Insomnia predicts treatment engagement and symptom change: a secondary analysis of a web-based CBT intervention for veterans with PTSD symptoms and hazardous alcohol use

Abstract

Posttraumatic Stress Disorder (PTSD) and hazardous drinking are highly comorbid, and often more severe than PTSD or hazardous drinking alone. Integrated, web-based interventions for PTSD/hazardous drinking may increase access to care, but have demonstrated equivocal results in reducing PTSD and hazardous drinking. One factor that may explain treatment engagement and symptom change is the presence of insomnia symptoms. The current study conducted secondary data analysis of a randomized controlled trial of an integrated web-based intervention for PTSD symptoms and hazardous drinking to examine insomnia symptoms as predictors of PTSD symptoms, alcohol use, and treatment engagement. In the parent study, 162 veterans in primary care reporting PTSD symptoms and hazardous drinking were randomized to receive either the intervention or treatment as usual. The current study examined insomnia among veterans who received the intervention (n = 81). Regression models tested baseline insomnia symptoms as predictors of treatment engagement, follow-up PTSD symptoms, and alcohol use. Hierarchical regression models tested change in insomnia during treatment as a predictor of follow-up PTSD symptoms and alcohol use. Results showed baseline insomnia predicted treatment engagement and follow-up drinking days, but not PTSD symptoms or heavy drinking days. Although overall change in insomnia was small, it predicted follow-up PTSD and heavy drinking days, but not drinking days. Results are consistent with previous research highlighting the importance of identifying and treating insomnia in the course of integrated treatment for PTSD/hazardous drinking. Future research should investigate how to best integrate insomnia, PTSD, and/or hazardous drinking interventions to maximize treatment engagement.

Veterans in treatment for PTSD and alcohol use who had more insomnia symptoms completed less treatment. Those who improved their insomnia during treatment reported fewer PTSD symptoms and alcohol use.

Implications.

Practice: Insomnia symptoms may interfere with treatment engagement and response, so it may be necessary to assess and provide intervention for insomnia symptoms during the course of treatment for comorbid PTSD and hazardous alcohol use.

Policy: Policymakers interested in increasing access to treatment for comorbid PTSD/hazardous alcohol use should consider how factors such as insomnia may influence intervention effectiveness.

Research: Future research should continue to investigate how treatment for insomnia symptoms can best be integrated into treatment for comorbid PTSD/hazardous alcohol use.

The comorbidity between posttraumatic stress disorder (PTSD) and hazardous alcohol use, typically defined as alcohol consumption above recommended limits, is well documented through decades of empirical and theoretical support. Theoretical explanations such as the self-medication hypothesis [1], which posits that alcohol is used to alleviate aversive affective experiences that accompany PTSD, have garnered support from empirical research demonstrating startlingly high rates of comorbidity; one recent systematic review reported rates of comorbidity ranging from 2% to 63%, with the majority exceeding 10% [2]. Unsurprisingly, among those whose hazardous alcohol use also meets diagnostic criteria for alcohol use disorder (AUD), individuals with comorbid PTSD/AUD tend to experience more symptoms and greater functional impairment compared to individuals with either PTSD or AUD [3]. Interventions for comorbid PTSD/AUD have demonstrated efficacy in reducing symptoms of PTSD and AUD (e.g., Concurrent Treatment of PTSD and Substance Use Disorders Using Prolonged Exposure, [4]; Seeking Safety, [5]), however treatment-seeking among individuals with comorbid PTSD/AUD is low [3] and integrated treatments are underutilized [6], possibly due to logistical barriers and stigma surrounding mental health services [7]. Additionally, existing integrated PTSD/AUD treatment may not be accessible as an early intervention for those with hazardous alcohol use and/or subthreshold PTSD symptoms that may not meet diagnostic criteria.

Recent research efforts have focused on adapting evidence-based interventions for more accessible and less stigmatizing environments and to address hazardous drinking and/or subthreshold PTSD symptoms. Adaptations of integrated PTSD/AUD interventions for delivery via web-based, self-directed platforms have yielded promising results. Both the Thinking Forward [8, 9] and VetChange [10, 11] interventions consist of modules based on cognitive-behavioral therapy (CBT) emphasizing coping skills for PTSD symptoms and hazardous drinking. Results suggest that the interventions effectively reduced alcohol consumption [8, 11] however only Brief and colleagues [11] reported significant reductions in PTSD symptoms. Brief and colleagues noted participant attrition as an important limitation of the study; only 34% of the intervention group completed all modules. High attrition and limited treatment engagement are common to PTSD and AUD treatment, and associated with poorer treatment outcomes [12, 13]. Investigating engagement with integrated and alternative treatment approaches is needed.

One factor that may be associated with PTSD symptoms, hazardous alcohol use, and treatment engagement is insomnia. Among those with AUD and/or PTSD, insomnia symptoms are highly comorbid; research has reported 60%–90% comorbidity with AUD [14] and 70%–91% comorbidity with PTSD [15]. Among those reporting clinically significant insomnia symptoms but without diagnoses, 51.3% screened positive for hazardous alcohol use and 93.3% screened positive for clinically significant PTSD symptoms [16]. Additionally, among primary care patients who represent a range of alcohol use (none, moderate, hazardous), use of alcohol to manage sleep problems was associated with a significantly higher likelihood of hazardous drinking compared to moderate drinking [17]. As sleep problems are highly common among those with PTSD, it is not surprising that a PTSD diagnosis has been associated with a greater likelihood of using alcohol as a sleep aid, which may lead to alcohol-related problems [18].

Insomnia symptoms have been associated with poorer response to both PTSD and AUD treatment [19, 20]. Research suggests that insomnia symptoms often constitute a distinct disorder among those with comorbid PTSD/AUD and that treatment of PTSD, AUD, or comorbid PTSD/AUD is often not sufficient to resolve insomnia symptoms. Residual insomnia symptoms may have adverse effects on PTSD/AUD symptomatology following treatment, and less is known about the impact of insomnia symptoms on treatment for hazardous drinking and/or subthreshold PTSD symptoms in the absence of a PTSD or AUD diagnosis.

In one study, up to 48% of individuals continued to experience significant insomnia symptoms despite receiving trauma-focused treatment and no longer meeting criteria for PTSD [21]. López and colleagues [22] demonstrated that insomnia symptoms persist after other PTSD symptoms respond to treatment and this residual insomnia may limit response to trauma-focused treatment; reductions in insomnia over the course of treatment predicted treatment response for individuals with comorbid PTSD and depression symptoms. Recent research on the effects of insomnia among individuals in integrated PTSD/AUD treatment have shown results consistent with studies of insomnia among individuals in PTSD-only treatment: PTSD, but not insomnia, decreased over time during a residential integrated PTSD/AUD treatment program [23]. Additionally, despite associations between PTSD and insomnia severity upon treatment initiation, baseline insomnia was not associated with post-treatment PTSD symptoms. One recent study of insomnia among individuals participating in integrated outpatient PTSD/AUD interventions did demonstrate statistically significant decreases in insomnia symptoms following treatment, but the majority of participants remained above cutoffs for clinically significant insomnia [24]. In addition to clarifying the relationships among insomnia, PTSD symptoms, and alcohol use, and treatment engagement, investigating the relationship between baseline and residual insomnia within the context of a web-based intervention is warranted. Earlier studies have examined changes in insomnia during the course of integrated PTSD/AUD treatment in traditional residential [23] and outpatient [24] settings, but no study to our knowledge has examined insomnia during the course of an integrated, web-based intervention.

The purpose of this study was to conduct secondary analyses to examine the relationships between insomnia symptoms, treatment engagement, and symptom change for Thinking Forward [8, 9], a web-based CBT for veterans in primary care with PTSD symptoms and hazardous alcohol use. In the parent randomized clinical trial, this intervention was compared to a primary care treatment as usual (TAU) control group. Compared to TAU, the intervention group significantly decreased days of heavy drinking; coping, social support, self-efficacy, and hope for the future mediated the treatment effect [8]. There were no significant differences between groups in PTSD symptoms; both groups showed significant decreases in PTSD symptoms over time. Of multiple alternative explanations for the lack of statistically significant differences in PTSD symptoms between intervention and TAU groups, one that may have important clinical implications is the possible impact of insomnia on treatment engagement and symptom change.

The primary aim of this study was to test baseline insomnia as a predictor of (a) change in PTSD symptoms, (b) change in alcohol use, and (c) treatment engagement operationalized as module completion in the web-based CBT intervention. Based on the aforementioned association between PTSD and insomnia symptoms [e.g., Ref. [22]], we hypothesized that baseline insomnia predicts change in PTSD symptoms, alcohol use, and module completion. Specifially, we hypothesized that more severe baseline insomnia would be associated with more severe PTSD symptoms and more alcohol use at follow-up, and fewer treatment modules completed. A secondary aim was to investigate change in insomnia during treatment as a predictor of change in PTSD symptoms and alcohol use. We hypothesized that individuals who experienced greater reductions in insomnia would report greater reductions in PTSD symptoms and alcohol use at follow-up. Results from this study extend those from previous research [22–24] demonstrating the effect of insomnia symptoms on PTSD symptoms and alcohol use, and engagement with traditional PTSD and/or integrated PTSD/AUD treatment to a web-based, self-guided intervention for PTSD symptoms and hazardous drinking among veterans in primary care.

METHODS

Participants

Participants were 162 veterans recruited from 4 Veterans Affairs (VA) primary care clinics for a randomized clinical trial examining a web-based intervention, Thinking Forward, for PTSD symptoms and alcohol use compared to primary care TAU. Inclusion criteria were: veteran of the post-2001 conflicts in Afghanistan and Iraq; at least subthreshold PTSD symptoms (meeting DSM-IV criteria of reexperiencing, and either two avoidance or three arousal symptoms, problem duration, and functional impairment, per the Clinician Administered PTSD Scale (CAPS; [25, 26]); hazardous alcohol use (≥ 8) on the Alcohol Use Disorders Identification Test (AUDIT; [27]); and not currently in treatment for PTSD or substance use outside of primary care). Considering the current focus on factors related to treatment engagement, only participants who were randomized to receive web-based CBT (n = 81) were included in the sample.

Participants were primarily White (n = 71, 87%), male (n = 74, 92%), Army (n = 54, 67%) veterans between the ages of 22–55 (M = 31.81; SD = 7.65). Most of the sample had at least a high school education (n = 74, 91%) with nearly half (n = 34, 42%) having some college. The majority of participants was not married (n = 49, 61%) and slightly less than half of the sample was employed (n = 39, 48%). Most participants (n = 64, 79%) met diagnostic criteria for PTSD, while the remaining 21% (n = 17) fell in the subthreshold range. The average score on the PTSD Checklist-Military (PCL-M; [28]) was 46.91 (SD = 10.64), above the recommended cutoff for clinically significant PTSD symptoms. All participants met criteria for hazardous drinking with 42% (n = 34) meeting DSM-IV diagnostic criteria for current alcohol dependence and 8.6% (n = 7) meeting criteria for current alcohol abuse per the MINI International Neuropsychiatric Interview (MINI; [29]); two participants’ diagnostic status was unable to be identified due to missing data. On the Insomnia Severity Index (ISI; [30]) 46% of participants reported clinically significant insomnia symptoms (36% moderate severity, 10% severe), 40% had subthreshold symptoms, and the remaining 10% had no clinically significant insomnia symptoms. Due to missing data, the final samples for the models of baseline insomnia predicting 12-week PTSD (N = 54) and alcohol use (N = 70) and the final samples for the models of insomnia change predicting 12-week PTSD (N = 54) and alcohol use (N = 55) varied. Independent samples t-tests revealed no significant differences between participants who provided complete follow-up data (N = 55; 67.9%) and those who did not (N = 26; 32.1%) on baseline variables (ps > .05).

Materials and procedures

All participants completed an in-person initial eligibility assessment session. Eligible participants were introduced to Thinking Forward at that time. An in-depth explanation of study procedures can be found in Acosta et al. [8]. Participants were asked to engage with the Thinking Forward website over the course of 12 weeks which included 24 brief modules: 12 core and 12 optional. Participants’ engagement with the modules was self-directed and self-paced. Core modules consisted of cognitive behavioral skills training related to trauma and hazardous alcohol use, and the optional modules focused on related conditions (e.g., insomnia). Possemato et al. [9] provides in-depth information on module content. The 81 participants randomized to Thinking Forward completed 9 modules on average (SD = 6.14); 34 participants (42%) completed all 12 core modules. Seven participants (8.6%) did not complete any modules, 27 participants (33.3%) completed at least one optional module, and one participant completed all 24 modules.

Participants completed in-person follow-up assessments every 4 weeks during the study including measures of PTSD, alcohol use, and insomnia. The PCL-M was the primary outcome measure for PTSD symptoms over the past month. The PCL-M is a psychometrically sound measurement using a Likert-type scale to rate an individual’s experience of 17 PTSD symptoms; higher scores indicate greater PTSD symptoms. The Timeline Followback (TLFB; [31]) was used to measure self-reported daily quantity of alcohol. Participants provided retrospective reports of the number of drinks on each day over the past 90 days at baseline, and since the last assessment point at follow-ups. Percentage of drinking days (PDD) and percentage of heavy drinking days (PHDD; ≥4 standard drinks per occasion for women and ≥5 for men) were the alcohol outcomes. The TLFB has demonstrated strong psychometric properties including test-retest reliability and high content, criterion, and construct validity [31]. Insomnia over the past month was measured at each assessment point with the ISI [30], a 7-item measure in which insomnia symptoms are rated on a 5-point Likert scale; higher scores indicate greater insomnia symptoms. This measure has demonstrated strong psychometric properties [30, 32].

Statistical analysis

To examine baseline insomnia as a predictor of PTSD symptoms and alcohol use at the 12-week follow-up, three separate hierarchical multiple regression models were run in which 12-week PCL-M, PHDD, or PDD was the dependent variable. Baseline PCL-M, PHDD, and PDD were entered into the first step of the each model as covariates, and baseline ISI was entered in the second step. To test baseline insomnia as a predictor of module completion, a hierarchical multiple regression model was run in which baseline PTSD and alcohol use variables were entered in the first step as covariates and baseline ISI was entered in the second step predicting the total (i.e., core and optional) number of modules completed.

Similar to the López et al. [22] study, three hierarchical multiple linear regression models were used to examine change in insomnia as a predictor of changes in PTSD symptoms and alcohol use. Twelve-week PCL-M, PHDD, and PDD were the dependent variables. Insomnia change scores were computed by subtracting 12-week ISI from baseline ISI; higher scores indicate greater reductions in insomnia symptoms. Demographic variables (age, race, employment statis, marital status) were tested for associations with baseline and outcome variables using bivariate correlations (continuous variables) and chi-square tests (categorical variables). No significant associations were found (all ps > .05) and therefore demographic variables were excluded from analyses. Each model had three steps, to account for baseline symptom scores (step 1: baseline PCL-M, PHDD, or PDD), and treatment engagement (step 2: total number of modules completed). Insomnia change was added in the final step of each model. Pair-wise correlations were computed to check for multicollinearity (r ≥ .80); among baseline variables that were included in the models (ISI, PCL-M, PDD, PHDD); no evidence was found. Analyses were conducted in SPSS version 22 and alpha was set to .05 for all statistical tests.

RESULTS

Preliminary results

Mean ISI score indicated clinically significant insomnia symptoms at baseline (M = 14.67, SD = 6.02) and follow-up (M = 13.49, SD = 6.46). Mean PCL-M score suggested clinically significant PTSD symptoms at baseline (M = 46.91, SD = 10.64) and follow-up (M = 40.89, SD = 12.77). Average PDD was 50.62% (SD = 29.02%) at baseline and 41.35% (SD = 29.97%) at follow-up. Average PHDD was 27.64% (SD = 24.93%) at baseline and 17.46% (SD = 20.30%) at follow-up. Baseline ISI and PCL-M (r = .47, p < .001) and baseline PDD and PHDD (r = .54, p < .001) were related. There were no other significant baseline associations.

Baseline insomnia as a predictor of symptom change and treatment engagement

Table 1 presents results from regression models testing baseline insomnia as a predictor of 12-week PTSD symptoms, PHDD, and PDD. The full model predicting 12-week PTSD symptoms accounted for 27.6% of the variance. Adding baseline insomnia did not significantly increase variance accounted for (ΔF = 0.21, ΔR² = .003, p = .65). Baseline PTSD was a significant predictor of 12-week PTSD symptoms (β = .52, p < .001); those with higher baseline PTSD symptoms reported more PTSD symptoms at follow-up. No significant relationship was found between baseline insomnia and 12-week PTSD symptoms (β = .06 p = .65).

Table 1.

| Hierarchical regression models testing baseline insomnia as a predictor of 12-week PTSD symptoms and alcohol use

	Predictor statistics			Change statistics		Full model
Model	B (SE)	β	p	ΔF	ΔR2	R  2
PTSD symptoms						.28
Baseline PTSD	0.63 (0.17)	.52	<.001	22.33	.30
Baseline insomnia	0.13 (0.28)	.06	.65	0.21	.003
Percent heavy drinking days						.31
Baseline PHDD	0.49 (0.09)	.56	<.001	28.60	.30
Baseline insomnia	−0.60 (0.34)	−.18	.08	3.15	.03
Percent drinking days						.43
Baseline PDD	0.70 (0.10)	.67	<.001	48.07	.41
Baseline insomnia	−0.95 (0.45)	−.19	.04	4.44	.04

N = 54 for PTSD model; n = 70 for drinking models. PTSD symptoms measured using the PTSD Checklist-Military (PCL-M; [28]). PHDD and PDD calculated using the Timeline Follow-Back (TLFB; [31]). Insomnia symptoms measured using the Insomnia Severity Index (ISI; [30]). In each model, baseline PTSD symptoms (PHDD, PDD) was entered into the first step, and baseline insomnia was entered into the second step. Predictor coefficients reported are from the full model including both baseline PTSD symptoms (PHDD, PDD) and baseline insomnia.

PHDD percent heavy drinking days; PDD percent drinking days.

The full model predicting 12-week PHDD accounted for 30.8% of the variance. Adding baseline insomnia to the model did not significantly increase variance explained (ΔF = 3.14, ΔR² = .03, p = .08). Baseline PHDD was a significant predictor of follow-up PHDD (β = .56, p < .001); those with higher baseline PHDD tended to report higher PHDD at follow-up. Baseline insomnia was not a significant predictor of 12-week PHDD (β = −.18, p = .08).

The full model predicting 12-week PDD accounted for 43.4% of the variance. Adding baseline insomnia was associated with significant increases in variance explained (ΔF = 4.44, ΔR² = .04, p = .04). Baseline PDD (β = .67, p < .001) and baseline insomnia (β = −.19, p = .04) were significant predictors of 12-week PDD. Individuals with higher baseline PDD tended to report higher PDD at follow-up. Baseline insomnia and follow-up PDD were inversely related; those with higher baseline insomnia tended to report lower PDD at follow-up.

The full model predicting total module completion accounted for 21.3% of the variance. Adding baseline covariates significantly increased variance explained (ΔF = 3.09, ΔR² = .11, p = .03) as did baseline insomnia (ΔF = 14.49, ΔR² = .14, p < .001). Baseline insomnia significantly predicted module completion (β = −.44, p < .001); participants with higher baseline insomnia completed fewer modules. To account for possible confounding effects of sleep-related items on the PCL-M, nightmare and sleep items (items 2 and 13) were removed from the total PCL-M score at baseline. Results remained significant with only a slight decrease in R² (from .21 to .20) and in the magnitude of the insomnia coefficient (from −.44 to −.40).

Change in insomnia symptoms as a predictor of change in PTSD symptoms and alcohol use

Results from models testing change in insomnia as a predictor of 12-week PTSD, PHDD and PDD are presented in Table 2. On average, insomnia symptoms reduced during the study (M = 0.54, SD = 5.90), however there was considerable variability (range = −13.00–18.00).

Table 2.

| Hierarchical regression models testing change in insomnia as a predictor of 12-week PTSD symptoms and alcohol use

	12-week PTSD symptoms			12-week Percent Heavy Drinking Days (PHDD)			12-week Percent Drinking Days (PDD)
	B (SE)	β	ΔR2	B (SE)	β	ΔR2	B (SE)	β	ΔR2
Step 1			.30***			.37***			.45***
Baseline covariate	0.67 (0.14)***	.55***		0.43 (0.08)***	.61***		0.71 (0.11)***	.67***
Step 2			.01			.01			.02
Baseline covariate	0.66 (0.14)***	.55***		0.41 (0.08)***	.58***		0.75 (0.11)***	.71***
Modules completed	−0.26 (0.27)	−.11		0.26 (0.36)	.08		−0.87 (0.56)	−.16
Step 3			.18***			.07*			.02
Baseline covariate	0.70 (0.12)***	.58***		0.43 (0.08)***	.61***		0.77 (0.11)***	.73***
Modules completed	−0.30 (0.23)	−.13		0.20 (0.34)	.07		−0.91 (0.56)	−.17
ISI change	−0.93 (0.22)***	−.43***		−0.81 (0.31)*	−.27*		−0.75 (0.52)	−.15

N = 54 for PTSD model; n = 55 for drinking models. Reported R² values are adjusted. Baseline covariate is baseline PTSD for the PTSD model, baseline PHDD for the PHDD model, and baseline PDD for the PDD model. PTSD symptoms measured using the PTSD Checklist-Military (PCL-M; [28]). PHDD and PDD calculated using the Timeline Follow-Back (TLFB; [31]). Insomnia symptoms measured using the Insomnia Severity Index (ISI; [30]). Insomnia change score calculated by subtracting post-treatment ISI score from baseline ISI score; higher scores indicate greater reductions in insomnia symptoms.

PHDD percent heavy drinking days; PDD percent drinking days.

*p < .05.

^**  p < .01.

^***  p < .001.

The full model predicting 12-week PTSD accounted for 49.0% of the variance in the outcome. Adding baseline PTSD (ΔF = 22.33, ΔR² = .30, p < .001) and change in insomnia (ΔF = 17.87, ΔR² = .18, p < .001) significantly increased variance explained; module completion did not. Change in insomnia significantly predicted 12-week PTSD symptoms (β = −.43, p < .001). Greater reduction in insomnia was associated with fewer PTSD symptoms at follow-up. To account for the possible confounding effects of sleep-related items on the PCL-M, items referencing sleep problems (items 2 and 13) were removed from the total PCL-M scores at baseline and follow-up. Results remained significant with only a slight decrease in R² (from .49 to .43) and in the magnitude of the insomnia change coefficient (from −.43 to −.40).

The full model predicting 12-week PHDD accounted for 41.8% of the variance. Adding baseline PHDD (ΔF = 31.41, ΔR² = .37, p < .001) and change in insomnia (ΔF = 6.69, ΔR² = .07, p = .01) significantly increased variance explained; module completion did not. Change in insomnia significantly predicted 12-week PHDD (β = -.27, p = .01). Greater reduction in insomnia was associated with fewer heavy drinking days at follow-up.

The full model predicting 12-week PDD accounted for 46.3% of the variance. Adding baseline PDD significantly increased explained variance (ΔF = 42.99, ΔR² = .45, p < .001). Change in insomnia was not significantly related to 12-week PDD (β = -.15, p = .15).

DISCUSSION

The purpose of this study was to extend the literature to examine effects of insomnia symptoms on treatment engagement, alcohol use and PTSD symptom change within an integrated, web-based intervention for PTSD symptoms and hazardous drinking. Results showed that veterans with higher baseline insomnia symptoms had fewer drinking days at follow-up and completed fewer treatment modules. No relationship was found between baseline insomnia and PTSD symptoms at follow-up. Overall, reductions in insomnia and PTSD symptoms were small. However, reductions in insomnia were associated with decreases in PTSD symptoms and heavy drinking days, but were not related to overall drinking days at follow-up.

Results from analyses of baseline insomnia, change in insomnia, and PTSD symptoms were consistent with hypotheses and previous research demonstrating both that baseline insomnia is not related to post-treatment PTSD [23] and that change in insomnia is related to treatment response [22–24]. Results from analyses of baseline insomnia, change in insomnia, and alcohol use were mixed with respect to hypotheses. Change in insomnia predicted 12-week PHDD in the expected direction, but was not significantly related to PDD. These findings are consistent with the parent study [8] which demonstrated reductions in PHDD, but not PDD, in the intervention group. PHDD can be considered a better representation of hazardous alcohol use than PDD, and as such may be more amenable to change via the intervention, particularly as Thinking Forward is based on a harm reduction framework. As results are consistent with our findings related to PTSD symptoms, it is possible that changes in insomnia, hazardous drinking, and PTSD symptoms reflect changes in clinically significant targets of the intervention.

Greater baseline insomnia was associated with fewer drinking days at follow-up, and although baseline insomnia did not significantly predict PHDD, the coefficient was in the same direction and approached significance (p = .08). The lack of statistical significance for PHDD may be explained by its low variability compared to PDD. As such, the overall pattern of results may suggest a relationship between baseline insomnia and alcohol use in the opposite of the originally anticipated direction. Although our results were in contrast to prior research among individuals with AUD [23, 24], research among individuals with hazardous alcohol use, but not necessarily an AUD diagnosis, has yielded less consistent findings. Only 42% of our sample met diagnostic criteria for current alcohol dependence, and were not necessarily seeking treatment for alcohol use, PTSD, or insomnia prior to study enrollment. Our sample characteristics are in contrast to previous research with treatment-seeking participants whose alcohol or substance use met diagnostic criteria, and may explain the results found in the current study.

Studies among individuals using alcohol at hazardous levels, but without an AUD diagnosis have found null [33, 34] or negative relationships [17] between alcohol consumption and sleep problems. This may suggest that the alcohol/sleep relationship observed among those with AUD may not translate to those using alcohol at hazardous levels, but who have not reached the level of diagnosable AUD. Previous research among those with hazardous alcohol use but without AUD supports relationships between more nuanced indicators of alcohol consumption (e.g., alcohol use as a sleep aid; [17]) and sleep problems (e.g., sleep problems due to worry; [33]), even when relationships between broad assessments of alcohol use and sleep problems were not significant. Our findings coupled with previous research suggest that broad and general assessments of alcohol consumption and sleep problems alone may not be a sufficient indicator of alcohol-related sleep problems among those without an AUD diagnosis, and future research should identify specific indicators of alcohol/sleep-related risk applicable to this population.

Another important finding from the current study is the relationship between baseline insomnia and module completion, a possible proxy of treatment engagement. Results suggest that insomnia symptoms may present a barrier to treatment engagement, perhaps particularly for web-based interventions considering the problematic attrition rates and relationship between engagement and effectiveness noted in the literature [35].

This study’s findings contribute to a growing body of literature suggesting the importance of focusing on the effect of insomnia symptoms on alcohol use and PTSD symptoms, and as a barrier to treatment engagement for PTSD/hazardous drinking [36]. Although existing PTSD measures assess for nightmares and sleep disturbances (e.g., PCL-M [28]), an insomnia-specific assessment, such as the ISI [30], may better capture the clinical significance of insomnia symptoms among those with PTSD symptoms and hazardous drinking.

If clinically significant insomnia is present, insomnia-specific intervention may be warranted. Research indicates that Cognitive Behavioral Therapy for Insomnia (CBT-I), a 4–8 session treatment, effectively reduces insomnia among individuals with both PTSD [37] and AUD [38]. Another benefit of CBT-I, particularly given this study’s focus on a web-based treatment, is the availability of the companion mobile application CBT-I Coach. Research has shown the application to be feasible and acceptable for improving sleep [39], and treating insomnia first may provide more advantageous conditions for trauma-focused therapy [36].

Results of the current study should be considered along with its limitations. Specifically, the study is limited by its use of self-report data, particularly for the alcohol use outcome, which is subject to a number of biases (e.g., memory, social desirability). However, confidence in the validity of the TLFB is bolstered by significant correlations between TLFB data and collateral reports of alcohol use from the parent study [8]. As a secondary data analysis, the study is unable to evaluate causality, which is an important direction for future research. The sample was primarily White males, which may limit generalizability of findings. Finally, the sample size was limited by study attrition, although the attrition rate observed in the current study is comparable, if not favorable, to similar studies [11]. Despite these limitations, findings highlight important clinical considerations and directions for future research.

In conclusion, results are consistent with previous findings demonstrating that changes in insomnia are related to PTSD symptoms, alcohol use, and engagement with integrated PTSD/AUD treatment, and extend these findings to a web-based, self-directed intervention among those with subthreshold PTSD/hazardous drinking. Findings highlight differential relationships among baseline and change in insomnia, alcohol use, and PTSD symptoms, and suggest important clinical applications such as assessing for insomnia symptoms among veterans with PTSD symptoms and hazardous drinking and, when present, addressing these symptoms during treatment. Future research should examine the effects of a dedicated insomnia intervention in conjunction with integrated treatment for PTSD symptoms/hazardous drinking in order to determine the effectiveness of different modalities (e.g., in-person, web/mobile application) and timing (e.g., prior to or concurrent with treatment for PTSD symptoms and/or hazardous drinking) of an insomnia intervention. Effectively addressing insomnia symptoms may reduce barriers to PTSD/hazardous drinking treatment and improve treatment engagement.

Compliance with Ethical Standards

Conflicts of Interest: Drs. Buckheit, Nolan, Possemato, Maisto, Rosenblum, and Acosta have no conflicts of interest. Dr. Marsch is affiliated with the business that developed the web-based psychosocial intervention platform that was used in the study. This relationship is extensively managed by Dr. Marsch and her academic institution.

Human Rights: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. This study was approved by the Syracuse VA, Western New York VA and National Development Research Institutes Institutional Review Boards.

Informed Consent: Informed consent was obtained from all individual participants in the study.

Welfare of Animals: This article does not contain any studies with animals by any of the authors.

Disclaimer

Views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States government.

Transparency Statements

1. The study was pre-registered at www.clinicaltrials.gov (study identifier: NCT01710943).

2. The analysis plan was not formally pre-registered.

3. De-identified data from this study are not available in a public archive. De-identified data from this study will be made available (as allowable according to institutional IRB standards) by emailing the corresponding author.

4. Analytic code used to conduct the analyses presented in this study are not available in a public archive. They may be available by emailing the corresponding author.

5. Materials used to conduct the study are not publicly available.