Responsiveness of Veterans Affairs Health Care System to Zolpidem Safety Warnings
Sedative hypnotic medications are routinely prescribed for insomnia treatment, but have been associated with significant risks of morning-after impairment. We evaluated responsiveness in the Veterans Health Administration (VHA) facilities to two drug safety warnings recommending against high-dose zolpidem use—a 2007 Veterans Administration Pharmacy Benefits Management Service warning and a 2013 Food and Drug Administration (FDA) warning.
We used interrupted time-series design to assess how the two warnings influenced prescribing within the VHA in outpatients from 2005 to 2014. We assessed two outcomes: monthly outpatient use of (1) higher-than-recommended dose of zolpidem among zolpidem users and (2) any-dose zolpidem among all VHA users. In sensitivity analyses, we compared zolpidem prescribing to prescribing other sleep medications not subject to safety warnings.
After the 2007 VHA warning, high-dose zolpidem use decreased significantly among both sexes from approximately 10% to 2%. Following the 2013 FDA warning, high-dose zolpidem use declined again; however, approximately half of women Veterans remained on high doses. Overall zolpidem use nearly quadrupled between the 2007 VHA and 2013 FDA warnings, but the overall use declined after the 2013 FDA warning. Increase in sedating antidepressant use was seen after the FDA warning, suggesting potential substitution.
Higher than recommended dose use within the VHA decreased after each zolpidem high dose warning. Although overall use also decreased after the FDA warning, almost 50% of high-dose use among women Veterans is concerning. Different strategies to communicate the warnings should be examined.
A commentary on this article appears in this issue on page 1093.
Kim HM, Gerlach LB, Yosef M, Stano C, Conroy DA, Valenstein M, Pfeiffer PN, Sales AE, Zivin K. Responsiveness of Veterans Affairs Health Care System to zolpidem safety warnings. J Clin Sleep Med. 2018;14(7):1135–1141.
Current Knowledge/Study Rationale: Sedative hypnotic medications are routinely prescribed for insomnia treatment, but have been associated with morning-after impairment risks. We evaluated responsiveness to two safety warnings recommending against high-dose zolpidem use in Veterans Health Administration facilities—a 2007 Veterans Administration warning and a 2013 Food and Drug Administration (FDA) warning.
Study Impact: Higher-than-recommended dose use decreased after each warning. Almost 50% of females, however, were using a high dose in 2014. In contrast, overall zolpidem use increased more than fourfold between 2007 and 2013, but decreased after the 2013 FDA warning in both sexes. Policies were effective in decreasing high dose use. Different strategies to communicate the warnings and understanding of patients, particularly women, that use a high dose should be examined.
Sleep disorders are common, as approximately 30% of the general population experiences insomnia.1 Despite evidence suggesting the efficacy of nonpharmacologic interventions such as healthy sleep behaviors and cognitive behavioral therapy for insomnia (CBT-I),2,3 sleep disorders are routinely treated with medication.4 During 2005–2010, roughly 4% of United States adults have been prescribed sleep medication within the past 30 days, with greater use among older adults (7% of those older than 80 years) and women (5%).5 Nonbenzodiazepine receptor agonists (NBRAs) are among the most commonly prescribed sedative hypnotics for insomnia treatment.6
Although NBRAs improve subjective quality of sleep, they only increase total sleep time by 25 minutes.7,8 However, potential adverse reactions to sedative hypnotic medications include risk of falls, fractures, impaired cognition, and suicidal ideation.9–12 NBRA use may lead to increased hospitalizations for major injury13 and motor vehicle accidents14 due to morning-after functional impairment.7 Morning-after impairment refers to impairment in driving and activities that require alertness the morning after use, including drowsiness or impairment of mental alertness, even if they feel fully awake.15 Despite these potential harms, NBRA uses increased throughout the United States from 1999–2010.16
In light of growing evidence of substantial risk associated with zolpidem use and lack of additional hypnotic efficacy at higher doses, the Department of Veterans Affairs Pharmacy Benefits Management (VA PBM) Services released a bulletin in May 2007 recommending that zolpidem be used at doses no greater than 10 mg per day.17–21 In August 2007, zolpidem was placed on the Veterans Health Administration (VHA) prescription formulary. In January 2013, the Food and Drug Administration (FDA) released a drug safety warning with zolpidem products regarding the risk of next-morning impairment and recommended to lower doses.22–26 For women, the FDA recommended the dose lowered to 5 mg for immediate-release products (Ambien and Edluar) and to 6.25 mg for extended-release products (Ambien CR). For men, the FDA recommended that health care professionals consider prescribing the lower doses. In May 2013, the FDA approved label changes for the new dosing recommendations to 5 mg for immediate-release zolpidem products and 6.25 mg for extended-release zolpidem products for women, and for men, either 5 mg or 10 mg for immediate-release zolpidem products and 6.25 mg or 12.5 mg for extended-release zolpidem products.25
We used the VHA pharmacy data to study how zolpidem use changed following the 2007 VHA zolpidem warning and 2013 FDA zolpidem warning. We expected high-dose use to decrease in response to each warning.
Patient Population and Outcomes
We linked the prescription data from the VA PBM with the VHA Corporate Data Warehouse patient data and used interrupted time-series design to examine changes in zolpidem use in the VHA from December 2005 to November 2014. The VHA is the largest integrated health care system in the United States,26 providing care at more than 140 medical centers. We calculated two outcomes every month at each facility (medical center) separately by sex: (1) the percentage of patients with at least one higher-than-recommended dose zolpidem fill among zolpidem users in the respective month at each facility and (2) the percentage of patients with any-dose zolpidem fill among all VHA users at each facility. Outcomes were percentages rather than counts of patients to account for increasing enrollment over the study years. Each month, numbers of all VHA users at each facility were calculated by adding all unique persons with any outpatient visit, inpatient stay, or prescription fill during the respective month. Number of zolpidem users were identified by creating monthly rolling cohorts of zolpidem users who filled at least one prescription of zolpidem during each month at each facility without distinguishing prevalent versus new users. Daily dosing was calculated as the quantity of tablets divided by the days-supply, multiplied by the dosage per tablet. We defined a higher-than-recommended dose use as higher than 10 mg per day for immediate-release zolpidem use and higher than 12.5 mg per day for extended-release zolpidem use. Because the 2013 FDA recommended dose was sex specific, after January 2013, we redefined higher-than-recommended dose for women using a lower threshold of 5 mg per day for immediate release or 6.25 mg per day for extended release.
We adjusted for potential variation in high-dose use and any dose use by facility region (Northeast, Upper Midwest, West, or South) and rurality (rural versus urban).
We used five analytic periods (Figure 1) to assess the effect of the two zolpidem safety warnings on higher-than-recommended dose zolpidem use (high-dose use) over the study period. We fit separate linear mixed-effects regression models by sex. Each model included facilities as random intercepts to account for facility-specific high-dose use levels and autoregressive correlation of facility monthly data. We estimated the level of high-dose zolpidem use and the monthly slope (rate of change) during each postwarning period, controlling for baseline trends and trends following prior warnings. We also evaluated trends in any-dose zolpidem use using a linear mixed-effects model. For any-dose zolpidem use, we fit a single model using data of both sexes, included sex as a predictor and tested for the change in the level of any-dose zolpidem use from before to after the 2013 FDA warning by sex. Because any-dose use pattern showed nonlinear trends during the study period, we included time in month as both linear and squared terms and the interaction terms of sex by month and month-squared.
For both high dose and any dose use, we explored if facility region and rurality of the facilities predicted facility-level variation in these outcomes. We also conducted sensitivity analyses assessing use of potential sleep medication substitutes that were not subject to the drug safety warnings to look for compensatory prescribing changes. We divided comparison drugs into the following medication classes: (1) sedating antidepressants (doxepin, trazodone, amitriptyline, nortriptyline, mirtazapine), (2) antipsychotics (quetiapine), (3) benzodiazepines (alprazolam, clonazepam, diazepam, lorazepam, temazepam) and (4) eszopiclone [Lunesta]. Although eszopiclone, an alternative sedative hypnotic agent, was not widely used in the VHA, we included it because it also had an FDA safety warning for next-day impairment recommending a lower dose in May 2014.27
Monthly use data from 142 facilities were included. During the study period, mean monthly facility zolpidem users increased from 120 to 793 men and 14 to 96 women from 2005 to 2014, respectively. Similarly, mean monthly facility patient population increased from 23,035 to 26,879 men and 1,396 to 2,330 women. At the time of their first zolpidem use, patients had a mean age of 55.9 (standard deviation = 16.2), 89.4% men, 78% white, and 6.6% Hispanic or Latino (n = 693,056). In each year, 16% to 18% of the facilities were in rural areas.
Higher Than Recommended Dose Zolpidem Use
Figure 1 shows trends in high-dose use by sex. Prior to the warnings, 9.8% of women and 10.9% of men were prescribed higher than recommended doses of zolpidem. Following the 2007 VHA warning, high-dose use declined significantly at a rate of 0.5% (P < .001) per month for women and 0.8% (P < .001, Table 1) per month for men. In both sexes, high-dose use continued to decline and remained at approximately 2% by the 2013 FDA warning. In women, due to the newly defined lower recommended dose, after the 2013 FDA warning, high-dose use was 68.0% (P < .001) higher than the level prior to the FDA warning, but it declined significantly at 3.3% (P < .001) per month until the May 2013 FDA label approval and continued to decline until the end of the study period at a rate of 0.5% (P < .001) per month. Despite the continued decrease after the FDA warning, 48.3% of female zolpidem users had high-dose zolpidem use in 2014. In contrast, high-dose use was already low (2%) in men prior to the FDA warning and continued to decline further until the end of the study period although the rate of decline was not statistically significant.
In women, we explored whether region and rurality of the facility predicted facility-level variation in high-dose use after the 2013 FDA warning. Accounting for the declining trend after the FDA warning, we found high-dose zolpidem use was 12.1% (P = .002) lower in rural than urban facilities, with no significant differences by region (results available upon request). For stability of the estimates, we excluded data from facilities with 5 or fewer zolpidem users in a given month, but the significant difference by rurality remained the same as results without the exclusion.
Any-Dose Zolpidem Use Over the Study Period
Figure 2 shows trends in any-dose zolpidem use. Between the 2007 VHA warning and 2013 FDA warning, any dose use increased in both sexes. In the month of the 2013 FDA warning, zolpidem use was 4.5% in women and was lower in men by 1.6% (P < .001, Table 2). Immediately following the FDA warning, zolpidem prescribing declined significantly, with a decline in level in the following month by 0.38% (P < .001) in women and by 0.17% (P < .001) in men. Following the FDA warning, zolpidem prescribing continued to decline further. However, at the end of the study period in 2014, the level of any-dose zolpidem use was still higher than the level before the 2007 VHA warning; overall zolpidem use in November 2014 was 2.6% for men and 3.7% in women. We also observed variation by region over the study period: the West and South of the United States had 0.68% (P < .001) and 0.44% (P = .02) higher use, respectively, than the Northeast region.
Sensitivity analyses of other potential substitutes for zolpidem showed differences in the level of use by sex for all comparison drug classes (Figure 3). Potential substitution, however, is shown only by sedating antidepressants as indicated by a slight increase in the use of them after the 2013 FDA warning when zolpidem use started to decline. We also looked for medication that zolpidem replaced during the period prior to the FDA warning when its use increased. We did not find a change in use suggesting potential substitution during the period, other than a slight decline in the use of benzodiazepines and quetiapine. Although a similar FDA warning was given for eszopiclone, its use level was low (< 0.1%) in VHA users during our study period. Because of low prevalence, it is not visible in Figure 3, but eszopiclone use continued to increase from 0.0003% in 2005 to 0.04% in 2014 in men and from 0.002% to 0.11% in women; no decline in its use was seen during our study period after the May 2014 eszopiclone warning.
This study used national VHA pharmacy data to examine responsiveness in prescribing patterns following two zolpidem safety warnings each recommending against use of high-dose zolpidem. High-dose use decreased immediately after both the 2007 VHA and 2013 FDA warnings. However, almost 50% of women Veterans remained on high-dose zolpidem as recently as 2014, which is concerning in light of significant harms associated with high dose. Given the different formulations of zolpidem and different recommendations by sex, it is possible that prescribers were confused about recommended zolpidem prescribing.
Lower prescribing of high-dose zolpidem in women in rural facilities was surprising, considering studies that report higher use of benzodiazepine,28 a potentially inappropriate medication, in rural-dwelling older adults than urban-dwelling older adults. In our data, although the magnitude of the difference between rural versus urban facilities was not nearly as large as the difference seen in women after the 2013 FDA warning, men in rural facilities also had lower use of high-dose zolpidem than in urban facilities.
Although the 2007 VHA warning focused on high-dose zolpidem use, it is noteworthy that the overall use of zolpidem increased nearly fourfold after the VHA warning in light of an abundance of studies as early as 2005 showing potential harm associated with sedative hypnotics.5 Low-dose zolpidem was not subject to warning, and the increased availability and ease of prescribing since zolpidem was placed on the VA National Formulary in 2007 could explain the increase in its use. The VA National Formulary is a listing of drugs that must be available for prescription at all VA facilities and cannot be made nonformulary by individual medical centers. Of note, the percentage of VHA users in whom posttraumatic stress disorder, depression, or substance use disorder was diagnosed that can potentially affect sleep increased continuously from 2001 to 2014,29 but it did not covary with trends in zolpidem use. The increase in use during this period is also consistent with the increasing trends in use in the general United States population.16 It remains unclear whether the observed increase in zolpidem prescribing was a substitution for other sleep medication. We did not find meaningful changes in potential substitution medication during the period before the 2013 FDA warning.
Our study showed that the 2013 FDA warnings had an effect on the intended direction of the VHA on the use of higher-than-recommended-dose zolpidem, although in women, the percentage of those being prescribed above the recommended dosage is nontrivial because of the lowered recommended dose. Additionally, any-dose zolpidem use also decreased after the FDA warnings. We observed a slight increase in the use of sedating antidepressants around that time, suggesting possible substitution of zolpidem by sedating antidepressants after the FDA warnings. In contrast, a recent study outside the VHA using the Optum Clinformatics research database from January 2011 to December 2013 showed a decrease in high-dose zolpidem, but the warning did not affect the already downward trajectory of high-dose use and had no effect on the number of total patients or new zolpidem users, suggesting that the warnings enhanced the use of lower dose forms.30
Recent declines in high- and any-dose zolpidem use seen in this study may be associated not only with drug safety warnings, but may have been influenced by increased evidence for and availability of nonpharmacological alternatives to sleep medications. Evidence-based nonpharmacologic treatments such as CBT-I show significant improvement across multiple sleep measures.31 Numerous professional organizations recommend nonpharmacologic treatment as first line for insomnia, with NBRAs reserved for short-term treatment in those who have failed behavioral interventions.3,32 In 2011, the VHA disseminated nationwide a training program for CBT-I, and by 2013, CBT-I was available31 to patients with insomnia.
This study has multiple strengths, including large national data with a long study period and an interrupted time-series design, a robust quasiexperimental design with strong internal validity.33 We also note some limitations. This study used monthly pharmacy fills or refills, which may not necessarily represent actual usage of the medication by the patients. Hence, our data represent an indirect measure of both VHA physician prescribing and patient medication usage. However, we do not expect the long-term trends in use assessed using pharmacy data to differ from long-term trends in patient medication use or provider prescribing. Medication dose accuracy is imperfect in administrative data, although we do not expect this to vary differentially over our study period. We did not consider duration or appropriateness of medication use. Other important reasons for dose changes can be individual history of response or adverse effects or the frequency of visits with the prescriber. We did not consider indication for zolpidem use because sleep-related disorders are known to be underdiagnosed34 and we cannot be certain of indication for use because our data are based on administrative data. Our results represent trends in overall use across all clinical indications. Last, we did not evaluate provider specialty, which could influence prescribing patterns within facility. Despite these limitations, our aim was to evaluate the overall responsiveness to zolpidem safety warnings, and utilizing data from a long period of time allowed stable estimates of overall trends and thus the estimation of changes in the pattern of prescribing.
Future research should examine in what ways the warning and associated guidance was communicated within the VHA facilities and if specific strategies are associated with greater effect. The VA PBM oversees not only the national drug plan for Veterans receiving services through the VA Health Care system, but also plays a key role in disseminating and monitoring safe and appropriate medication use.18,35 The VA Center for Medication Safety (VA MedSAFE) within the VA PBM employed several strategies to communicate the zolpidem drug safety warnings including publishing National PBM Bulletins to alert providers to the FDA warnings.26 Additionally, the VA MedSAFE tracks and evaluates the use of potentially high-risk medications through medication use evaluations (MUEs) to assess safe use of medications across facilities.36 The Medication Use Evaluation Tracker (MUET) is an optional web-based application that can provide VA prescribers with a secure list of patients who may benefit from reconsideration of their medication use or dose.35 In another part of the current study, we are also currently directly examining in what ways the drug safety warnings and associated guidance were communicated within VA facilities and if specific strategies (eg, use of MUETs, academic detailing for physicians, integration of clinical pharmacy services) are associated with greater effect in reducing potentially inappropriate prescribing. It would also be important to investigate reasons for the high percentage of women at a higher than recommended dose after 2013 FDA warning. Finally, large studies demonstrating the effect of high-dose use on negative health consequences such as falls or motor accidents could re-enforce the importance of adhering to dose recommendations.
All authors have seen and approved the manuscript. This study was funded by the Department of Veterans Affairs (VA IIR 14-324). The authors report no conflicts of interest.
cognitive behavioral therapy for insomnia
Veterans Health Administration
United States Food and Drug Administration
nonbenzodiazepine receptor agonists
Department of Veterans Affairs Pharmacy Benefits Management Services
5 Prescription sleep aid use among adults: United States, 2005-2010. NCHS Data Brief; 2013127:1-8.
14 Sedative hypnotic medication use and the risk of motor vehicle crash. Am J Public Health; 2015;1058:e64-e69.
15 Questions and Answers: Risk of next-morning impairment after use of insomnia drugs; FDA requires lower recommended doses for certain drugs containing zolpidem (Ambien, Ambien CR, Edluar, and Zolpimist).
Accessed September 17, 2017https://www.fda.gov/drugs/drugsafety/ucm334041.htm. Updated December 23, 2014.
16 Trends in outpatient visits for insomnia, sleep apnea, and prescriptions for sleep medications among US adults: findings from the National Ambulatory Medical Care survey 1999-2010. Sleep; 2014;378:1283-1293, 25083008.
17 Department of Veterans AffairsNational PBM Drug Safety Alert Distribution. Washington DC: Department of Veterans Affairs; 2008.
20 Appendix 1: Review of Various Trials using zolpidem at Higher Doses.
Accessed December 15, 2016http://www.pbm.va.gov/PBM/vacenterformedicationsafety/Appendix1Detailsofhighdosezolpidemtrials.pdf. Published May 2007.
21 Department of Veterans AffairsDepartment of Veterans Affairs Pharmacy Benefits Management Service and Medical Advisory Panel Guidance for Treatment of Insomnia in Veterans in the Primary Care Setting November 2007. Washington DC: Department of Veterans Affairs; 2007.
23 Zolpidem and driving impairment - identifying persons at risk. N Engl J Med; 2013;2698:689-691.
24 U.S. Food & Drug AdministrationRisk of next morning impairment after use of insomnia drugs; FDA requires lower recommended doses for certain drugs containing zolpidem (Ambien, Ambien CR, Edluar, and Zolpimist).
Accessed September 17, 2017https://www.fda.gov/drugs/drugsafety/ucm334041.htm.
25 U.S. Food & Drug AdministrationFDA Drug Safety Communication: FDA approves new label changes and dosing for zolpidem products and a recommendation to avoid driving the day after using Ambien CR.; 2013
Accessed September 17, 2017http://www.fda.gov/Drugs/DrugSafety/ucm352085.htm.
26 Department of Veterans Affairs Veterans Health Administration (VHA) Pharmacy Benefits Management (PBM) & Medical Advisory Panel (MAP) VA Center for Medication Safety (VAMedSAFE) National PBM Communication, January 16, 2013.
Accessed September 30, 2016http://www.pbm.va.gov/vacenterformedicationsafety/nationalpbmcommunication/ZolpidemandProposedLowerDosesDuetoImpairedMentalAlertnessNATIO.pdf.
27 Department of Veterans Affairs Pharmacy Benefits Management Services (PBM), Medical Advisory Panel (MAP), and Center for Medication Safety (VA Medsafe). Eszopiclone (Lunesta): Lowered Dose Recommendations due to Next-day Impairment; May, 2014.
Accessed September 30, 2016http://www.pbm.va.gov/PBM/vacenterformedicationsafety/nationalpbmbulletin/Eszopiclone_Lunesta_Next_Day_Impairment_and_Lowered_Doses_NATIONAL_PBM_Bulletin.pdf.
29 Suicide Among Veterans and Other Americans 2001-2014Office of Suicide Prevention.; 201683
Accessed January 16, 2018https://www.mentalhealth.va.gov/docs/2016suicidedatareport.pdf.
33 A critical review of methods to evaluate the impact of FDA regulatory actions. Pharmacoedidemiol Drug Saf; 2013;229:986-994.
34 Oxford Textbook of Sleep DisordersOxford, England: Oxford University Press; 2017.