (8.9)

0.33 242,793 (62.6) 145,010 (37.4)

(11.1) (21.5) (21.8) (17.6) (13.1) (8.9) (4.2) (1.9)

0.35 (8.5) (14.9) (11.1) (0.6) (64.5) (0.5)

0.27 184,412 (47.6) 156,202 (40.3) 47,189 (12.2)

1410 2736 2786 2241 1669 1136 530 246

Antidepressant Users (n ¼ 12,754)

0.01

0.44 (5.5) (9.2) (7.2) (0.9) (69.8) (6.7)

0.01

PS-Matched Cohort (C-statistics ¼ 0.65)

Age-Sex Exact-Matched Cohort Benzodiazepine Users (n ¼ 12,754)

0.01 108,458 (35.2) 199,670 (64.8)

0.06

(9.5) (15.9) (11.4) (0.6) (62.2) (0.5)

aSD

174 (1.4) 118 (0.9) 5015 2492 3030 898

(39.3) (19.5) (23.8) (7.0)

749 (5.9) 12,005 (94.1) 648 (5.1) 206 (1.6) 5821 3561 3804 1122

0.21 0.06

(45.6) (27.9) (29.8) (8.8)

0.13 0.20 0.14 0.07

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

Age group 50e54 55e59 60e64 65e69 70e74 75e79 80e84
85 Sex Female Male Insurance type Health insurance Medical aid Medical institution type Tertiary hospital General hospital Primary hospital Nursing home Clinic Others CCI 0 12 3þ Polypharmacy (
6 medications) Yes No No. of inpatient visits (
3 times) Yes No Comorbidities Depression Traumatic brain injury Hypertension Diabetes mellitus Hyperlipidemia Coronary artery disease Atrial fibrillation Cerebrovascular disease Peripheral vascular disease Myocardial infarction Stroke

Benzodiazepine Users (n ¼ 387,803)

0.00 274 (3.0) 8862 (97.0)

277 (3.0) 8859 (97.0)

174 (1.9) 106 (1.2)

204 (2.2) 108 (1.2)

3735 2054 2357 702

(40.9) (22.5) (25.8) (7.7)

3735 2041 2329 691

(40.9) (22.3) (25.5) (7.6)

0.02 0.00 0.00 0.00 0.01 0.01

3968 (1.0) 21,228 (5.5)

3140 (0.8) 17,737 (4.6)

0.02 0.04

2782 (0.9) 15,065 (4.9)

2771 (0.9) 15,127 (4.9)

0.00 0.00

148 (1.2) 804 (6.3)

245 (1.9) 2163 (17.0)

0.06 0.34

130 (1.4) 770 (8.4)

133 (1.5) 750 (8.2)

0.00 0.01

19,687 (5.1)

12,516 (3.2)

0.09

11,497 (3.7)

11,727 (3.8)

0.00

724 (5.7)

868 (6.8)

0.05

559 (6.1)

574 (6.3)

0.01

2523 (0.7) 14,463 (3.7)

1974 (0.5) 12,721 (3.3)

0.02 0.02

1759 (0.6) 10,666 (3.5)

1767 (0.6) 10,653 (3.5)

0.00 0.00

82 (0.6) 548 (4.3)

109 (0.9) 1748 (13.7)

0.03 0.33

65 (0.7) 542 (5.9)

61 (0.7) 502 (5.5)

0.01 0.02

(continued on next page) 5

6

Table 1 (continued ) Benzodiazepine Users vs Nonusers

Benzodiazepine Users vs Antidepressant Users PS Matched Cohort (C-statistics ¼ 0.69)

Age-Sex Exact-Matched Cohort

Benzodiazepine Users (n ¼ 9136)

Antidepressant Users (n ¼ 9,136)

aSD

984 (7.7) 850 (6.7) 2425 (19.0)

0.02 0.03 0.02

723 (7.9) 592 (6.5) 1769 (19.4)

731 (8.0) 601 (6.6) 1789 (19.6)

0.00 0.00 0.01

1887 (14.8) 9418 (73.8)

2023 (15.9) 9245 (72.5)

0.03 0.03

1361 (14.9) 6567 (71.9)

1399 (15.3) 6597 (72.2)

0.00 0.01

0.00 0.00 0.01 0.00 0.00 0.01 0.01 0.00 0.00

54 3372 7,262 557 329 3276 6479 191 1682

(0.4) (26.4) (56.9) (4.4) (2.6) (25.7) (50.8) (1.5) (13.2)

201 3376 7,142 1357 796 3213 6535 203 3120

(1.6) (26.5) (56.0) (10.6) (6.2) (25.2) (51.2) (1.6) (24.5)

0.12 0.00 0.02 0.24 0.18 0.01 0.01 0.01 0.29

54 2350 5,044 539 304 2267 4595 142 1520

(0.6) (25.7) (55.2) (5.9) (3.3) (24.8) (50.3) (1.6) (16.6)

47 2329 5,116 537 320 2258 4598 143 1511

(0.5) (25.5) (56.0) (5.9) (3.5) (24.7) (50.3) (1.6) (16.5)

0.01 0.01 0.02 0.00 0.01 0.00 0.00 0.00 0.00

0.00 0.00 0.00 0.00 0.00 0.00

1719 2048 4958 1773 1730

(13.5) (16.1) (38.9) (13.9) (13.6)

2354 2807 5903 2734 1525

(18.5) (22.0) (46.3) (21.4) (12.0)

0.14 0.15 0.15 0.20 0.05 0.00

1330 1642 3739 1517 1159

(14.6) (18.0) (40.9) (16.6) (12.7)

1298 1613 3721 1488 1147

(14.2) (17.7) (40.7) (16.3) (12.6)

0.01 0.01 0.00 0.01 0.00 0.05

1992 1467 1394 1170 1019 893 900 1490 1422 1007

(15.6) (11.5) (10.9) (9.2) (8.0) (7.0) (7.1) (11.7) (11.1) (7.9)

1992 1467 1394 1170 1019 893 900 1490 1422 1007

(15.6) (11.5) (10.9) (9.2) (8.0) (7.0) (7.1) (11.7) (11.1) (7.9)

1350 1043 1013 841 697 662 666 1090 1070 704

(14.8) (11.4) (11.1) (9.2) (7.6) (7.2) (7.3) (11.9) (11.7) (7.7)

1445 1029 982 830 700 593 611 1141 1075 726

(15.8) (11.3) (10.7) (9.1) (7.7) (6.5) (6.7) (12.5) (11.8) (7.9)

aSD

Benzodiazepine Users (n ¼ 12,754)

17,867 (5.8) 16,652 (5.4) 45,436 (14.7)

17,805 (5.8) 17,029 (5.5) 44,924 (14.6)

0.00 0.01 0.01

1058 (8.3) 769 (6.0) 2511 (19.7)

0.20 0.33

31,134 (10.1) 206,812 (67.1)

30,762 (10.0) 205,484 (66.7)

0.00 0.01

(0.3) (15.8) (38.7) (2.1) (1.2) (17.5) (32.6) (0.8) (8.4)

0.02 0.26 0.34 0.11 0.09 0.17 0.33 0.06 0.14

910 59,921 146,539 7831 4679 62,575 123,896 2837 29,872

(0.3) (19.4) (49.6) (2.5) (1.5) (20.3) (40.2) (0.9) (9.7)

897 59,567 145,354 7913 4697 61,881 122,581 2802 30,232

(0.3) (19.3) (47.2) (2.6) (1.5) (20.1) (39.8) (0.9) (9.8)

36,977 41,254 124,541 44,392 26,252

(9.5) (10.6) (32.1) (11.4) (6.8)

0.08 0.08 0.09 0.04 0.23 0.00

32,122 35,730 103,691 36,779 26,280

(10.4) (11.6) (33.7) (11.9) (8.5)

32,460 36,118 103,821 37,065 25,959

(10.5) (11.7) (33.7) (12.0) (8.4)

82,093 60,081 48,824 40,689 35,948 32,865 27,395 22,889 21,082 15,937

(21.2) (15.5) (12.6) (10.5) (9.3) (8.5) (7.1) (5.9) (5.4) (4.1)

64,272 47,442 38,818 32,431 28,949 26,282 21,877 18,260 16,879 12,918

(20.9) ()15.4 (12.6) (10.5) (9.4) (8.5) (7.1) (5.9) (5.5) (4.2)

65,056 46,902 37,991 31,939 28,506 26,714 22,215 18,626 17,139 13,040

(21.1) (15.2) (12.3) (10.4) (9.3) (8.7) (7.2) (6.0) (5.6) (4.2)

aSD

Benzodiazepine Users (n ¼ 308,128)

29,883 (7.7) 26,584 (6.9) 72,837 (18.8)

18,308 (4.7) 18,283 (4.7) 46,234 (11.9)

0.12 0.09 0.19

55,929 (14.4) 282,492 (72.8)

31,360 (8.1) 222,823 (57.5)

1376 102,064 214,056 15,123 9544 93,853 189,223 5065 48,816

(0.4) (26.3) (55.2) (3.9) (2.5) (24.2) (48.8) (1.3) (12.6)

993 61,173 149,990 8113 4781 67,997 126,588 2925 32,494

46,390 51,151 141,131 49,782 52,479

(12.0) (13.2) (36.4) (12.8) (13.5)

82,093 60,081 48,824 40,689 35,948 32,865 27,395 22,889 21,082 15,937

(21.2) (15.5) (12.6) (10.5) (9.3) (8.5) (7.1) (5.9) (5.4) (4.1)

Antidepressant Users (n ¼ 12,754)

aSD, absolute standardized difference; CCI, Charlson comorbidity index; COPD, chronic obstructive pulmonary disease; ERT, estrogen replacement therapy; H2RA, histamine 2 receptor antagonists; TCA, tricyclic antidepressants. *Others included bicalutamide, buspirone, digoxin, and tiropramide.

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

COPD Chronic liver disease Chronic pulmonary disease Comedications Anticholinergics Anti-inflammatory analgesics Antipsychotics Antiparkinson agents Narcotic analgesics Antiepileptics TCA Fluoroquinolones H2RA ERT Platelet aggregation inhibitors Anticoagulants Statins Antihypertensives Antidiabetes drugs Others* Index year 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016

aSD

Nonusers (n ¼ 308,128)

Nonusers (n ¼ 387,803)

Benzodiazepine Users (n ¼ 387,803)

PS-Matched Cohort (C-statistics ¼ 0.65)

Age-Sex Exact-Matched Cohort

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

7

Table 2 Risk of Dementia in Benzodiazepine Users Compared With Nonusers and Antidepressant Users by Applied Lag Time Age-Sex Exact-Matched Cohort Person-Years (mean) Benzodiazepine users vs nonusersy Lag-time applied: 7 years Nonusers 2,144,836 (5.5) Benzodiazepine 2,159,414 (5.6) users Lag-time applied: 5 y Nonusers 2,144,836 (5.5) Benzodiazepine 2,159,414 (5.6) users Lag-time applied: 3 y Nonusers 2,144,836 (5.5) Benzodiazepine 2,159,414 (5.6) users Lag-time applied: 1 y Nonusers 2,144,836 (5.5) Benzodiazepine 2,159,414 (5.6) users Benzodiazepine users vs antidepressant Lag-time applied: 7 y Antidepressant 50,641 (4.0) users Benzodiazepine 60,189 (4.7) users Lag-time applied: 5 y Antidepressant 50,641 (4.0) users Benzodiazepine 60,189 (4.7) users Lag-time applied: 3 y Antidepressant 50,641 (4.0) users Benzodiazepine 60,189 (4.7) users Lag-time applied: 1 y Antidepressant 50,641 (4.0) users Benzodiazepine 60,189 (4.7) users

PS-Matched Cohort

No. of Events

IR*

Crude HR (95% CI)

Adjusted HR (95% CI)

Person-Years (Mean)

No. of Events

IR*

Crude HR (95% CI)

Adjusted HR (95% CI)

662 1,479

0.31 0.68

1.00 (Reference) 2.21 (2.01e2.42)

1.00 (Reference) 1.31 (1.28e1.34)

1,685,744 (5.5) 1,718,409 (5.6)

540 1136

0.32 0.66

1.00 (Reference) 2.04 (1.84e2.26)

1.00 (Reference) 1.17 (1.04e1.30)

1642 3728

0.77 1.73

1.00 (Reference) 2.25 (2.13e2.39)

1.00 (Reference) 1.35 (1.33e1.37)

1,685,744 (5.5) 1,718,409 (5.6)

1305 2862

0.77 1.67

1.00 (Reference) 2.13 (2.00e2.28)

1.00 (Reference) 1.23 (1.14e1.32)

2786 6548

1.30 3.03

1.00 (Reference) 2.33 (2.23e2.44)

1.00 (Reference) 1.46 (1.44e1.48)

1,685,744 (5.5) 1,718,409 (5.6)

2212 5029

1.31 2.93

1.00 (Reference) 2.22 (2.11e2.33)

1.00 (Reference) 1.33 (1.26e1.40)

3972 9702

1.85 4.49

1.00 (Reference) 2.42 (2.34e2.51)

1.00 (Reference) 1.54 (1.52e1.56)

1,685,744 (5.5) 1,718,409 (5.6)

3159 7452

1.87 4.34

1.00 (Reference) 2.31 (2.21e2.40)

1.00 (Reference) 1.43 (1.37e1.49)

33

0.65

1.00 (Reference)

1.00 (Reference)

37,342 (4.1)

23

0.62

1.00 (Reference)

1.00 (Reference)

40

0.66

0.97 (0.61e1.54)

1.03 (0.84e1.26)

42,024 (4.6)

24

0.57

0.95 (0.54e1.68)

1.01 (0.81e1.27)

93

1.84

1.00 (Reference)

1.00 (Reference)

37,342 (4.1)

68

1.82

1.00 (Reference)

1.00 (Reference)

103

1.71

0.89 (0.67e1.17)

0.88 (0.78e0.99)

42,024 (4.6)

69

1.64

0.89 (0.64e1.25)

0.69 (0.60e0.78)

193

3.81

1.00 (Reference)

1.00 (Reference)

37,342 (4.1)

137

3.67

1.00 (Reference)

1.00 (Reference)

208

3.46

0.87 (0.72e1.06)

0.98 (0.89e1.08)

42,024 (4.6)

138

3.28

0.88 (0.69e1.11)

0.71 (0.65e0.79)

394

7.78

1.00 (Reference)

1.00 (Reference)

37,342 (4.1)

279

7.47

1.00 (Reference)

1.00 (Reference)

298

4.95

0.63 (0.54e0.73)

0.73 (0.68e0.79)

42,024 (4.6)

206

4.90

0.65 (0.54e0.78)

0.54 (0.50e0.59)

usersz

IR, incidence rate. *Incidence rate was calculated per 1000 person-years. y Exact-matched cohort: adjusted for CCI, polypharmacy, the number of inpatient visits, and all comorbidities and concomitant medications; PS-matched cohort: adjusted by CCI, depression, and cerebrovascular disease. z Exact-matched cohort: adjusted for CCI, the number of inpatient visits, depression, hypertension, diabetes mellitus, hyperlipidemia, cerebrovascular disease, stroke, antipsychotics, antiepileptics, tricyclic antidepressants, platelet aggregation inhibitors, anticoagulants, statins, antihypertensives, and anti-diabetes drugs; PS-matched cohort: CCI, polypharmacy, depression, stroke, antipsychotics, narcotic analgesics, tricyclic antidepressants, and anticoagulants.

mean follow-up period of 5.6 years (1 year: adjusted HR 1.43, 95% CI 1.37e1.49; 3 years: adjusted HR 1.33, 95% CI 1.26e1.40; 5 years: adjusted HR 1.23, 95% CI 1.14e1.32; 7 years: adjusted HR 1.17, 95% CI 1.04e1.30). Table 3 shows the risk of dementia according to the duration of exposure and action of benzodiazepines. The durationresponse association was supported, and concurrent users had the highest risk of dementia. Table 4 shows the risk of dementia according to age, sex, subtype of dementia, insurance type, and medical institution type. Although higher risk of dementia was observed in younger patients, male patients, and national health insurance beneficiaries, there was no difference in the risk associated with medical institution type and dementia subtype. Table 5 shows the results of the secondary analyses of the risk of dementia according to the redefined outcome definitions. The association was consistent regardless of the outcome definitions. Benzodiazepine Users vs Antidepressant Users When benzodiazepine use was compared with antidepressant use, no significant increased risk of dementia was observed for any lag-

time (Table 2). The duration of exposure and action of benzodiazepines were not associated with an increased risk of dementia compared with matched antidepressant users (Table 3). All results from the subgroup analyses showed no statistical significance (Table 4). All secondary analyses were consistent with the results of the primary analysis (Table 5). Discussion Our nationwide cohort study found that, after the first 5 years of exposure, the risk of dementia was 23% higher in benzodiazepine users than in nonusers. Although, initially, there appeared to be a higher risk of dementia with the use of benzodiazepines, no association was observed with benzodiazepine use compared with the use of antidepressants, the active comparator. This result suggests that the association between benzodiazepine and dementia is likely due to confounding by indication. Further, we found that the increased HR of dementia with benzodiazepine use compared with nonuse reduced with longer lag-time. This indicates the potential for protopathic bias and suggests that a causal association is unlikely between

8

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

Table 3 Risk of Dementia, According to Duration-Response and Duration of Action, of Benzodiazepine Use Compared With Nonuse and Antidepressant User

Benzodiazepine users vs nonusersy PDC of benzodiazepine users Short-term (PDC <0.15) Medium-term (0.15  PDC <0.3) Long-term (PDC
0.3) Duration of action Short-acting benzodiazepine users Long-acting benzodiazepine users Concurrent users Benzodiazepine users vs antidepressant usersz PDC of benzodiazepine users Short-term (PDC <0.15) Medium-term (0.15PDC <0.3) Long-term (PDC
0.3) Duration of action Short-acting benzodiazepine users Long-acting benzodiazepine users Concurrent users

Age-Sex Exact-Matched Cohort*

PS-Matched Cohort*

Crude HR (95% CI)

Adjusted HR (95% CI)

Crude HR (95% CI)

Adjusted HR (95% CI)

2.02 (1.90e2.14) 4.81 (4.22e5.48) 5.80 (5.18e6.51)

1.85 (1.74e1.97) 3.74 (3.28e4.27) 4.45 (3.96e5.01)

2.02 (1.94e2.11) 5.58 (5.09e6.12) 7.12 (6.59e7.69)

2.02 (1.94e2.11) 5.11 (4.66e5.61) 6.32 (5.85e6.82)

2.08 (1.95e2.22) 2.43 (2.27e2.60) 2.98 (2.46e3.61)

1.87 (1.74e2.00) 2.18 (2.03e2.34) 2.64 (2.18e3.19)

2.14 (2.04e2.25) 2.48 (2.36e2.60) 3.19 (2.78e3.66)

2.11 (2.02e2.22) 2.48 (2.36e2.60) 3.34 (2.74e3.60)

0.55 (0.34-0.89) 0.60 (0.37-0.97) 0.69 (0.47-1.00)

0.60 (0.36e0.99) 0.61 (0.37e1.01) 0.79 (0.53e1.18)

0.49 (0.27e0.86) 0.64 (0.74e1.10) 0.78 (0.50e1.22)

0.48 (0.27e0.86) 0.67 (0.39e1.16) 0.80 (0.51e1.25)

1.08 (0.79-1.48) 0.66 (0.44-0.98) 0.62 (0.15-2.50)

1.22 (0.89e1.69) 0.73 (0.49e1.10) 0.62 (0.15e2.52)

1.08 (0.75e1.57) 0.64 (0.39e1.05) 0.84 (0.21e3.45)

1.06 (0.73e1.53) 0.64 (0.39e1.06) 0.76 (0.19e3.11)

*5-year lag-time was applied. y Exact-matched cohort: adjusted for CCI, polypharmacy, the number of inpatient visits, and all comorbidities and concomitant medications; PS-matched cohort: adjusted by CCI, depression, and cerebrovascular disease. z Exact-matched cohort: adjusted for CCI, the number of inpatient visits, depression, hypertension, diabetes mellitus, hyperlipidemia, cerebrovascular disease, stroke, antipsychotics, antiepileptics, tricyclic antidepressants, platelet aggregation inhibitors, anticoagulants, statins, antihypertensives, and anti-diabetes drugs; PS-matched cohort: CCI, polypharmacy, depression, stroke, antipsychotics, narcotic analgesics, tricyclic antidepressants, and anticoagulants.

benzodiazepine use and increased risk of dementia. Notably, the HRs for dementia moved closer to null association with a longer duration of lag-time in both cohorts (benzodiazepine users vs nonusers and benzodiazepine users vs antidepressant users). Heterogeneity has been observed in the findings of other studies on the association between benzodiazepine use and the risk of dementia.12e15,18,19,30e35 Though a nested case-control study conducted in Taiwan suggested that the current use of benzodiazepine was associated with a 2.71-fold increased risk of dementia,12 the risk may have been overestimated due to the lack of consideration of disease latency, therefore, including patients potentially prescribed benzodiazepines for prodromal symptoms of dementia. Prospective cohort studies conducted in Europe and the US found 1.6-fold13 and 3.5-fold19 increase in the risk of dementia in benzodiazepine users, respectively, compared with nonusers. However, subsequent prospective cohort studies reported no association.15,18 The association between the long-term use of benzodiazepines and the risk of dementia has been controversial due to inconsistencies in the durationresponse results.12,18,32 Likewise, a nested case-control study using the healthcare database in the United Kingdom found no association with the application of a 4-year lag-time.35 One of the limitations of prospective cohort-based studies is the possible misclassification of person-time. Previous prospective studies followed up participants every 2e3 years to assess their exposure status, using questionnaires.13,15,19 This may have led to exposure misclassification whereby incorrectly classifying persontime as either “exposed” or “unexposed” may overestimate or underestimate the true association. Prospective studies have strength in outcome validity of dementia with the use of cognition evaluation tools, or validation by trained psychologists13,15,18,19; outcome validation is particularly important to distinguish pseudodementia symptoms in dementia research. To overcome limitations in outcome validity, we used prespecified algorithms to define dementia and conducted multiple sensitivity analyses with redefined outcome definitions. To address bias from confounding by indication, we selected antidepressants as an active comparator for benzodiazepines, as these drugs have overlapping indications, such as depression and

anxiety,23,24,36 which frequently coexist.37 Depression and anxiety also share common symptoms encompassing general distress. A recent prospective cohort study found no association between SSRI use (excluding paroxetine) and increased risk of dementia compared with nonuse.38 This result supports the use of antidepressants as an effective negative control; however, risk in paroxetine users can be partially attributable to the negative association observed when an active comparator was used in our study.38 Nonetheless, this association is unlikely to be causal given the potential risk overestimation owing to the high prevalence of cerebrovascular disease in paroxetine users and the nondose-response relationship.38 The increased risk of dementia, reported in previous studies, with the use of benzodiazepine or antidepressant may have been confounded by depression. Depression may be a risk factor or prodromal symptom of dementia.39 Observational studies have observed a 2- to 5-fold increase in the risk of dementia among patients with depression.39 This may support our results that the risk of dementia in benzodiazepine users, compared with nonusers, decreased after the time-varying adjustment of major confounders of dementiaddepression and cerebrovascular disease. The risk estimates declined from a crude HR of 2.13 to 1.23 after the adjustment. The observed duration response in benzodiazepine users compared to nonusers may not be a true effect of the use of benzodiazepines as the entire matched nonusers were used as a reference for the estimation of HR. Long-term benzodiazepine users are more likely to have more severe and long-term mental health disorders than short-term users, therefore, resulting in a higher risk of dementia because of the condition rather than the treatment of that condition. In addition, the duration-response with higher PDC with benzodiazepines was not observed when antidepressants were used as the comparator. Although we did not identify an increased risk of dementia with benzodiazepines, these drugs should still be prescribed with caution as long-term benzodiazepine use is associated with addiction, prolonged sedation, and delirium.40 High prevalence of potentially inappropriate benzodiazepine use in patients with dementia41 is problematic as long-term benzodiazepine use may aggravate symptoms.

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

9

Table 4 Stratified Analysis for Risk of Dementia in Benzodiazepine Users Compared With Nonusers and Antidepressant Users Age-Sex Exact-Matched Cohort*

Benzodiazepine users vs nonusersy Age group 50e54 55e59 60e64 65e69 70e74 75e79 80e84
85 P for interaction Sex Female Male P for interaction Insurance type Healthcare insurance Medical aid P for interaction Medical institution type Tertiary hospital General hospital Primary hospital Nursing home Clinic Others P for interaction Type of dementia Alzheimer’s disease Vascular dementia Others{ P for interaction Age group 50e54 55e59 60e64 65e69 70e74 75e79 80e84
85 P for interaction Benzodiazepine users vs. antidepressant Sex Female Male P for interaction Insurance type Healthcare insurance Medical aid P for interaction Medical institution type Tertiary hospital General hospital Primary hospital Nursing home Clinic Others P for interaction Type of dementia Alzheimer’s disease Vascular dementia Othersk P for interaction

PS-Matched Cohort*

Crude HR (95% CI)

Adjusted HR (95% CI)

Crude HR (95% CI)

Adjusted HR (95% CI)

5.22 3.95 2.86 2.30 1.99 1.84 1.64 1.67

(3.43e7.93) (2.98e5.24) (2.32e3.52) (2.00e2.65) (1.76e2.24) (1.63e2.09) (1.42e1.90) (1.36e2.04)

5.25 (3.42e8.05) 3.78 (2.83e5.06) 2.74 (2.21e3.40) 2.18 (1.88e2.52) 1.96 (1.73e2.23) 1.83 (1.60e2.08) 1.66 (1.42e1.93) 1.54 (1.25e1.91) <.0001

4.96 4.19 3.22 2.27 1.95 1.80 1.57 1.49

(3.53e6.96) (3.36e5.22) (2.74e3.79) (1.93e2.67) (1.70e2.24) (1.56e2.07) (1.33e1.84) (1.18e1.87)

4.90 (3.49e6.87) 4.16 (3.33e5.19) 3.22 (2.74e3.79) 2.29 (1.94e2.69) 1.95 (1.70e2.24) 1.80 (1.56e2.07) 1.57 (1.34e1.85) 1.49 (1.18e1.87) <.0001

2.02 (1.87e2.19) 2.56 (2.35e2.79)

1.83 (1.68e1.99) 2.33 (2.13e2.55) <.0001

2.10 (1.99e2.22) 2.59 (2.43e2.75)

2.08 (1.97e2.20) 2.57 (2.41e2.74) <.0001

2.18 (2.06e2.32) 2.20 (1.80e2.69)

2.00 (1.88e2.13) 1.99 (1.61e2.46) <.0001

2.30 (2.20e2.41) 2.10 (1.85e2.37)

2.29 (2.19e2.39) 2.10 (1.86e2.37) <.0001

2.02 2.14 2.47 2.59 2.52 2.59

(1.57e2.60) (1.79e2.55) (1.95e3.12) (1.26e5.31) (2.35e2.71) (1.95e3.45)

2.11 (1.62e2.74) 2.08 (1.72e2.50) 2.20 (1.73e2.81) 2.62 (1.18e5.84) 2.19 (2.03e2.36) 2.35 (1.74e3.18) .1985

2.23 2.32 2.75 2.85 2.53 2.01

(1.87e2.66) (2.06e2.62) (2.33e3.24) (1.83e4.46) (2.40e2.66) (1.59e2.54)

2.41 (2.02e2.88) 2.46 (2.18e2.77) 2.68 (2.27e3.17) 2.97 (1.89e4.65) 2.49 (2.36e2.62) 2.00 (1.58e2.52) .2751

0.97 (0.82e1.14) 1.00 (0.94e1.07) 1.05 (0.90e1.23)

1.03 (0.86e1.23) 1.02 (0.95e1.09) 1.04 (0.88e1.23) *

0.98 (0.81e1.19) 1.04 (0.97e1.12) 1.04 (0.87e1.24)

1.08 (0.96e1.22) 1.07 (1.02e1.12) 1.10 (0.99e1.22) *

0.46 0.51 0.40 1.19 0.74 0.78 -

0.53 (0.17e1.63) 0.59 (0.26e1.37) 0.36 (0.18e0.72) 1.28 (0.68e2.40) 0.78 (0.41e1.47) 0.51 (0.15e1.70) <.0001

0.67 0.47 0.37 1.18 0.82 0.94 -

0.64 (0.18e2.29) 0.51 (0.19e1.39) 0.37 (0.16e0.82) 1.22 (0.60e2.45) 0.81 (0.38e1.73) 0.63 (0.18e2.22) <.0001

1.01 (0.70e1.46) 0.74 (0.48e1.14)

1.16 (0.79e1.70) 0.80 (0.51e1.26) .0001

0.99 (0.63e1.56) 0.80 (0.49e1.31)

0.95 (0.60e1.50) 0.78 (0.48e1.29) .0168

0.90 (0.66e1.21) 0.90 (0.43e1.89)

1.00 (0.73e1.37) 0.98 (0.45e2.13) <.0001

0.96 (0.67e1.37) 0.55 (0.22e1.37)

0.96 (0.67e1.38) 0.42 (0.16e1.12) <.0001

0.80 0.84 0.70 0.99 1.57 1.28

0.95 (0.38e2.40) 0.86 (0.49e1.51) 0.73 (0.30e1.80) 0.19 (0.00e10.13) 1.65 (0.98e2.80) 1.23 (0.06e24.07) .9911

0.73 0.88 0.65 1.05 1.38 0.52

0.73 (0.27e1.97) 0.88 (0.48e1.60) 0.63 (0.26e1.54) 1.91 (0.26e13.85) 1.26 (0.70e2.29) 0.79 (0.05e13.77) .8527

1.65 (0.51e5.41) 0.85 (0.59e1.24) 2.54 (0.60e10.79)

1.98 (0.80e4.93) 0.78 (0.53e1.15) 0.77 (0.28e2.14)

(0.16e1.33) (0.23e1.12) (0.21e0.78) (0.65e2.17) (0.41e1.34) (0.27e2.26)

(0.19e2.38) (0.17e1.27) (0.17e0.83) (0.59e2.36) (0.39e1.70) (0.29e3.07)

usersz

(0.33e1.92) (0.48e1.45) (0.30e1.63) (0.18e5.43) (0.94e2.61) (0.08e20.48)

1.84 (0.79e4.26) 0.77 (0.56e1.06) 0.91 (0.39e2.11)

{

(0.27e1.96) (0.48e1.60) (0.27e1.55) (0.19e5.72) (0.77e2.47) (0.03e8.33)

1.80 (0.48e6.75) 0.70 (0.46e1.07) 0.72 (0.15e3.45) {

*5-year lag-time was applied. y Exact-matched cohort: adjusted for CCI, polypharmacy, the number of inpatient visits, and all comorbidities and concomitant medications; PS-matched cohort: adjusted by CCI, depression, and cerebrovascular disease. z Exact-matched cohort: adjusted for CCI, the number of inpatient visits, depression, hypertension, diabetes mellitus, hyperlipidemia, cerebrovascular disease, stroke, antipsychotics, antiepileptics, tricyclic antidepressants, platelet aggregation inhibitors, anticoagulants, statins, antihypertensives and anti-diabetes drugs; PS-matched cohort: CCI, polypharmacy, depression, stroke, antipsychotics, narcotic analgesics, tricyclic antidepressants, and anticoagulants. k Others include dementia in other specified diseases classified elsewhere, unspecified dementia, dementia in Pick’s disease and dementia with Lewy bodies. { P value for interaction was not calculated, because subtype of dementia was an outcome variable.

10

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

Table 5 Sensitivity Analyses for Risk of Dementia in Benzodiazepine Users Compared With Nonusers and Antidepressant Users, According to Re-defined Outcome Definitions

Benzodiazepine users vs nonusersy Strict outcome definitions Restricted to secondary diagnosis and treatment for dementiaz Diagnosis from tertiary hospital and treatment for dementiax Other instrumental definitions
1 diagnosis and treatment for dementiak
1 inpatient or 2 outpatient diagnosis or treatment for dementia** Dementia diagnosis or treatment for dementiayy Benzodiazepine users vs antidepressant userszz Strict outcome definitions Restricted to secondary diagnosis and treatment for dementiaz Diagnosis from tertiary hospital and treatment for dementiax Other instrumental definitions
1 diagnosis and treatment for dementiak
1 inpatient or 2 outpatient diagnosis or treatment for dementia** Dementia diagnosis or treatment for dementiayy

Age-Sex Exact-Matched Cohort*

PS-Matched Cohort*

Crude HR (95% CI)

Adjusted HR (95% CI)

Crude HR (95% CI)

Adjusted HR (95% CI)

2.19 (2.06e2.33) 2.35 (2.18e2.53)

1.96 (1.84e2.09) 2.08 (1.92e2.25)

2.08 (1.94e2.23) 2.21 (2.03e2.40)

2.07 (1.93e2.21) 2.19 (2.02e2.39)

2.23 (2.11e2.36) 2.13 (2.05e2.22) 2.14 (2.06e2.22)

2.00 (1.89e2.12) 1.90 (1.82e1.98) 1.90 (1.82e1.97)

2.11 (1.99e2.24) 2.01 (1.91e2.10) 2.02 (1.93e2.11)

2.10 (1.97e2.23) 2.01 (1.92e2.10) 2.00 (1.91e2.10)

0.96 (0.71e1.29) 0.84 (0.59e1.19)

1.07 (0.79e1.44) 0.96 (0.67e1.38)

0.95 (0.67e1.35) 0.92 (0.60e1.41)

0.94 (0.67e13.4) 0.92 (0.61e1.41)

0.86 (0.66e1.11) 0.88 (0.72e1.07) 0.80 (0.67e0.96)

0.97 (0.74e1.26) 1.01 (0.82e1.23) 0.91 (0.75e1.09)

0.90 (0.66e1.23) 1.00 (0.79e1.26) 0.90 (0.72e1.12)

0.90 (0.66e1.23) 1.00 (0.79e1.26) 0.90 (0.72e1.11)

*5-year lag-time was applied. y Exact-matched cohort: adjusted for CCI, polypharmacy, the number of inpatient visits, and all comorbidities and concomitant medications; PS-matched cohort: adjusted by CCI, depression, and cerebrovascular disease. z Presence of
1 inpatient or 2 outpatient dementia diagnosis, restricted to primary and secondary diagnosis, and a prescription for dementia medication x Presence of dementia diagnosis from tertiary hospital and a prescription of dementia medication. k Presence of
1 dementia diagnosis record from either inpatient or outpatient setting and a prescription for dementia medication **Presence of either
1 in- or 2 outpatient dementia diagnosis or a prescription of dementia medication yy Presence of either dementia diagnosis or a prescription of dementia medication zz Exact-matched cohort: adjusted for CCI, the number of inpatient visits, depression, hypertension, diabetes mellitus, hyperlipidemia, cerebrovascular disease, stroke, antipsychotics, antiepileptics, tricyclic antidepressants, platelet aggregation inhibitors, anticoagulants, statins, antihypertensives, and anti-diabetes drugs; PS-matched cohort: CCI, polypharmacy, depression, stroke, antipsychotics, narcotic analgesics, tricyclic antidepressants, and anticoagulants.

Our study has multiple strengths. First, we assembled a large population-based cohort by using data from a nationwide database over 15 years. Second, the database includes all inpatient and outpatient prescription records, as South Korea’s healthcare system is based on fee-for-service reimbursement, thereby eliminating immeasurable time bias, which arises as a result of the absence of inpatient drug exposure information.42 Third, we applied epidemiologic designs to overcome time-related bias as well as bias from confounding by indication. A new-user design was applied, thereby eliminating biases related to prevalent users. Time-lag bias was avoided by matching the duration between cohort entry and initiation date of follow-up for the selection of comparators.26 The consideration of lag time and active comparator design reduced bias from reverse causation and confounding by indication, respectively.43 Finally, we applied rigorous statistical methods, including PS matching to increase comparability between groups, and treated comorbidity and comedication as timedependent variables. However, our findings should be interpreted with caution due to the following limitations. Our study has the potential for misclassification of dementia diagnosis. A validation study comparing the diagnoses recorded in the South Korean healthcare database with electronic medical records found an overall positive predictive value of 82%. By its nature, disentangling dementia from pseudodementia symptoms, such as depression or anxiety, is challenging. Therefore, we used prespecified algorithms to reduce outcome misclassification and tested the robustness of results by redefining outcome definitions. Residual confounding from unmeasured confounders (ie, alcohol consumption, smoking, family history of dementia, physical activity, and body mass index) may remain. Higher HRs for dementia in benzodiazepine users among younger and male patients in the stratified analysis may be because they are more susceptible to drug-induced dementia risk, or residual confounders may have exerted excessive influence specifically in younger and male patients who had low dementia incidences.

Conclusions and Implications There was a significant association between benzodiazepine use and the risk of dementia. However, the strength of the association was reduced with the application of longer lag-times and removed when an active comparator was used, which suggests the absence of a causal association. Nonetheless, considering the controversial reports on the association between benzodiazepines and dementia, the use of benzodiazepines still warrants the attention of clinicians and policymakers. Acknowledgments We sincerely thank National Health Insurance Service for their cooperation in providing access to the database (Data number: NHIS-2018-1-071). Lisa Kalisch Ellett is supported by an NHMRC-ARC Dementia Research Development Fellowship (Grant identification number APP1101788). References 1. Donoghue J, Lader M. Usage of benzodiazepines: A review. Int J Psychiatry Clin Pract 2010;14:78e87. 2. Cheng JS, Huang WF, Lin KM, et al. Characteristics associated with benzodiazepine usage in elderly outpatients in Taiwan. Int J Geriatr Psychiatry 2008; 23:618e624. 3. Stranks EK, Crowe SF. The acute cognitive effects of zopiclone, zolpidem, zaleplon, and eszopiclone: A systematic review and meta-analysis. J Clin Exp Neuropsychol 2014;36:691e700. 4. Limon A, Reyes-Ruiz JM, Miledi R. Loss of functional GABA(A) receptors in the Alzheimer diseased brain. Proc Natl Acad Sci USA 2012;109:10071e10076. 5. Puustinen J, Nurminen J, Vahlberg T, et al. CNS medications as predictors of precipitous cognitive decline in the cognitively disabled aged: A longitudinal population-based study. Dement Geriatr Cogn Disord Extra 2012;2:57e68. 6. Dean L. Diazepam therapy and CYP2C19 genotype. In: Pratt V, McLeod H, Rubinstein W, et al., editors. Medical Genetics Summaries. Bethesda (MD): National Center for Biotechnology Information (US); 2012. 7. Wedlund PJ. The CYP2C19 enzyme polymorphism. Pharmacology 2000;61: 174e183.

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11 8. Zhang YA, Reviriego J, Lou YQ, et al. Diazepam metabolism in native Chinese poor and extensive hydroxylators of S-mephenytoin: Interethnic differences in comparison with white subjects. Clin Pharmacol Therapeut 1990;48:496e502. 9. Lucchetta RC, da Mata BPM, Mastroianni PC. Association between development of dementia and use of benzodiazepines: A systematic review and metaanalysis. Pharmacotherapy 2018;38:1010e1020. 10. Gomm W, von Holt K, Thome F, et al. Regular benzodiazepine and Z-substance use and risk of dementia: An analysis of German claims data. JAD 2016;54: 801e808. 11. Mawanda F, Wallace RB, McCoy K, et al. PTSD, psychotropic medication use, and the risk of dementia among US veterans: A retrospective cohort study. J Am Geriatr Soc 2017;65:1043e1050. 12. Wu CS, Ting TT, Wang SC, et al. Effect of benzodiazepine discontinuation on dementia risk. Am J Geriatr Psychiatry 2011;19:151e159. 13. Billioti de Gage S, Begaud B, Bazin F, et al. Benzodiazepine use and risk of dementia: Prospective population based study. BMJ (Clinical research ed.) 2012;345:e6231. 14. Billioti de Gage S, Moride Y, Ducruet T, et al. Benzodiazepine use and risk of Alzheimer’s disease: Case-control study. BMJ (clinical research ed.) 2014;349: g5205. 15. Shash D, Kurth T, Bertrand M, et al. Benzodiazepine, psychotropic medication, and dementia: A population-based cohort study. Alzheimer Dement 2016;12: 604e613. 16. Imfeld P, Bodmer M, Jick SS, et al. Benzodiazepine use and risk of developing Alzheimer’s disease or vascular dementia: A case-control analysis. Drug Safety 2015;38:909e919. 17. Bietry FA, Pfeil AM, Reich O, et al. Benzodiazepine use and risk of developing Alzheimer’s disease: A case-control study based on Swiss claims data. CNS Drugs 2017;31:245e251. 18. Gray SL, Dublin S, Yu O, et al. Benzodiazepine use and risk of incident dementia or cognitive decline: Prospective population based study. BMJ (clinical research ed) 2016;352:i90. 19. Gallacher J, Elwood P, Pickering J, et al. Benzodiazepine use and risk of dementia: Evidence from the Caerphilly Prospective Study (CaPS). J Epidemiol Commun Health 2012;66:869e873. 20. Shin JY, Choi NK, Jung SY, et al. Overlapping medication associated with healthcare switching among Korean elderly diabetic patients. J Korean Med Sci 2011;26:1461e1468. 21. Kim DS, Lee HJ, Son IJ, et al. Retrospective drugs utilization review study for chronic kidney disease using national health insurance database. Yakhak Hoeji 2009;53:138e144. 22. Park EC, Jang SI, Jeon SY, et al. Report of the evaluation for validity of discharged diagnoses in Korean Health Insurance database. Health Insurance Review and Assessment Service; 2017. 23. Lee JY, Kang WS, Kim JW, et al. Prescription patterns of benzodiazepine for outpatients in a psychiatric department in Korea. Anxiety Mood 2015;11:143e148. 24. Lampe L. Drug treatment for anxiety. Aust Prescriber 2013;36:186e189. 25. Richardson K, Fox C, Maidment I, et al. Anticholinergic drugs and risk of dementia: Case-control study. BMJ (clinical research ed) 2018;361:k1315.

11

26. Suissa S, Azoulay L. Metformin and the risk of cancer: Time-related biases in observational studies. Diabetes Care 2012;35:2665e2673. 27. Campigotto F, Weller E. Impact of informative censoring on the Kaplan-Meier estimate of progression-free survival in phase II clinical trials. J Clin Oncol 2014;32:3068e3074. 28. Parsons L. Reducing bias in a propensity score matched-pair sample using greedy matching techniques 2001, Proceedings of the twenty-sixth Annual SASÒ users group international conference. Cary, NC: SAS Institute Inc; 2001. p. 214e216. 29. Knopman DS. Dementia and cerebrovascular disease. Mayo Clin Proc 2006;81: 223e230. 30. Paterniti S, Dufouil C, Alperovitch A. Long-term benzodiazepine use and cognitive decline in the elderly: The Epidemiology of Vascular Aging Study. J Clin Psychopharmacol 2002;22:285e293. 31. Lagnaoui R, Tournier M, Moride Y, et al. The risk of cognitive impairment in older community-dwelling women after benzodiazepine use. Age Ageing 2009;38:226e228. 32. Wu CS, Wang SC, Chang IS, et al. The association between dementia and longterm use of benzodiazepine in the elderly: Nested case-control study using claims data. Am J Geriatr Psychiatry 2009;17:614e620. 33. Lagnaoui R, Begaud B, Moore N, et al. Benzodiazepine use and risk of dementia: A nested case-control study. J Clin Epidemiol 2002;55:314e318. 34. Verdoux H, Lagnaoui R, Begaud B. Is benzodiazepine use a risk factor for cognitive decline and dementia? A literature review of epidemiological studies. Psychol Med 2005;35:307e315. 35. Richardson K, Mattishent K, Loke YK, et al. History of nenzodiazepine prescriptions and risk of dementia: Possible bias due to prevalent users and covariate measurement timing in a nested case-control study. Am J Epidemiol 2019;188:1228e1236. 36. Lee JY, Kim MS, Lee SH, et al. Korean medication algorithm project for generalized anxiety disorder 2009 (II): Medication algorithm and long-term medication treatment strategy. J Korean Neuropsychiatr Assoc 2010;49:553e563. 37. Lydiard RB. Coexisting depression and anxiety: Special diagnostic and treatment issues. J Clin Psychiatry 1991;52:48e54. 38. Heath L, Gray SL, Boudreau DM, et al. Cumulative antidepressant use and risk of dementia in a prospective cohort study. J Am Geriatr Soc 2018;66: 1948e1955. 39. Byers AL, Yaffe K. Depression and risk of developing dementia. Nat Rev Neurol 2011;7:323e331. 40. Markota M, Rummans TA, Bostwick JM, et al. Benzodiazepine use in older adults: Dangers, management, and alternative therapies. Mayo Clin Proc 2016; 91:1632e1639. 41. Eshetie TC, Nguyen TA, Gillam MH, et al. Potentially inappropriate prescribing in people with dementia: An Australian population-based study. Int J Geriatr Psychiatry. 2019;34:1498-1505. 42. Suissa S. Immeasurable time bias in observational studies of drug effects on mortality. Am J Epidemiol 2008;168:329e335. 43. Stella F, Radanovic M, Balthazar ML, et al. Neuropsychiatric symptoms in the prodromal stages of dementia. Curr Opin Psychiatry 2014;27:230e235.

11.e1

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

Supplementary Table 1 Definition of Exposure, Outcomes, and Comorbidities Exposure (ATC codes) Benzodiazepines Antidepressants SSRIs SNRIs Outcomes Dementia diagnosis (ICD-10) Medication prescription for treating dementia (ATC codes) Comorbidities (ICD-10) Depression Traumatic brain injury Hypertension Diabetes mellitus Hyperlipidemia Coronary artery disease Atrial fibrillation Cerebrovascular disease Peripheral vascular disease Myocardial infarction Stroke Chronic obstructive pulmonary disease Chronic liver disease Chronic pulmonary disease

N05CD, N05BA, N03AE N06AB03-N06AB06, N06AB08, N06AB10 N06AX16, N06AX17, N06AX21, N06AX23 F00-F03, G30, G31.00, G31.82 N06DA01-N06DA04, N06DX

F32, F33 S020, S021, S027, S029, S060-S069 I100, I150, I152, I158, I159 E10-E14 E780-E785 I20-I25 I48 G45,G46, I60-I69, I71 I739, I790, R02, Z958, Z959 I21, I22 G45, H341, I60-I64 J41-J44 B15-B19, C22, K70, K73, K74 J40- J47, J60- J67, J684, J701, J703, J841, J920, J961, J982, J983

ICD, International Classification of Diseases, 10th Revision; ATC, Anatomical Therapeutic Chemical Classification System.

Supplementary Table 2 Classification of Benzodiazepines and Antidepressants Long-Acting Benzodiazepines

Short-Acting Benzodiazepines

Clordiazepoxide Chlordiazepoxideþclinidium Clobazam Clonazepam Clorazepate Diazepam Estazolam Ethyl loflazepate Flunitrazepam Flurazepam

Alprazolam Bromazepam Brotiazolam Blotiazepam Etizolam Lorazepam Midazolam Oxazepam Tofisopam Triazolam

Selective Serotonin Reuptake Inhibitors

Serotoninenorepinephrine Reuptake Inhibitors

Citalopram Escitalopram Fluoxetine Fluvoxamine Paroxetine Sertraline

Duloxetine Milnacipran Venlafaxine Desvenlafaxine

Y.-H. Baek et al. / JAMDA xxx (2019) 1e11

11.e2

Supplementary Table 3 List of Comedications Drug Class

Drug Name

Anticholinergics

Benztropine, Biperiden, Cimetropium, Fesoterodine, Flavoxate, Glycopyrrolate, Mebeverine, Orphenadrine, Oxybutynin, Pinaverium, Procyclidine, Scopolia Ext, Solifenacin, Tolterodine, Trihexyphenidyl, Trospium Aceclofenac, Acetaminophen, Aspirin, Dexibuprofen, Dexketoprofen, Diclofenac, Diflunisal, Etodolac, Fenbufen, Fenoprofen, Flurbiprofen, Ibuprofen, Indomethacin, Ketoprofen, Ketorolac, Loxoprofen, Meclofenamate sodium, Mefenamic acid, Naproxen, Oxaprozin, Sodium salicylate, Sulindac, Amisulpiride, Aripiprazole, Chlorpromazine, Clozapine, Haloperidol, Olanzapine, Paliperidone, Pimozide, Quetiapine, Risperidone, Sulpiride, Thioridazine, Trifluoperazine, Ziprasidone Haloperidol, Levosulpiride, Metoclopramide, Metoclopramide, Olanzapine, Risperidone, Sulpiride Buprenorphine, Butorphanol, Codeine, Dihydrocodeine, Dihydrocodeine,combinations, Dihydrocodeineþcodeine,combinations, Fentanyl, Hydromorphone, Morphine, Nalbuphine, Oxycodone, Oxycodone,combinations, Pentazocine, Pethidine, Tapentadol, Tramadol, Tramadol,combinations Carbamazepine Clonazepam, Ethosuximide, Fosphenytoin, Gabapentin, Lamotrigine Levetiracetam Oxcarbazepine, Perampanel, Phenobarbital, Phenytoin Pregabalin Primidone, Rufinamide, Stiripentol, Topiramate, Valproic acid, Vigabatrin, Zonisamide, Amitriptyline, Amoxapine, Clomipramine, Dothiepin, Doxepin, Imipramine, Maprotiline, Nortriptyline, Quinupramine, Ciprofloxacin, Enoxacin, Fleroxacin, Gatifloxacin, Gemifloxacin, Levofloxacin, Lomefloxacin, Moxifloxacin, Norfloxacin, Ofloxacin, Pefloxacin, Sparfloxacin, Tosufloxacin Cimetidine, Famotidine, Lafutidine, Nizatidine, Ranitidine, Roxatidine Estradiol, Estriol Beraprost sodium, Cilostazol, Clopidogrel, Dipyridamole, Ginkgo biloba leaf extract, Ginkgo biloba leaf extract þ cilostazol, Ginkgo biloba leaf extract þ ticlopidine, Limaprost, Ozagrel sodium, Sarpogrelate, Sulodexide, Ticlopidine, Triflusal Apixaban, Aspirin, Edoxaban, Fondaparinux, Parnaparin Rivaroxaban, Tenecteplase, Urokinase, Warfarin Atrovastatin, Atrovastatinþs-amlodipine, Pravastatin, Rosuvastatin, s-Amlodipine, besylate, Simvastatin Acebutolol, Alacepril, Amiloride, Amlodipine, Arotinolol hcl, Atenolol, Atenololþchlorthalidone, Azosemide, Barnidipine, Benazepril, Betaxolol, Bevantolol hcl, Bisoprolol hemifumarate, Candesartan cilexetil, Candesartanþhydrochlorothiazide, Captopril, Captoprilþhydrochlorothiazdie, Carteolol, Carvedilol, Chlorthalidone, Enalapril, Enalaprilþhydrochlorothiazide Eprosartan mesylate, Eprosartanþhydrochlorothiazide, Esmolol, Felodipine Felodipineþmetoprolol, Fimasartan potassium, Fimasartanþhydrochlorothiazide, Fosinopril, Furosemide, Hydrochlorothiazide, Imidapril, Indapamide, Indapamide, Irbesartan, Irbesartanþhydrochlorothiazide, Labetalol, Lercanidipine, Lisinopril, Lisinoprilþhydrochlorothiazide, Losartan potassium, Losartanþchlorthalidone, Losartanþhydrochlorothiazde, Metoprolol succinate, Metoprolol tartrate, Metoprololþhydrochlorothiazide, Moexipril, Nebivolol hydrochloride, Nicardipine hcl, Nifedipine, Nilvadipine, Nimodipine, Nitroglycerin, Olmesartan cilexetil, Olmesartan, medoxomil, Olmesartanþhydrochlorothiazide, Perindopril tertbutylamine, Perindoprilþindapamide, Propranolol, Propranololþhydrochlorothiazide, Quinapril, Ramipril, Ramiprilþfelodipine, Ramiprilþhydrochlorothiazide, S-atenolol, Sotalol, Spironolactone, Spironolactoneþhydrochlorothiazide, Telmisartan, Telmisartanþhydrochlorothiazide, Temocapril, Torasemide, Valsartan, Valsartanþhydrochlorothiazide, Valsartanþlercanidipine, Zofenopril calcium Acarbose, Albiglutide, Alogliptin, Alogliptinþmetformin, Alogliptinþpioglitazone, Anagliptin, Chlorpropamide, Dapagliflozin, Dulaglutide, Empagliflozin, Evogliptin, Exenatide, Gemigliptin, Glibenclamidee, Gliclazide, Glimepiride, Glimepirideþrosiglitazone, Gliquidone, Insulin, Ipragliflozin, Linagliptin, Lixisenatide, Lobeglitazone, Metformin, Metforminþanagliptin, Metforminþdapagliflozin Metforminþempagliflozin, Metforminþevogliptin, Metforminþgemigliptin, Metforminþglibenclamide, Metforminþgliclazide, Metforminþglimepiride, Metforminþlinagliptin, Metforminþlobeglitazone, Metforminþmitiglinide, Metforminþnateglinide, Metforminþpioglitazone, Metforminþrepaglinide, Metforminþrosiglitazone, Metforminþsaxagliptin, Metforminþsitagliptin, Metforminþteneligliptin, Metforminþvildagliptin, Metforminþvoglibose, Miglitol, Mitiglinide, Nateglinide, Pioglitazone, Repaglinide, Rosiglitazone, Saxagliptin, Sitagliptin, Teneligliptin, Vildagliptin, Voglibose Bicalutamide, Buspirone, Digoxin, Tiropramide

Antiinflammatory analgesics

Antipsychotics Antiparkinson agents Narcotic analgesics

Antiepileptics Tricyclic antidepressants Fluoroquinolones Histamine 2 receptor antagonists Estrogen replacement therapy Platelet aggregation inhibitors Anticoagulants Statins Antihypertensives

Antidiabetics

Others