Abstract
Background:
Benzodiazepines and related drugs (BZDR) are occasionally used to treat certain symptoms of Alzheimer’s disease (AD). However, the risks related to BZDR use are high in older persons. Although frequent BZDR use has been reported in persons with AD, no previous study has focused specifically on the incidence of BZDR use in this population.
Objective:
We investigated the incidence of BZDR use in persons with and without AD during a five-year follow-up.
Methods:
The Finnish nationwide, register-based MEDALZ cohort includes all AD cases who received a clinically verified AD diagnosis in 2005–2011 (n = 70,718) and their matched comparison persons. Incidence of BZDR, including benzodiazepines (lorazepam, oxazepam, temazepam, alprazolam, chlordiazepoxide, diazepam, and nitrazepam) and Z-drugs (zolpidem and zopiclone), use was investigated in the cohort from two years before to three years after the diagnosis of AD. Further, initial BZDRs were investigated.
Results:
The incidence of BZDR use was higher in persons with AD starting from 12 months before the diagnosis and peaked at six months after the diagnosis of AD (incidence rate ratio [IRR] = 2.6, 95% confidence interval [CI] = 2.5–2.8). Benzodiazepines were more frequently initiated by persons with AD, with the incidence peaking at six months after the diagnosis (IRR = 4.5, 95% CI = 4.1–4.9) and remaining over three times higher than in comparison persons until three years after the diagnosis.
Conclusion:
Early symptomatic treatment with BZDRs is contrary to AD treatment guidelines. As BZDRs impair cognition, the observed early treatment with BZDRs may complicate the monitoring of AD treatment effectiveness.
INTRODUCTION
The number of persons with Alzheimer’s disease(AD), the most common cause of dementia, isincreasing as the population ages [1]. Up to 97% of persons with dementia experience behavioral andpsychological symptoms of dementia (BPSD) during the course of the disorder [2], and similar prevalence has been observed specifically in persons with AD [3].
Benzodiazepines and related drugs (BZDR) can be used in treatment of particular BPSDs, including anxiety, agitation, and sleep disturbances [4]. They can be used in short-term treatment in persons with AD if non-pharmacological and other pharmacological treatments have not been adequate or if only infrequent treatment is needed. Older persons, however, are more sensitive to the effects of these drugs, resulting in an increased risk of adverse drug events [5]. BZDR use increases the risk of falls and hip fractures [6, 7], which may further lead to other severe outcomes including increased health care service use, morbidity, and mortality [8]. High age and multiple comorbidities predispose older persons to BZDR dependence [9]. Especially long-term benzodiazepine use has been associated with cognitive decline and an increased risk of AD [10, 11].
A previous cross-sectional and nationwide study in Finland demonstrated that 29% of persons with AD used BZDRs during the first year after the diagnosis of AD [12]. Further, a longitudinal study in Finland demonstrated 45% of persons with AD using BZDRs during a four-year follow-up [13]. In both studies, the prevalence of BZDR use was higher in persons with AD compared with their matched comparison persons without AD. Several other studies have reported a high prevalence of BZDR use in persons with AD in various settings [3, 14–26]. Although the risks related to BZDR use in older persons are well known and the adverse effects on cognition have been demonstrated, there are no previous studies regarding the incidence of BZDR use in persons with AD. Similarly, the temporal relationship between the initiation of BZDR use and AD diagnosis has not been previously assessed.
The aim of this study was to investigate theincidence of benzodiazepine and related drug use from two years before to three years after the diagnosis of AD in a Finnish nationwide cohort of community-dwelling persons with and without AD.
MATERIALS AND METHODS
Cohort
This nationwide, register-based study was based on the MEDALZ (Medication use and Alzheimer’s disease) cohort [12]. The cohort included all Finnish persons diagnosed with AD between 2005 and 2011 (n = 73,005). Persons with dementia related to Parkinson’s disease were excluded, leaving 70,718 persons with AD in the study population. For each person with AD, four age-, gender-, and region of residence-matched comparison persons without AD were selected from a nationwide register including all residents. As a result, the cohort consisted of 353,576 persons in total (Fig. 1). The cohort [12] and the applied registers [27] have been previously described in more detail.
Persons with the verified diagnosis of AD were identified from the Special Reimbursement Register maintained by the Social Insurance Institution (SII). The verified diagnosis included symptoms consistent with mild to moderate AD, decrease in social capacity, computed tomography or magnetic resonance imaging scan, exclusion of alternative diagnoses, and confirmation of the diagnosis by a neurologist or geriatrician [28]. The diagnosis of AD was based on the NINCDS-ADRDA and DSM-IV criteria [29, 30] and it has been described in more detail previously [27]. The validity of the AD diagnosis based on the Special Reimbursement Register has been demonstrated [31].
Drug use
Data on drug use from years 1995–2012 were obtained from the Prescription Register. The Prescription Register contains information on each reimbursed drug purchase for all residents in Finland [27]. All drugs are coded according to the Anatomical Therapeutic Chemical (ATC) classification [32]. The register data relevant for this study included package size, strength, dosage form, date of purchase, and the purchased amount in defined daily doses [33]. Drugs used during stays at hospitals or public nursing homes are not recorded in the register and, therefore, only community-dwelling persons were included in the analyses.
BZDRs were categorized according to the ATC classification as benzodiazepines (classes N05BA and N05CD) and benzodiazepine-related drugs (“Z-drugs”, N05CF). Lorazepam, oxazepam, and temazepam were categorized as medium-acting benzodiazepines whereas alprazolam, chlordiazepoxide, diazepam, and nitrazepam as long-acting benzodiazepines. Zolpidem and zopiclone were categorized as Z-drugs. Midazolam, triazolam, and zaleplon were excluded from the analyses because they were not reimbursed during the study period. Furthermore, clobazam was excluded from the analyses because it is indicated only for the treatment for epilepsy. The Prescription Register provides valid information on BZDR use in older community-dwelling Finnish persons, especially in regular drug use [34].
Fixed dosage assumptions based on tablets or defined daily doses are not valid in older Finnish persons [35, 36]. Thus, BZDR use was modeled with the PRE2DUP method [37]. This method was based on modeling individual purchasing behavior by calculating the sliding averages of the daily dose. The method considered regularity of drug purchases, possible stockpiling of drugs and hospitalization periods. Each drug was modeled separately according to the ATC classification. The model provided drug use periods, i.e., dates of the beginning and the end of continuous drug use. The use of benzodiazepines and Z-drugs was modeled similarly. For this study, we combined overlapping benzodiazepine and Z-drug use periods to retrieve the duration of BZDR use. During BZDR use, persons were allowed to switch between drug substances unless there were any breaks in drug use. BZDR polypharmacy (use of two or more BZDRs concomitantly) was defined as overlapping drug use periods of ≥60 days.
Study design
One-year washout period for BZDR use was utilized to exclude prevalent users from the analyses (Fig. 1). The washout period began three years before and ended two years before the diagnosis of AD, including the corresponding matching date of comparison persons. Persons with any BZDR use during the washout period were excluded from all analyses. Other exclusion criteria were ≥6 months hospital stay during the washout period and ongoing ≥90 days hospital stay at the end of the washout period because drug use during hospitalizations was not known. Furthermore, comparison persons for the excluded persons with AD were excluded. Finally, persons with AD without any comparison persons were excluded to maintain the matching.
For the included persons, the study follow-up began two years before and ended three years after the diagnosis of AD or the corresponding matching date unless the follow-up was censored for any of the following reasons: ≥90 days hospitalization, death, end of study (December 31, 2012), or diagnosis of AD for the comparison persons. The AD status was considered as a time-dependent variable and, in total, 7,802 comparison persons were diagnosed with AD within three years after the matching date. For these comparison persons, the study follow-up in the comparison cohort ended two years before the diagnosis of AD. The incidence of BZDR use was investigated during the five-year study follow-up.
Comorbidities
Comorbid conditions recorded by the time of AD diagnosis, including the corresponding date of comparison persons, were obtained for each person from the Special Reimbursement Register since 1972. The Special Reimbursement Register contains information on each person’s entitlement to higher reimbursement of drugs due to chronic diseases [27]. A modified Charlson’s comorbidity index (CCI, including chronic heart failure, coronary artery disease, diabetes, asthma or chronic obstructive pulmonary disease [COPD], disseminated connective tissue diseases, rheumatoid arthritis and other comparable conditions, uremia requiring dialysis, severe anemia in connection with chronic renal failure, leukemia and other malignant diseases of blood and bone marrow including malignant diseases of the lymphatic system, breast cancer, prostate cancer, gynecological cancer, and malignant neoplasms) was computed for each person based on the comorbidities as in a previous study [38]. In addition, data on epilepsy, hypothyroidism and all chronic cardiovascular diseases (coronary artery disease, heart failure, hypertension, and arrhythmias) were obtained from the register.
Data on history of psychiatric disorders were collected from the Hospital Discharge Register maintained by the National Institute of Health and Welfare. Data since 1972 were applied in this study. The register contains information on admission and discharge dates as well as person’s diagnoses related to the inpatient admissions according to the International Statistical Classification of Diseases and Related Health Problems (ICD) classification [39]. Only psychiatric diagnoses from hospitalizations at least five years before the diagnosis of AD, including the corresponding date of comparison persons, were considered to exclude hospitalizations due to the prodromal symptoms of AD. Diagnoses for schizophrenia, schizotypal or delusional disorders (ICD-10 codes F20-29; ICD-9 codes 295, 297, 298, 3010 and 3012; ICD-8 codes 295, 297, 298, 29999, 30100 and 30120), bipolar disorder (ICD-10 codes F30-31; ICD-9 codes 2962, 2963, 2964 and 2967; ICD-8 codes 29610, 29620, 29630, 29688 and 29699), and severe depression (ICD-10 codes F32-39; ICD-9 codes 2961, 2968, 3004 and 3011; ICD-8 codes 29600, 30040, 30041 and 30110) were collected from the register.
Statistical analyses
The five-year incidence of BZDR use was defined as having at least one purchase of the drugs during the study follow-up. Incidence rates (IR) were calculated for each six-month interval during the five-year follow-up as a rate per 100 person-years. The IRs were compared between persons with AD and comparison persons with the Poisson regression and the results were reported as incidence rate ratios (IRR) with 95% confidence intervals (CI). Separate IR values for benzodiazepine and Z-drug use were calculated from persons initiating drug use with monotherapy.
Participants’ characteristics according to incident BZDR use were compared with the chi-squared test for categorical variables and with t-test or Mann-Whitney U test for the continuous variables. Mean, median, range, and interquartile range (IQR) were applied in describing the continuous variables.
The prevalence of BZDR polypharmacy was analyzed as having at least one polypharmacy period during the five-year follow-up. In addition, the duration of first BZDR use period was investigated. Initial drugs were investigated among BZDR users who initiated with monotherapy. Further, the number of BZDR initiations within 10 days after hospital discharge was investigated. All analyses were performed using SAS (version 9.3; SAS Institute Inc., Cary, NC, USA).
No ethics approval or informed consent from the cohort was required by the Finnish legislation as the persons in the sample were not contacted and only de-identified data were utilized. Permissions for data use were received from the maintainers of registers.
RESULTS
Study population
The study population included 211,955 persons with AD and comparison persons without AD in total (Table 1). The mean age of the study population at the time of AD diagnosis was 79.1 years (range 34.1–104.6). Most persons (60.3%) were women. The median follow-up time was 1,826 days (IQR = 1,371–1,826) in the whole study population, 1,687 days (IQR = 1,259–1,826) in persons with AD, and 1,826 days (IQR = 1,402–1,826) in comparison persons (p < 0.0001).
Incident benzodiazepine and related drug use
Altogether, 13,341 (25.7%) persons with AD and 27,028 (16.9 %) comparison persons initiated BZDR use during the five-year follow-up (p < 0.0001). BZDR users were more likely to be women (Table 1). Furthermore, BZDR users were more likely to have higher CCI, cardiovascular diseases, asthma/COPD, hypothyroidism, and history of psychiatric disorders.
The IR of BZDR use was higher in persons with AD compared with those without AD starting from 12 months before the diagnosis of AD and the difference remained significant until the end of the five-year follow-up (Fig. 2a). The difference between the IRs of BZDR use was at highest at six months after the diagnosis, when the IR in persons with AD was 11.2/100 person-years and in comparison persons 4.1/100 person-years (IRR = 2.6, 95% CI = 2.48–2.76). The IR of BZDR use decreased gradually in comparison persons during the five-year follow-up.
The IR of benzodiazepine use was higher in persons with AD throughout the five-year follow-up (Fig. 2b). The difference between the IRs was at highest at six months after the diagnosis, the IR being 6.4/100 person-years in persons with AD and 1.4/100 person-years in comparison persons (IRR = 4.5, 95% CI = 4.13–4.88). After the diagnosis, the IR of benzodiazepine use in persons with AD remained three to four times higher than in those without AD.
The difference between the IRs of Z-drug use was at highest at six months after the AD diagnosis (Fig. 2c), when the IRs were 4.4/100 person-years and 2.7/100 person-years in persons with and without AD, respectively (IRR = 1.6, 95% CI = 1.46–1.71). When compared to persons without AD, the IR of Z-drug use in persons with AD was higher from the diagnosis to 18 months after diagnosis.
Among the incident BZDR users, initiation with benzodiazepines was more frequent in persons with AD (54.0%) than in comparison persons (35.2% , p < 0.0001) during the five-year follow-up (Table 2). Medium-acting benzodiazepines were the initial drugs for 47.4% of BZDR users with AD and 26.7% of comparison persons using BZDRs (p < 0.0001). Use of long-acting benzodiazepines was less frequent in BZDR users with AD (6.6%) compared with comparison persons (8.6 %) using BZDRs (p < 0.0001). BZDR users with AD (46.0%) initiated drug use with Z-drugs less frequently compared with comparison persons (64.8% , p < 0.0001).
The median duration of the first BZDR use period was 121 days (IQR = 40–349) in persons with AD and 88 days (IQR = 37–211) in comparison persons (p < 0.0001). BZDR polypharmacy was twice as frequent in incident BZDR users with AD (9.1% , n = 1,213) compared with those without AD (4.4% , n = 1,189) (p < 0.0001). Zopiclone and oxazepam were the most common drugs used in combination by concomitant BZDR users with and without AD (24.7% and 27.3% , respectively, p = 0.1462). Among the incident BZDR users, 20.3% (n = 2,706) of persons with AD and 12.2% (n = 3,289) of comparison persons initiated their BZDR use within 10 days after discharge from hospital care (p < 0.0001).
DISCUSSION
To our knowledge, this is the first study to report the incidence of BZDR use in persons with AD. This incidence was higher in persons with AD compared with those without AD from one year before to three years after the diagnosis of the disease. Benzodiazepines were the most frequently initiated drugs in persons with AD and a considerable increase in benzodiazepine incidence occurred around the time of AD diagnosis, peaking at six months after the diagnosis.
The early symptomatic treatment with BZDRs in persons with AD is not in line with treatment guidelines [4, 28]. Cholinesterase inhibitors (ChEI) are the first-line pharmacological treatment for AD [4, 28] and more than 80% of persons with AD in Finland use them shortly after the diagnosis [38]. It may take up to six months to observe the beneficial effects of ChEIs [28]. Concomitant treatment with BZDRs may prevent monitoring the effectiveness of ChEI treatment because BZDR use has been associated with cognitive decline in persons with AD [19, 40].
Older persons are more sensitive to the central nervous system effects of BZDRs due to age-related changes in pharmacodynamics and pharmacokinetics [5] and, therefore, BZDR use may cause daytime sedation and impair alertness in this population [41]. BZDRs affect cognition, gait, and balance as well, resulting in an increased risk of falls and fractures [42]. Although the risk of injuries is at highest at the beginning of BZDR use, the risk remains high throughout the treatment. Therefore, the longer duration of the first BZDR use period in persons with AD compared with those without AD in the current study is alarming. Longer duration of use in persons with AD has been demonstrated in prevalent users as well, as a previous nationwide study in Finland reported 30% of persons with AD and 26% of persons without AD having long-term (≥180 days) BZDR use during a four-year follow-up (13).
Initiation of BZDR use during the first ten days after hospital discharge was more frequent in persons with AD compared with those without AD in our study. A previous study reported a similar finding in general older population as 13% of older persons initiated benzodiazepine use within 10 days after discharge [43]. The same study reported a three-fold risk of receiving initial BZDR prescription associated with hospital care. High initiation rate in our study was observed especially in persons with AD, which may result from increasing number and severity of BPSDs during acute hospitalization [44].
BZDR polypharmacy was twice as frequent inincident BZDR users with AD than in those withoutAD in this study. A previous longitudinal study reported that the 3-month prevalence of concomitant use of ≥2 benzodiazepines was 25% among benzodiazepine users with AD [17]; however, in that study, the total duration of concomitant use was not considered. An increased risk of adverse drug events has been associated with BZDR polypharmacy in general older population, as polypharmacy increased the risk of hip fractures more than BZDR monotherapy [45, 46]. The higher prevalence of BZDR polypharmacy in persons with AD raises concern as AD itself has been shown to be a risk factor for falls and fractures [47, 48] and, therefore, BZDR as well as psychotropic polypharmacy should be avoided in this population [28].
As BPSDs are highly prevalent in newly-diagnosed persons with AD [49], the higher incidence of BZDR use in persons with AD in this study may reflect the treatment of BPSDs. However, nonpharmacological treatment should be the first-line option in BPSD treatment [2, 4] and if pharmacological treatment is needed, cholinesterase inhibitors are the primary option [50, 51]. Further, if cholinesterase inhibitors are not effective enough, benzodiazepines can be used in infrequent or short-term treatment of anxiety and agitation [4]. However, the evidence on the effectiveness of benzodiazepines in the treatment for BPSDs in AD is insufficient [51] and the risk of associated adverse drug events is considerable [2]. It is noteworthy that benzodiazepine use may actually worsen the BPSDs. In the current study, benzodiazepine use consisted mainly of medium-acting drugs which are preferred over long-acting drugs because of smaller risk of accumulation due to delayed elimination and associated excess daytime sedation [52]. The Z-drugs can be used if treatment of mainly sleep disturbances is needed [4], but there are no randomized controlled trials on BZDR use for this indication in persons with AD [53]. The pharmacological treatment of BPSDs is problematic as there are safety concerns related to other pharmacological treatment options as well [51].
Due to the nationwide cohort in this study, the selection bias was minimal. Therefore, our results are representative of the community-dwelling Finnish persons with AD. In addition, major strengths of our research include the clinically verified diagnoses of AD and follow-up for the study population both before and after the diagnosis. The Finnish registers are comprehensive [27] and, therefore, they enable high-quality pharmacoepidemiological research. Our exposure modeling approach, based on the purchased amounts of drugs and the intervals between purchases obtained from the Prescription Register, enabled more realistic estimation on duration of drug use than the fixed dosage assumptions. With the drug use model, we defined BZDR polypharmacy as the concomitant use for ≥60 days to differentiate it from drug switches. Our data, with one-year washout period, enabled us to investigate the incidence of drug use. Furthermore, we were able to compare the incidence between persons with AD and the matched comparison persons without AD.
There are some recognized limitations regarding the use of register-based data. The Prescription Register did not include some small packages of non-reimbursed BZDRs, which may have led to minor underestimation of BZDR use in this study population. Moreover, the register-based data did not include BZDR use during hospital care or nursing home stays. The register data did not contain indication for BZDR use, either. Another limitation of this study concerns the lack of information on severity of AD and BPSDs. Further, history of psychiatric disorders was likely to be underestimated as the Hospital Discharge Register included only the most serious cases requiring inpatient treatment. However, the amount of underestimation was unlikely to differ between persons with and without AD.
In conclusion, the incidence of benzodiazepine and related drug use was higher in persons with AD compared with those without the disease from one year before the diagnosis. As clinicians, we are concerned about the incidence peak at six months after the diagnosis since early treatment with these drugs is contrary to treatment guidelines of the disease. Further, the concomitant use of these and antidementia drugs complicates the assessment of the effectiveness of antidementia treatment because benzodiazepines and related drugs have adverse effects on cognition. Due to the increasing use of benzodiazepines and related drugs before the diagnosis of AD, further studies on the associated risks are urgently needed.
