Adherence to behavioral recommendations of cognitive behavioral therapy for insomnia predicts medication use after a structured medication taper

Abstract

Study Objectives:

Cognitive behavioral therapy for insomnia (CBTI) has been paired with supervised medication tapering to help hypnotic-dependent individuals discontinue their hypnotics. This study examined the hypothesis that higher participant adherence to behavioral recommendations of CBTI will predict lower odds of using sleep medications 3 months after completion of a combined CBTI/sleep medication tapering protocol.

Methods:

Fifty-eight individuals who used sedative hypnotics completed four CBTI sessions followed by sleep medication tapering. Logistic regression was used to examine the association of stability of time in bed and stability of rise time (measured as the within-person standard deviation) at completion of CBTI with two outcomes at 3-month follow-up: use of sedative hypnotics and use of any medication/substance for sleep.

Results:

Participants with more stability in their rise time after CBTI than at baseline (ie, a decrease in their within-person standard deviation) had 69.5% lower odds of using sedative hypnotics at follow-up (odds ratio = 0.305, 95% confidence interval = 0.095–0.979, P = .046) than individuals who had no change or a decrease in the stability of their rise time. Results were similar for time in bed: participants with more stability in their time in bed after CBTI than at baseline had 83.2% lower odds of using sedative hypnotics (odds ratio = 0.168, 95% confidence interval = 0.049–0.580, P = .005). Increase in stability of rise time and stability of time in bed was also associated with reduced odds of using any medication/substance for sleep at follow-up.

Conclusions:

Participants who implement behavioral recommendations of CBTI appear to have more success with discontinuing use of sleep medications.

Clinical Trial Registration: Registry: ClinicalTrials.gov; Name: The Role of Tapering Pace and Selected Traits on Hypnotic Discontinuation; URL: https://clinicaltrials.gov/ct2/show/NCT02831894; Identifier: NCT02831894.

Citation:

Edinger JD, Wamboldt FS, Johnson RL, et al. Adherence to behavioral recommendations of cognitive behavioral therapy for insomnia predicts medication use after a structured medication taper. J Clin Sleep Med. 2023;19(8):1495–1503.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Protocols that combine CBTI with sleep medication tapering are based on the assumption that participants will utilize CBTI skills to manage insomnia rather than utilizing medications to manage insomnia. This study examined whether higher participant use of specific CBTI skills (ie, adherence to behavioral recommendations with regard to stability of rise time and stability of time in bed) will predict lower odds of using sleep medications three months after completion of a combined CBTI/sleep medication tapering protocol.

Study Impact: Findings of this study provide initial information about the role of CBTI in interventions designed to help people discontinue use of medications for sleep. Participants who implement the behavioral recommendations of CBTI appear to have more success with discontinuing use of medications for sleep.

INTRODUCTION

Chronic insomnia is a prevalent condition that reduces quality of life, increases health care costs, leads to daytime impairment, and elevates risks for serious psychiatric and cardiometabolic illnesses for millions worldwide.^1–14 Effective insomnia treatments include pharmacotherapy with sleep-promoting agents that are commonly called hypnotics, as well as cognitive behavioral therapy for insomnia (CBTI). Most individuals who seek treatment for insomnia seek help via primary care, where a hypnotic prescription is the first—and usually the only—treatment offered for insomnia.^15,16 Multiple concerns have been raised about prescription hypnotics, especially the benzodiazepines (BZDs) and benzodiazepine receptor agonists (BZRAs). Some of these medications, especially classic BZDs, have been associated with tolerance and declining efficacy over time, physical and psychological dependence, loss of self-efficacy with regard to sleep, residual daytime sedation and cognitive impairment, and rebound insomnia upon abrupt withdrawal.^17–22 Many individuals who use medications to treat their insomnia would prefer to discontinue such use.²³

An evidence-based method to help people discontinue use of hypnotics for insomnia is CBTI in conjunction with supervised medication tapering (SMT).^24–26 CBTI helps patients develop cognitive and behavioral skills to manage/reduce their sleep difficulties. Protocols that combine CBTI with SMT are based on the assumption that participants will utilize CBTI skills to manage insomnia rather than utilizing medications to manage insomnia. However, research has not examined whether participant use of CBTI skills is associated with success in discontinuing medications after a combined CBTI/SMT protocol.

Two key behavioral skills that are emphasized in CBTI are to have a stable rise time every day and to have a stable amount of time in bed each night. Research on adherence to CBTI has focused on both stability of rise time^27–29 and stability of time in bed (TIB),^27–29 and a recent systematic review of adherence to CBTI includes the recommendation that researchers utilize variance in rise time and TIB when reporting adherence to CBTI.³⁰ The current study examined whether stability of rise time and stability of TIB were associated with medication use outcomes after SMT. This study involves secondary analysis of data collected during an intervention that involved CBTI followed by SMT with the goal of discontinuing use of BZDs and BZRAs for insomnia.³¹ The current analyses examine the hypothesis that greater stability of rise time and TIB will be associated with better outcomes (ie, not using BZD-related hypnotics or any other medication or substance for sleep) following participation in a combined CBTI/SMT protocol. Due to the design of the intervention study, the current analyses focus exclusively on outcomes at 3-month follow-up. Outcomes immediately after completion of SMT are not examined in the current study because the intervention protocol included a blinded control group that was maintained on a consistent dose of sleep medications throughout the SMT. All individuals in the control group who followed the intervention protocol were still on their medications at the end of SMT.

METHODS

Study participants

The current study utilizes data collected during an intervention that recruited adults who were both long-term (≥ 12 consecutive months) and frequent (≥ 5 nights per week) BZD or BZRA hypnotic users interested in discontinuing their hypnotic use. The study was approved by the National Jewish Health Institutional Review Board (Protocol HS-2891). To be included in the study participants had to (1) be using one or more BZD or BZRA hypnotics at bedtime for insomnia, (2) have been using one or more such agents at least 5 nights per week for at least the past 12 months, (3) express interest in discontinuing hypnotic use and learning to manage insomnia without medications, (4) report at least one failed attempt to discontinue hypnotic use in the past, (5) score ≥ 10 on the Insomnia Severity Index and have a score of 2 or more on either the interference or distress items, and (6) provide written consent.

Exclusion criteria were designed to permit enrollment of a broadly representative sample yet eliminate subgroups that would introduce extensive error variance into analyses. Those excluded had (1) an untreated, unstable, or “in-treatment” psychiatric disorder (eg, major depression in psychotherapy or on a medication regimen that had been changed within the past 2 months); (2) a lifetime diagnosis of any psychotic or bipolar disorder; (3) an imminent risk for suicide; (4) evidence of alcohol or drug abuse (other than hypnotics) within the past year; (5) an unstable or terminal physical illness (eg, cancer), neurological degenerative disease (eg, dementia), or sleep disruptive medical condition (eg chronic pain); (6) current use of medications known to cause insomnia (eg, corticosteroids); (7) a history or screening evidence of restless legs syndrome, circadian rhythm sleep disorder (eg, delayed sleep phase syndrome), sleep apnea (apnea-hypopnea index > 5), or periodic limb movement disorder (Periodic Limb Movements Index > 15); (8) habitual bedtimes later than 2:00 am or rising times later than 10:00 am; and (9) consumption of > 2 alcoholic beverages per day at least 5 times per week. We also excluded pregnant women and mothers with care-taking responsibilities for infants due to the sleep disruption caused by these circumstances.

Study design

Overview

The intervention employed a double-blinded design involving three tapering conditions. Participants (n = 78) completed baseline measures and then underwent four CBTI sessions scheduled across a 6-week time span. After the four CBTI sessions, participants completed additional assessment measures and were randomly assigned to one of three, double-blind, 20-week hypnotic tapering protocols. One-third of the sample was assigned to a no-taper control group in which their baseline hypnotic dosage remained unchanged; a second third of the sample was assigned to a protocol in which their hypnotic dosage was reduced by 10% every 2 weeks, and the final third was assigned to a faster tapering protocol in which they underwent a 25% reduction in their medication dosage every 2 weeks. All three tapering schedules were implemented in a double-blind fashion; only the study statistician and the pharmacist who prepared the medications were aware of the treatment assignment. Individuals who were randomized to the no-taper control group had the opportunity to participate in an open-label taper after completion of the blinded taper. Medication usage was assessed at the end of the 20-week SMT and at 3 months following completion of SMT, regardless of tapering pace (in order to maintain blinding). For individuals who chose to start an open-label taper (22 of the 26 participants in the no-taper control group chose to do so), this assessment was scheduled 3 months after the end of the open-label taper. All individuals who chose to stop participating in the intervention (ie, CBTI and taper) had the opportunity to continue providing outcome data. Edinger et al³¹ provide more detailed information about the intervention.

CBTI treatment

All participants were provided four sessions of CBTI prior to the hypnotic tapering phase of the study. CBTI was delivered via one-on-one sessions with a licensed clinical psychologist (J.D.E.) who is certified in behavioral sleep medicine and who had 36 years of experience in delivering insomnia therapies at the time of study initiation. CBTI sessions occurred at weeks 1, 2, 4, and 6 of the treatment phase. The CBTI therapist was kept blind to the tapering group assignment of each participant in order to control for experimenter bias. Treatment was guided by the first author’s published treatment manuals.^32–34 Session 1 focused on treatment rationale, discussion of normal sleep, discussion of putative insomnia perpetuating mechanisms, and presentation of combined stimulus control and sleep restriction instructions. Session 2 focused on use of constructive worry exercises to address sleep-disruptive worrying in bed. Session 3 focused on cognitive restructuring via use of thought records and behavioral experiments to address unhelpful beliefs about sleep. Session 4 reviewed all of the treatment strategies and included sleep hygiene education concerning lifestyle (eg, use of caffeine, alcohol, and tobacco, as well as dietary and exercise practices) and environmental factors (light, noise, temperature) that affect sleep, as well as discussion of maintaining treatment gains and avoiding relapse.

The guidelines outlined in the first author’s published treatment manual³⁴ were followed when administering sleep restriction. This approach sets time in bed at average total sleep time plus 30 minutes. TIB is adjusted up in 15-minute increments when sleep efficiency is ≥ 85% and the patient reports daytime sleepiness/fatigue. TIB is adjusted down in 15-minute increments when sleep efficiency remains < 85%. TIB is kept constant when sleep efficiency is ≥ 85% and the patient does not have daytime fatigue/sleepiness and is satisfied with her/his current sleep pattern.

SMT protocol

After completing CBTI, participants began their randomly assigned 20-week blinded medication taper protocol. During SMT, participants received all of their medications from our medical center’s pharmacy. Participants assigned to the 25% tapering protocol received 100% of their baseline dose during the first 2 weeks of SMT and then the weight of active medication placed in their study capsules was reduced by 25% during weeks 3 and 4, and so on, while a dextrose filler was accordingly increased. A similar procedure was used to accomplish the 10% biweekly reductions for our 10% tapering condition. All hypnotic medications were provided to study participants by pharmacy personnel, and both the study investigators and the participants were kept blind to treatment condition. Participants were provided a new supply of medications every 2 weeks containing the nightly dosage required by their respective tapering protocol for each 2-week phase of the 20-week tapering period. After completion of the blinded taper, individuals who had been randomized to the no-taper control group had the opportunity to participate in an open-label taper. These individuals were able to choose their preferred tapering pace: 10% or 25%.

During tapering, one of the study physicians (F.S.W.) met with each participant for scheduled visits every 2 weeks. More frequent in-person or telephone visits were made available to participants upon request if they had concerns that required physician attention. Since this physician was kept blind to each participant’s assigned protocol, there was no risk that he would convey information that would unblind any participant. The goal of these visits was to address any distressing symptoms or safety concerns that arose during study participation. During these visits, the study physician maintained a supportive attitude but did not provide any formal advice about sleep management techniques other than query what the person had learned to do during CBTI.

Measures

A baseline assessment was conducted prior to the intervention. The baseline assessment included questions regarding demographic information, history of insomnia and of sleep medication use, and current use of medications for sleep. The Insomnia Severity Index was completed to characterize the severity of insomnia at baseline. The Insomnia Severity Index is a 7-item questionnaire that assesses nighttime and daytime insomnia symptoms and has well-established reliability and validity.³⁵

An online sleep diary system was used to collect self-reported estimates of sleep and wake times. This web-based diary was designed to mimic the Consensus Sleep Diary³⁶ and solicit information about the participant’s bedtime, sleep onset latency, number and length of nocturnal awakenings, time of final waking, rising time, sleep quality ratings, and sleep medication use. Participants were instructed to complete sleep diaries each morning upon arising for 14 days at baseline (prior to starting CBTI) and for 14 days following completion of CBTI (prior to starting the SMT). Rise time was measured by a single question in the diary, and TIB was calculated per the standard definition published by Buysse et al.³⁷ For both indicators of adherence (ie, rise time and TIB), adherence was operationalized as the within-person standard deviation. In order to ensure that enough data were available to adequately measure³⁸ rise time and TIB, the within-person standard deviation was calculated for participants who had sleep diary data for at least 7 days at baseline and at least 7 days at post-CBTI. The within-person standard deviation was calculated for each indicator of adherence at baseline and for each indicator of adherence at post-CBTI. A within-person standard deviation of 0 would indicate perfect stability from night to night, with higher values indicating less stability.

Medication use data at the 3-month follow-up assessment were obtained from self-report during interviews and from the medication tracking information provided via the sleep diaries. Two medication use outcomes were examined in the current analyses: (1) use of BZD-related hypnotics at 3-month follow-up and (2) use of any medication or substance for sleep at 3-month follow-up. The use of any medication or substance for sleep is a composite outcome that reflects whether each participant used any of the following categories of medications/substances for sleep at 3-month follow-up: BZD-related hypnotics, non-BZD-related prescription sleep medications (eg, amitriptyline, gabapentin, trazodone), over-the-counter medications for sleep (eg, Tylenol PM, Benadryl), and substances to promote sleep (ie, alcohol, cannabidiol, tetrahydrocannabidinol).

Statistical analyses

Data were analyzed using SPSS Statistics Version 28. Significance tests were two-sided with a significance level of .05. For the current analyses, participants were included if they completed at least 7 days of sleep diaries at both baseline and post-CBTI and also provided outcome data at 3-month follow-up. Due to missing data, the current analyses included 58 individuals (see Figure 1 for participant flow diagram).

Figure 1. Participant flow diagram.

Preliminary analyses to examine characteristics of participants

Characteristics of participants were summarized using mean and standard deviation for continuous variables and number and percentage for categorical variables.

Analyses to examine the effect of adherence on medication use outcomes

For both indicators of adherence (ie, rise time and TIB), we examined three methods of conceptualizing adherence. Method 1 is the within-person standard deviation at post-CBTI. Method 2 is the change in within-person standard deviation from baseline to post-CBTI, measured as a continuous value. Method 3 is a categorical value that captures whether each participant’s within-person standard deviation decreased between baseline and post-CBTI (ie, whether stability increased during this timeframe). The use of continuous and categorical methods of measuring change allowed us to examine consistency of findings across different methods of measuring change.

Three logistic regression models were calculated to examine the association of stability of rise time with use of BZD-related hypnotics at 3-month follow-up: one model in which stability of rise time was measured via Method 1, one model in which it was measured via Method 2, and one model in which it was measured via Method 3. Three logistic regression models were calculated to examine the association of stability of rise time with use of any medication or substance for sleep at 3-month follow-up—using Method 1, Method 2, and Method 3 as described previously. For TIB, these analyses were repeated using stability of TIB instead of stability of rise time as the predictor. Twelve logistic regression models were calculated in total. In all models, outcomes were coded as 0 = no use and 1 = yes use (of BZD-related hypnotics or use of any medication/substance for sleep).

All logistic regression models included as a covariate the study group to which the participant had been assigned in the intervention study. Study group was coded as blinded taper and open-label taper (rather than as no-taper control, 10% reduction rate, and 25% reduction rate) because blinded status—but not tapering rate—was consistently associated with outcomes in the intervention study.³¹ All analyses used outcomes measured at 3-month follow up. We did not examine outcomes immediately post-taper due to the design of the intervention. In the intervention study, all individuals who were assigned to either the 10% taper group or the 25% taper group who followed the protocol were not using medications at the end of taper. Thus, outcomes at the end of SMT were largely determined by treatment assignment, and the number of treatment failures at the end of the SMT was extremely small.

Analyses to examine missing data

Individuals who were removed from the sample due to missing data were compared to individuals who were included in analyses with regard to baseline characteristics. Continuous variables that were approximately normally distributed were examined via t tests. Continuous variables that were not normally distributed were examined via the Mann-Whitney U test. Categorical variables were examined via chi-square tests. These analyses provide information regarding bias that may be introduced due to missing data.

In addition, sensitivity analyses were conducted. These analyses included participants who had been excluded due to missing outcome data. In these analyses, missing outcome data were assumed to be treatment failures (ie, assumed to using BZD-related hypnotics and assumed to be using any medication or substance for sleep).

RESULTS

Characteristics of participants

Baseline characteristics of the sample are in Table 1. The mean age was 56.40 (SD = 12.36). The majority of participants were female (63.79%) and Caucasian (89.47%). The mean duration of insomnia was 14.57 years (SD = 13.05), and the mean duration of sleep medication use was 10.44 years (SD = 7.93). On average, participants used a BZD-related hypnotic for sleep 6.84 nights per week at baseline (SD = 0.52), and the mean dose in diazepam equivalents was 56.38 mg per week (SD = 40.39). At baseline, the mean value in our sample for the within-person SD for rise time was 50.81 minutes. The mean value in our sample at baseline for the within-person SD for TIB was 57.23 minutes.

Table 1.

Characteristics of the sample at baseline (n = 58).

Demographic and Health Characteristics at Baseline
Age in years	56.40 (12.36)
Sex
Male	21 (36.21%)
Female	37 (63.79%)
Race/ethnicity
Caucasian non-Hispanic	51 (89.47%)
Minority	6 (10.53%)
Education (number of years)	16.36 (2.47)
Duration of insomnia in years	14.57 (13.05)
Insomnia severity (Insomnia Severity Index)	15.52 (5.26)
Duration of sleep medication use in years	10.44 (7.93)
BZD-related hypnotic frequency of use (nights per week)	6.84 (0.52)
Diazepam equivalents in mg (per week)	56.38 (40.39)
Rise Time and Time in Bed at Baseline
Rise time: within-person SD at baseline (in minutes)	50.81 (25.48)
Time in bed: within-person SD at baseline (in minutes)	57.23 (29.14)

Values are mean (standard deviation) or n (%). BZD = benzodiazepine, SD = standard deviation.

Table 2 provides information about our measures of participant adherence to behavioral recommendations of CBTI. On average, the within-person SD for stability of rise time decreased by 1.21 minutes between baseline and post-CBTI. More than half (58.62%) of the sample had a decrease in their within-person SD for rise time between baseline and post-CBTI. Stated another way, 58.62% of the sample had an increase in the stability of their rise time. Among this 58.62% of the sample, the within-person SD for rise time decreased by 19.41 minutes on average. For TIB, 62.07% of the sample had an increase in the stability of their TIB between baseline and post-CBTI. Among this 62.07% of the sample, the within-person SD for TIB decreased by 21.29 minutes on average.

Table 2.

Indicators of participant adherence to behavioral recommendations of CBTI (n = 58).

Rise Time
Within-person SD at post-CBTI (in minutes)	49.60 (41.94)
Change from baseline to post-CBTI (continuous measure, in minutes)*	−1.21 (37.16)
Change from baseline to post-CBTI (dichotomous measure)
Within-person SD did not decrease between baseline and post-CBTI	24 (41.38%)
Within-person SD decreased between baseline and post-CBTI	34 (58.62%)
Change from baseline to post-CBTI, for each group
Within-person SD did not decrease between baseline and post-CBTI	24.57 (44.14)
Within-person SD decreased between baseline and post-CBTI	−19. 41 (14.26)
Time in Bed
Within-person SD at post-CBTI (in minutes)	52.38 (27.97)
Change from baseline to post-CBTI (continuous measure; in minutes)*	−4.85 (29.63)
Change from baseline to post-CBTI (dichotomous measure)
Within-person SD did not decrease between baseline and post-CBTI	22 (37.93%)
Within-person SD decreased between baseline and post-CBTI	36 (62.07%)
Change from baseline to post-CBTI, for each group
Within-person SD did not decrease between baseline and post-CBTI	22.04 (19.92)
Within-person SD decreased between baseline and post-CBTI	−21.29 (21.42)

Values are mean (standard deviation) or n (%).

^*A negative value indicates that the mean value in the sample for the within-person standard deviation decreased at post-CBTI compared to baseline.

CBTI = cognitive behavioral insomnia therapy, SD = standard deviation.

Effect of adherence on medication use outcomes

Medication use outcomes are shown in Table 3. Slightly more than one-third of the sample (36.21%) was using BZD-related hypnotics at 3-month follow-up. More than two-thirds of the sample (70.69%) was using some type of medication or substance for sleep at 3-month follow-up. Detailed information about the types of medications and substances used for sleep at 3-month follow-up other than BZD-related hypnotics is provided in Table S1 ^{(493.9KB, pdf)} in the supplemental material.

Table 3.

Medication use outcomes at 3-month follow-up (n = 58).

Medication Use Outcomes at 3-Month Follow-Up	n (%)
Use of BZD-related hypnotics
No	37 (63.79)
Yes	21 (36.21)
Use of any medication or substance for sleep
No	17 (29.31)
Yes	41 (70.69)

BZD = benzodiazepine.

Results of logistic regression models are in Table 4. When using Method 1 to measure adherence (ie, the predictor is measured as the within-person SD at post-CBTI), none of the models show a statistically significant association between adherence and outcomes (all P values are > .05). When using Method 2 to measure adherence (ie, the predictor is a continuous measure of change from baseline to post-CBTI), three of the four models show a statistically significant association between adherence and outcomes. In the model in which TIB is a predictor of use of BZD-related hypnotics, the odds ratio = 1.033 (95% confidence interval =1.005–1.061, P = .019). This means that, for every 1-minute increase in the within-person SD for TIB between baseline and post-CBTI, the odds of using any medication or substance for sleep increase by 3.3%.

Table 4.

Logistic regression models for the association of adherence to behavioral recommendations of CBTI with medication use at 3-month follow-up^* (n = 58).

	Use of BZD-Related Hypnotics	Use of Any Medication or Substance for Sleep
Rise Time
Method 1: within-person SD post-CBTIa	1.001 (0.988–1.014), .887	1.012 (0.989–1.035), .322
Method 2: change BL to post-CBTI continuousb	1.000 (0.986–1.015), .966	1.053 (1.015–1.091), .005
Method 3: change BL to post-CBTI dichotomousc
Within-person SD did not decrease	Reference	Reference
Within-person SD decreased	0.305 (0.095–0.979), .046	0.130 (0.026–0.658), .014
Time in Bed
Method 1: within-person SD post-CBTIa	1.021 (1.000–1.043), .055	1.011 (0.987–1.035), .362
Method 2: change BL to post-CBTI continuousb	1.033 (1.005–1.061), .019	1.042 (1.013–1.072), .004
Method 3: change BL to post-CBTI dichotomousc
Within-person SD did not decrease	Reference	Reference
Within-person SD decreased	0.168 (0.049–0.580), .005	0.126 (0.025–0.644), .013

Values are odds ratio (95% confidence interval), P value.

^*All models adjust for treatment assignment in the intervention study (ie, open-label taper vs blinded taper); in all models, the outcomes are coded such that 0 = not using medication/substance and 1 = using medication/substance (ie, 1 = treatment failure)

^a For Method 1, an odds ratio greater than 1.000 indicates that a higher value for the within-person SD at post-CBTI (ie, more variability) is associated with higher odds of treatment failure at 3-month follow-up.

^b For Method 2, an odds ratio greater than 1.000 indicates that an increase in variability over time (from baseline to post-CBTI) is associated with higher odds of treatment failure at 3-month follow-up.

^c For Method 3, an odds ratio less than 1.000 indicates that participants who had a reduction in variability over time (from baseline to post-CBTI) had lower odds of treatment failure at 3-month follow-up than participants who had an increase or no change in variability.

BL = baseline, BZD = benzodiazepine, CBTI = cognitive behavioral therapy for insomnia, SD = standard deviation.

When using Method 3 to measure adherence (ie, a categorical measure of change from baseline to post-CBTI), all four models show a statistically significant association between adherence and outcomes. In the model in which TIB is a predictor of use of BZD-related hypnotics, the odds ratio = 0.168 (95% confidence interval = 0.049–0.580, P = .005). This means that participants who had a decrease in their within-person standard deviation for TIB between baseline and post-CBTI had 83.2% lower odds of using BZD-related hypnotics at 3-month follow-up than participants who had no change or an increase in their within-person standard deviation between baseline and post-CBTI.

Missing data analysis

Table S2 ^{(493.9KB, pdf)} provides the baseline characteristics of individuals who were excluded from the study due to missing data and results of analyses to compare these individuals to those who were included in the study. The two groups differed from each other with regard to 1 of the 11 characteristics: individuals who were excluded from the study had more severe insomnia (P = .036). The mean score for excluded individuals was 18.35 (SD = 4.72) as compared to 15.52 (SD = 5.26) for included individuals. In addition, individuals who were excluded from the study exhibited a trend toward less frequent use of BZD-related hypnotics at baseline (P = .098). On average, individuals who were excluded from the study used BZD-related hypnotics 6.60 nights per week (SD = 0.75) as compared to 6.84 nights per week (SD = 0.52) for individuals who were included in the study. Table S3 ^{(493.9KB, pdf)} contains the results of sensitivity analyses in which missing data were assumed to be treatment failures. None of the findings are substantively different in these models when compared to the results shown in Table 4. All P values that were < .05 in the original models were < .05 in the sensitivity analyses.

DISCUSSION

Our findings suggest that changes in sleep-related behaviors between baseline and the end of CBTI may influence rates of medication discontinuation following SMT. While stability of rise time and stability of TIB at the end of CBTI did not predict medication use outcomes, change in stability did predict outcomes—both with regard to use of BZD-related hypnotics and use of any medication or substance for sleep. Specifically, individuals who increased the stability of their rise time between baseline and the end of CBTI had lower odds of using BZD-related hypnotics 3 months after completing SMT, as compared to individuals with no change or a decrease in the stability of their rise time. Individuals who increased the stability of their rise time also had lower odds of using any medication or substance for sleep at this time point. An increase in the stability of TIB was also associated with lower odds of using BZD-related hypnotics and lower odds of using any medication or substance for sleep 3 months after completing SMT. These results suggest that changes in sleep behaviors may lead to greater success in discontinuing use of medications for sleep. However, an alternative explanation is that an underlying trait influences our predictors and our outcomes: people who adhere to CBTI also adhere to discontinuing use of their medications.

Change in sleep behavior was examined as both a continuous and a categorical variable. When comparing results for the continuous method of measuring change vs the categorical method, there were mixed findings with regard to whether change in rise time was associated with the use of BZD-related hypnotics. Change in rise time stability—measured as a continuous variable—was not associated with use of BZD-related hypnotics. When measured as a categorical variable, change in rise time stability was associated with use of BZD-related hypnotics. Further research is needed to more fully understand the association of change in rise time with subsequent use of BZD-related hypnotics.

Protocols that combine CBTI with SMT are an evidence-based method to help people discontinue using hypnotics for insomnia. These protocols are based on the assumption that participants will utilize CBTI skills rather than medications to manage insomnia. Findings of the current study suggest that participants who are adherent to the behavioral recommendations of CBTI do indeed appear to be more likely to be successful at achieving the goal of discontinuing medication use. The current study used information from sleep diaries to measure adherence to CBTI. This is consistent with existing research that has utilized measures of variability in various parameters captured in sleep diaries to measure adherence to behavioral components of CBTI.^{27–29,39–42} The use specifically of variability in rise time and variability in TIB has been employed in several studies^27–29 and is recommended by Agnew et al as a measure of adherence to CBTI.³⁰ Our values were similar to those reported by other authors. For example, our within-person standard deviation for TIB at baseline of 57.23 minutes is comparable the value of 63.99 minutes that was reported by Reidel and Lichstein.²⁹ Our within-person standard deviation for rise time at baseline was 50.81 minutes, as compared to 42.32 minutes as reported by Riedel and Lichstein.²⁹

With regard to change in these parameters after participating in CBTI, the mean within-person standard deviation for rise time decreased by 1.21 minutes in our study. This compares to a 16.0-minute decrease in the mean within-person standard deviation reported by Riedel and Lichstein²⁹ and a 3.5-minute decrease in the median within-person standard deviation reported by Ludwin et al.²⁸ For TIB, the mean decrease was 4.85 minutes in our study, compared to a mean decrease of 29.5 minutes in Riedel and Lichstein²⁹ and median decrease of 4.4 minutes in Ludwin et al.²⁸ Direct comparison of our study with these studies is hampered somewhat by differences between studies. Both of the comparison studies utilized more treatment sessions than the current study and included fewer than 25 participants. At present, there are no guidelines regarding what is considered a clinically meaningful change in stability of rise time or TIB.

The current study focused on stability of rise time and TIB. A future step in this line of research would be to examine the role of patient adherence to additional CBTI skills in interventions to help people discontinue use of sleep medications. It is likely that a broader range of cognitive and behavioral skills that are taught in CBTI (in addition to stability of rise time and TIB) contributes to success in discontinuing use of sleep medications. This line of research may identify a subset of CBTI skills that is most relevant to success in discontinuing use of sleep medications. Ultimately, such research would address the important clinical question of whether a simplified version of CBTI—that can more easily be delivered by physicians and that would focus on the CBTI skills that are most salient for medication discontinuation—could be used in protocols to help people discontinue use of medications for insomnia.

Individuals who stop using BZD-related hypnotics after participating in a combined CBTI/SMT protocol may later resume use of these medications. In one study, 19% of the combined CBTI/SMT group were using BZD hypnotics at 3-month follow-up, which increased to 29% at 12 months and 35% at 24 months.⁴³ In another study, 17% of the combined CBTI/SMT group were using benzodiazepines at 3-month follow-up, which increased to 30% at 12-month follow-up.²⁶ Researchers have posited that decreased use of CBTI skills over time is one potential reason for the observed increase in medication use.^26,43 In the current study, approximately one-third (36.21%) of our sample was using BZD-related hypnotics at 3-month follow-up and nearly twice as many participants (70.69%) were using some type of medication or substance for sleep at this time point. Thus, rather than discontinuing the use of medications/substances for sleep altogether, many individuals shifted from using BZD-related hypnotics to some other medication or substance for sleep. Further research is needed to examine use of CBTI skills over longer periods postintervention, as well as long-term patterns of medication use—including use of non-BZD-related medications or substances for sleep in addition to BZD-related hypnotics.

Limitations

When considering our findings, it is important to note several limitations of this study. First, these findings are based on secondary analyses of data from a study that was not designed to investigate behavioral adherence to CBTI as a predictor of medication use outcomes. As such, data are not available regarding sleep behavior at all relevant timepoints (eg, throughout CBTI, throughout the 20-week tapering period, at 3-month follow-up). Thus, we cannot confirm that CBTI recommendations were closely followed throughout the duration of CBTI or that participants who reported more stability in their sleep behaviors after CBTI maintained these gains throughout the tapering period or during longer-term follow-up. Second, we operationalized adherence to CBTI strictly with regard to stability of rise time and TIB. As such, our analyses do not encompass adherence to the full range of cognitive and behavioral skills that are taught in CBTI and may not capture participants’ overall adherence with CBTI. Third, one-quarter of the sample from the intervention study was excluded from the current analyses due to missing data. Individuals who were excluded due to missing data had more severe insomnia, which limits the generalizability of our findings. Fourth, the study sample was relatively small, and the number of treatment failures (n = 21 using BZD-related hypnotics and n = 41 using any medication or substance for sleep) precludes the inclusion of multiple covariates in our statistical models. As such, our models did not adjust for duration of insomnia, anxiety symptoms, or other patient characteristics that may influence success at discontinuing medication use. Fifth, all analyses are based on participant-reported data (rather than objective assessments). It would have been especially useful to have data from actigraphy to measure adherence to CBTI and data from urine drug screens for a more complete assessment of medication use. Sixth, the sample was comprised largely of Caucasians, so the generalizability of the study’s results to larger groups of ethnically diverse patients remains to be tested. Seventh, individuals who completed an open-label taper did so after completing a blinded taper in which they had been assigned to the no-taper control group. These individuals provided data regarding medication use outcomes 3 months after completing the open-label taper. As such, the amount of time elapsed between CBTI and 3-month follow-up was longer for these participants, which may influence the extent to which they were still practicing CBTI skills at follow-up. Finally, due to the inclusion criteria, the sample was comprised entirely of individuals who expressed an interest in discontinuing hypnotic use and learning to manage their insomnia without medications. Hence, further research that addresses these limitations is needed.

CONCLUSIONS

An increase in the stability of rise time and stability of TIB after completing CBTI is associated with lower odds of using BZD-related hypnotics and any medication or substance for sleep 3 months after completing SMT. These sorts of changes are recommended as part of stimulus control and sleep restriction therapies that comprise the core behavioral treatment components of CBTI. Adherence to these recommendations, in theory, should enhance the patient’s perception of sleep stability/predictability, which in turn reduces the perceived need for medications to provide such control over sleep. Further research to test this speculation is needed.

DISCLOSURE STATEMENT

All authors have reviewed and approved this manuscript. Institution where work was performed: National Jewish Health. All authors received funding from a grant provided by the National Institute on Drug Abuse of the National Institutes of Health under grant award number 1R34DA038847-01A1 for their roles in conducting the research that provided the results summarized in this manuscript. Jack Edinger has received previous funding from Merck and donated sleep recording equipment from Philips-Respironics. He also serves as a paid consultant to Somly, a company that provides an online CBTI treatment program. Sheila Tsai is on the Sleep Physician Advisory Board for ResMed. Charles Morin has research contracts from Eisai, Idorsia, and Lallemand Health. He also serves as a member of the Advisory Board/Consultant for Eisai, Idorsia, Pear Therapeutics, and Sunovion and has royalties from Mapi Research Trust. The other authors report no conflicts of interest. The content presented herein is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

ABBREVIATIONS

BZD

benzodiazepine

BZRA

benzodiazepine receptor agonists

CBTI

cognitive behavioral therapy for insomnia

SMT

supervised medication tapering

TIB

time in bed