Retrospective Population Cohort Study on Hip Fracture Risk Associated with Zolpidem Medication

Abstract

Study Objective:

Few studies have evaluated the hip fracture risk for zolpidem users. We assessed the risk for subjects taking zolpidem.

Design:

Population-based retrospective cohort study using claims data of a universal insurance system.

Participants:

We identified 6,978 patients newly prescribed for zolpidem in 2000-2001 age 18 y and older, and 27,848 nonusers frequency matched with sex, age, and date visiting a clinic.

Measurements and Results:

Both cohorts were followed up to the end of 2008 to measure the hip fracture incidence and risk, which considered factors such as sex, age, occupation, days of drug use, and osteoporosis status. The zolpidem users had a 2.23-fold higher hip fracture incidence than nonusers (3.10 versus 1.39 per 1,000 person-y). The risk increased with age for both cohorts. The elderly users had a 21-fold higher incidence than the younger users, or twofold higher than the elderly nonusers. Among 33 patients (20.4%) with hip fracture occurring during presumed medication days, which was accountable for an incidence of 1,083.0 per 1,000 person-y. Those taking the medicine for 8 days or longer had a moderately higher fracture rate than those taking it for less days (6.02 versus 4.48 per 100 person-times) with a ratio of 1.34 (95% confidence interval 0.42-4.56). Subjects with blue collar occupations were at a higher fracture risk.

Conclusion:

The hip fracture risk of zolpidem users is higher than that of nonusers. Fracture prevention awareness should be disseminated to the users.

Citation:

Lin FY; Chen PC; Liao CH; Hsieh YW; Sung FC. Retrospective population cohort study on hip fracture risk associated with zolpidem medication. SLEEP 2014;37(4):673-679.

INTRODUCTION

The trend of using sedatives-hypnotics to treat psychiatric problems such as anxiety, insomnia, and other disorders has significantly increased.¹ Benzodiazepines have been recognized as the mainstay of psychoactive drugs for medications over barbiturates in the past few decades.² Benzodiazepines potentially lead to abuse, dependence, cognitive function impairment, falls, and hip fractures.^3–11 Fractures are major life-threatening public health concerns, especially for the aging population. Fracture risk is also a concern for patients taking benzodiazepines and newer sedative-hypnotic types including zopiclone, zolpidem, and zaleplon (z-hypnotics). The use of these new types has been greatly promoted because of their superior safety over benzodiazepines as short-acting hypnosedatives.¹²

The newer nonbenzodiazepine hypnotics have a short half-life and are effective and well tolerated. Thus, these hypnotics have been prescribed in a rapidly increasing trend.¹³ Zopiclone was introduced to Taiwan in 1990, zolpidem in 1993, and zaleplon in 2003. The prescriptions of z-hypnotics for the elderly has been increasing in Taiwan, with zolpidem the most frequently prescribed.¹⁴ Thus, our attention has focused on zolpidem across all age groups to determine whether patients taking the drug are at higher hip fracture risk than those taking benzodiazepines.

This concern has been increased by studies that compared the risks of hip and other nonvertebral fractures between zolpidem and other psychotropic medications.^15–17 However, one case-control study included a small sample size of patients taking zolpidem.¹⁴ A cohort study compared fracture risk before and after initiation for each drug treatment.¹⁵ Both studies found a 1.9- to 2.5-fold increase in the risk of hip fracture and nonvertebral fractures among zolpidem users, which was approximately similar to that of those who used diazepam. No risk variations by sex, age, and medication duration are examined in these studies.

Thus far, data on the association between zolpidem use and hip fractures for populations are insufficient. In this report, we conducted a population-based study, comparing the newly prescribed zolpidem patients with nonusers to evaluate the subsequent hip fracture risk using the claims data of a universal insurance system. We also evaluated the risk associated with medication duration, and examined if osteoporosis interacts with zolpidem to induce a higher risk.

METHODS

Study Population

This study was based on information derived from the National Health Research Database (NHRD) of Taiwan, which included a randomly selected sample with one million insured people from the National Health Insurance (NHI). The NHI covers more than 96% of both health care providers and all 23 million citizens since the universal health insurance program was established in 1995.¹⁸ The available claims data included demographic characteristics, enrollment dates, and outpatient data for the period of 1996 to 2008. Diagnoses are coded according to the International Classification of Diseases, Ninth Revision (ICD-9). The outpatient prescription data include information on drug name, dose, quantity, duration, and date dispensed for each year from 1997 to 2008. We linked files using scrambled personal identifications to secure individuals' confidentiality, preventing the ethical violations.

Study Design

We conducted a retrospective cohort study for subjects age 18 y and older who were identified from the NHRD to compare hip fracture rates between cohorts with and without zolpidem usage. All study subjects were enrolled in the insurance for at least 2 y before the study cohorts were established. Patients who received an initial zolpidem prescription from 2000 to 2001 without any other sedative-hypnotic preparation since 1997 were identified as the zolpidem user cohort. Subjects in the database without any history of zolpidem and other sedative-hypnotic use were designated as the nonuser cohort. Prescriptions of hypnotics and sedatives were defined by the WHO Anatomical Therapeutic Chemical classification system. For each zolpidem user, four nonusers were randomly selected and frequency matched by sex, age, and the date when the zolpidem user was identified. Individuals with a hip fracture history at the baseline were excluded from the study cohorts. Users of anxiolytic and antiepileptic medications (including benzodiazepines) were included in this analysis.

Hip fractures often occur concurrently with falls. However, we had no data in this study on the time of day in which a fall, if any, might have occurred. Incident hip fracture cases were identified for both cohorts with the first hip fracture diagnosis (ICD-9 codes 820.xx) during the follow-up period. All study subjects were followed up from baseline until the hip fracture event was identified or censored for loss, withdrawn from the insurance, or until the end of 2008, depending on which date came first.

We used the category that was registered in the insurance as a proxy measure for socioeconomic status.¹⁹ We categorized civil servants, administrative employers, and business and industrial setting employees for the white-collar group with higher socioeconomic status. Farmers, fishermen, crewman, and industrial laborers were in the blue-collar group. Other insured groups were military personnel and veterans, including also those with low income and who were retired or unemployed. In this study, the total zolpidem medication dispensed for each patient was measured based on the number of days the medicine was prescribed for the patient during the follow-up period.

Statistical Analysis

Data analysis first evaluated the chronological trends of prescriptions for zolpidem, benzodiazepines, and other nonbenzodiazepines from 1997 to 2008 to show whether zolpidem had become the primary hypnotic prescribed in Taiwan. Demographic characteristics, and the histories of osteoporosis (ICD-9 codes 733.0x), insomnia (ICD-9 codes 307.4x and 780.5x), and of taking anxiolytics and antiepileptics were compared in the zolpidem user and nonuser cohorts. The differences between cohorts were examined using the chi-square test. The incident hip fracture cases and incidence rates in the follow-up period were estimated by the demographic variable and the status of osteoporosis, insomnia, and of taking anxiolytics and antiepileptics for both cohorts. Poisson regression analysis was used to measure the zolpidem user cohort against nonuser cohort crude and adjusted incidence rate ratios (IRR) and 95% confidence intervals (CI). The Kaplan-Meier procedure was performed to determine the cumulative hip fracture probabilities between both cohorts by the end of 2008, which was examined using the log-rank test.

For zolpidem users, we compared the proportional distribution of hip fracture events and the rates that occurred during the periods of taking the drug and afterward to represent drug influencing and noninfluencing periods, respectively. We counted the postmedication events by 1-7, 8-14, 15-30, 31-60, 61-120, 121-210 and > 210 days after the medication. For those with the fracture that occurred during medication, we further calculated that the fracture rate was estimated in per 100 person-times by days the drug prescribed for. The insurance in Taiwan allowed prescribing zolpidem for a patient generally for up to 3 days at each clinic visit. Other specialized health care services could provide prescription for 1 week to several weeks. Zolpidem was available at two different dosages, 6.25 mg and 10 mg in Taiwan. These different dosages were combined into a single category in all analyses described in the following paragraphs, because physicians prescribed only 10 mg medicine. Approximately 85% of prescriptions for zolpidem were prescribed mainly for nightly use, and others were prescribed for use on an as-needed basis.

Data analysis also evaluated whether osteoporosis interacting with zolpidem usage was associated with the hazards of hip fracture. We used the multivariable Cox proportional hazards regression to estimate the hazard ratio (HR) and 95% CI of the hip fracture controlling for other potential risk factors. The proportionality test for the Cox model had a P value of 0.17. The statistical significant level was at P < 0.05. All analyses were performed with SAS 9.2 (SAS Institute, Cary, NC).

RESULTS

The chronologic pattern of major hypnotics prescribed for the study population in Taiwan showed a steady increase of zolpidem use from 1998 to 2008 (Figure 1). Zolpidem has become the most frequently prescribed hypnotic in Taiwan, and approximately 65% of hypnotic prescriptions were for zolp-idem in 2008.

Figure 1.

Prescriptions of each dispensed sedative-hypnotic in Taiwan from 1997 to 2008. Prescriptions are given in numbers of each drug prescribed by total dispensed sedatives-hypnotics per 100.

The study cohorts comprised 6,978 subjects in the zolpidem user cohort and 27,848 nonusers identified from 2000 to 2001. Table 1 shows the characteristics of the study populations. Both cohorts had more women and fewer elderly persons. More than 70% of subjects had white-collar jobs, and the proportion was higher in zolpidem users than for nonusers. The rates of using zolpidem were higher among those with osteoporosis, insomnia, and those taking anxiolytics and antiepileptics than for those individuals not having these conditions or taking those classes of medications (P < 0.0001).

Table 1.

Baseline characteristics of study subjects between zolpidem users and nonusers

During the follow-up period, 162 and 319 hip fracture events were recorded in the zolpidem and nonuser cohorts, respectively (Table 2). The incidence was 2.23-fold higher for zolpidem users than non-users (3.10 versus 1.39 per 1,000 person-y). The incidence increased with increasing age in both cohorts with the highest value for the elderly zolpidem users, which was 2.08-fold higher than in the elderly nonusers (11.1 versus 5.34 per 1,000 person-y). However, the age-specific adjusted IRR was the highest at 6.50 (95% CI = 5.45-7.76) for those age 18-44 y. The hip fracture risk was much higher for those with blue-collar jobs, as well as those with low income, those who were retired, and those who were unemployed. No significant difference was observed in military-related personnel. Among users, the incidence was 3.5-fold higher for those with blue-collar jobs than those with white-collar jobs (6.86 versus 1.95 per 1,000 person-y). However, the multivariable Poisson model shows that low- income, retired, and unemployed populations remained at higher IRR, compared to those with white-collar jobs. Zolpidem users with osteoporosis or insomnia, or those taking anxiolytics were also more likely to have hip fractures than nonusers. The Poisson model shows that osteoporosis, insomnia, and antiepileptic consumption were factors that also predicted the fracture risk.

Table 2.

Incidence of hip fractures and zolpidem users to nonusers rate and rate ratio by demographic status and baseline medical status

The Kaplan-Meier model showed the cumulative hip fracture probability for both cohorts (Figure 2). The cumulative hip fracture probability was significantly higher for zolpidem users than in nonusers (log-rank P < 0.0001).

Figure 2.

Kaplan-Meier survival curves comparing the incident hip fracture probability for zolpidem users or nonusers.

Table 3 shows that 20.4% (n = 33) of hip fractures were identified during the period of taking zolpidem. With an overall exposure time of 31.8 person-y, these cases were accountable for an incidence rate of 1,038.0 per 1,000 person-y. Only 0.6% of fracture cases appeared 1-7 days after using the drug, accountable for an incidence of 13.7 per 1,000 person-y. The remainder of fractures appeared to decline with washout time. Among 33 patients with hip fracture occurred during the medication days, those taking the medicine for 8 days or longer had 1.34-fold (95% CI 0.42-4.56) of higher hip fracture rate than those taking it for less days (6.02 versus 4.48 per 100 person-times; Table 4).

Table 3.

Proportional distribution and incidence of hip fracture events in zolpidem users according to when events occurred

Table 4.

Distribution of hip fracture occurring during medication in patients with hip fracture by days of prescription (N = 33)

Table 5 shows that the Cox model assessed results for the interaction between osteoporosis and zolpidem use. Overall, the incidence of hip fracture increased further, to 4.65 per 1,000 person-y for zolpidem users with osteoporosis, with an unadjusted HR of 4.68 (95% CI 3.65-6.01), compared with nonusers without osteoporosis who had an incidence of 0.98 per 1,000 person-y. The adjusted HR was 2.44 (95% CI 1.85-3.22) after controlling for covariates, with a P value of 0.8 for the interaction term of osteoporosis × zolpidem.

Table 5.

Interaction between osteoporosis and zolpidem use in association with hip fracture in Cox regression analysis controlling for demographic factors, insomnia, anxiolytics, and antiepileptics

DISCUSSION

This Asian population-based study reports that patients taking zolpidem were associated with elevated risk of hip fracture. Studies on sedative-hypnotic uses have consistently found increased fracture risks mainly for the elderly who used benzodiazepine or other drugs related to benzodiazepines.^{6,7,9–11,15–17} A case-control study compared the hip fracture risk associated with sleep-promoting agents in the elderly population in Medicare and in other special assistance programs in New Jersey.¹⁵ Zolpidem use is associated with an odds ratio of nearly 2.0 for fracture, which is similar to that of those taking benzodiaze-pine or other drugs found by Wagner et al. in a cohort study.⁸ In a longitudinal study on health maintenance organization members, Finkle et al. compared the nonvertebral fracture rates before and after zolpidem and benzodiazepine treatments.¹⁶ They also found a twofold increase in risk after using zolpidem, which was similar to using diazepam. The risk decreased with time since the initial treatment.

It is not clear whether the association between hip fracture and hypnotic use remain confounded by other factors. In this study we found that the risk of the association between hip fracture and zolpidem use was influenced by female sex, older age, lower socioeconomic status, presence of insomnia, and usage of antiseizure medications, but the association remained after adjustment for these variables. There is an additive effect between zolpidem use and osteoporosis in association with hip fracture, but the effect declines after controlling for covariates.

Our study comparing the first-time zolpidem users with those without any sedative-hypnotic treatment revealed an overall hip fracture risk similar to the finding of Finkle et al.¹⁶ The incidence is especially higher for the elderly or those with low socioeconomic status. The risk further increased for those in whom osteoporosis and insomnia had been diagnosed. Among the zolpidem users, the fracture incidence was nearly 21-fold higher for the elderly users than for the young users (11.1 versus 0.52 per 1,000 person-y) or 2.1-fold higher than the elderly nonusers (5.34 per 1,000 person-y). Therefore, zolpidem use increases the risk among the elderly. We further evaluated zopiclone users and found that the incidence of hip fracture was 2.9-fold greater in users than in nonusers (data not shown). The fracture incidence was nearly 16.3-fold higher for the elderly users than for the young users. Therefore, the elderly “Z” drug users are at increased risk. Zaleplon was not covered under the national insurance plan in Taiwan during our study period, and therefore data related to its use could not be analyzed in this study.

This study also showed that men and persons working in low-income occupations are at much higher risk of hip fracture. Subjects with such occupations have to wake earlier for jobs in the morning and have greater exposure to potentially detrimental effects of zolpidem because of potentially shorter time in bed after taking the medication. In addition, hard physical work associated with many low-income jobs may itself represent a risk for hip fractures. Absence of information in this database on time of day when hip fracture as a result of a fall occurred did not allow us to determine these possibilities. Thus, the association between zolpidem use and falls remains subject to possible residual confounding by these and other variables.

The biologic mechanisms by which zolpidem and benzodiazepines cause fractures remain uncertain. Zolpidem is similar to most benzodiazepines during liver oxidation, and are impaired with advancing age.^20,21 Visual hallucination distortion is one of the neuropsychiatric reactions induced by zolp-idem,^22,23 which could increase the incidence of falling leading to hip fracture. Zolpidem is effective in initiating sleep with a half-life of 2 to 3 h.^24,25 Zolpidem is similar to benzodiazepines in inducing hallucinations and also results in adverse cognitive effects. Increasing the zolpidem concentration in the blood also increases its potential harmful effects.^22,23,26

In a study of Michigan nursing homes, Avidan et al. failed to find that hypnotic drug use was associated with risk of hip fracture, probably because of small sample size.²⁷ Insomnia was related to hip fracture in that study. In our much larger database derived from insurance claims, we could not determine whether poor sleep independent of use of zolpidem was associated with hip fractures. Without concurrent information in our study about zolpidem efficacy and measures of sleep quality on particular nights on which falls occurred, this possibility remains.

The strength of this study is use of a retrospective cohort comparison between hypnotic users and nonusers to observe the natural history of the subsequent risk. We attempted to evaluate whether prolonged use of zolpidem increased hip fracture risk, but found that the risk is maintained at a constant level after taking zolpidem for longer than 1 week. The most important finding in this study is that approximately 20% of fracture episodes occur during the medication days, accountable for an incidence rate of 1,038.0 per 1,000 person-y. The long-term users are at slightly higher incident fracture than short-term users, but the difference is not significant. The incidence declines sharply when medication is not being administered. This finding supports the hypothesis that the medication induces hallucinations, which are assumed to be responsible for fractures because of falls and other accidents.^22,23,26 This study indicates that, when placed in the context of number of days since the final day of each patient's zolpidem prescription, the incidence of hip fracture events is minor 2 weeks after the date of the last prescription.

Nonetheless, some of the associations that have emerged in our analyses may reflect individuals using zolpidem on an as-needed basis many days after their original prescription was filled. Our data might also imply that observed association between zolpidem and hip fracture may be partially artifactual, because subjects of lower socioeconomic status and subjects with osteoporosis were at much higher risk. Labor-oriented occupations and those with osteoporosis have increased risks to falls. Osteoporosis is more prevalent among the elderly. This skeletal disorder increases with fracture risk.^28,29 Women, particularly postmenopausal women with accelerated bone loss, are at increased risk for osteoporosis.³⁰ However, the current study failed to prove that women are at a higher risk than men in the zolpidem cohort, probably because men are predominant in the labor-oriented job group.

This study has other limitations. The drug-dispensing data reflect the routine practice for the representative populations rather than clinical studies. A filled prescription was used as a proxy for zolpidem use. Also, the results in this study do not completely reflect the actual usage or the effect of over-the-counter (OTC) medications. Although most sedative-hypnotic users were monitored by health administration, the OTC medications may confer reduced hip fracture risk relative to prescription medications. We observed all study subjects for at least 2 y prior to the study for detailed diagnoses and drug exposure and to ensure that no sedative-hypnotic preparation was taken before the exposure assessment period. However, under the WHO Anatomical Therapeutic Chemical classification system that we used to define sedative and hypnotic prescriptions, anxiolytics included benzodiazepines, and use of such medications may have included some individuals whose benzodiaze-pine prescription was a de facto hypnotic. The misclassification of zolpidem exposure thus still cannot be ruled out. Benzodiazepines and benzodiazepine-related drugs were considered hip fracture risk factors. Therefore, a restraint was made on the exposure subjects only for zolpidem users to reduce the confounding effect from other sedatives-hypnotics in this study. Other hip fracture risk factors include low bone mineral density, weight loss,³¹ protein binding affinity,²² and cytochrome P450 inhibition, which can reduce zolpidem clearance during their metabolism.²⁰ Our database did not include cognition levels, physical function status (e.g., walking difficultly, chair rise ability), body mass index, or biochemical measures, such as osteoporosis diagnosis (as a proxy for lower bone mineral density), to be used in data analyses to minimize the potential confounding effect by other hip fracture risks.

In summary, the results of our cohort study demonstrate an increased hip fracture risk among zolpidem users. The fracture episodes occur mainly during the medication days. The risk declines to the level of nonusers in the medication washout period. This finding suggests that great caution should be exercised when taking zolpidem to prevent falls or other accidents.

DISCLOSURE STATEMENT

This study was supported by the Executive Yuan Department of Health, Food, and Drug Administration (Grant number DOH 100-FDA-61403), the China Medical University Hospital (Grant number 1MS1), and the Taiwan Department of Health (Grant numbers DOH101-TD-B-111-004 and DOH101-TD-C-005). The authors have indicated no financial conflicts of interest.