Associations between Medication Assisted Therapy Services Delivery and Mortality in a National Cohort of Veterans with Posttraumatic Stress Disorder and Opioid Use Disorder

Abstract

Objective:

Opioid use disorder (OUD) is a notable concern in the United States (US) and strongly associated with mortality. There is a high prevalence of OUD in patients with posttraumatic stress disorder (PTSD) and the mortality associated with OUD may be exacerbated in patients with PTSD. Medication-assisted treatment (MAT) for OUD has become standard of care for OUD and has been shown to reduce mortality. However, there has been little study of MAT and mortality in patients with PTSD and OUD.

Methods:

We conducted a retrospective cohort study in U.S. veterans who had newly engaged in PTSD treatment, were diagnosed with OUD and were provided MAT for at least one day between 2004 and 2013. We assessed mortality for one year following the index diagnosis date. We calculated all-cause mortality as well as death by external cause, overdose plus suicide, overdose, and suicide rates per 100,000. We used hazard ratios (HR) and 95% confidence intervals (CI) to compare death rates between patients with high versus low adherence to MAT. We evaluated the impact of high versus low exposure to general substance abuse care. We considered a confidence interval that did not cross one to be significant.

Results:

A total of 5,901 patients met inclusion criteria. Most patients were men and the average age was 43.3 years (SD=13.8). The all-cause mortality rate was 1,370 per 100,000 patients. High adherence to MAT resulted in a non-significant, decreased risk for death due to all-cause (HR=0.73, 95% CI [0.47, 1.13]), external cause (HR=0.71, 95% CI [0.38, 1.35]), and overdose or suicide (HR=0.66, 95% CI [0.33, 1.35]). Patients with high exposure ( 60 days) to general substance abuse care were significantly less likely to die due to external cause (HR=0.39, 95% CI [0.18, 0.85]) and overdose or suicide (HR=0.31, 95% CI [0.12, 0.77]).

Conclusions:

In patients with PTSD and OUD, improved adherence to MAT and greater exposure to general substance abuse care may result in lower mortality. Studies with longer follow-up and larger sample sizes to assess the impact of MAT on suicide are needed to confirm our findings.

Introduction

Misuse of opioids has reached epidemic proportions and has been declared a public health emergency in the United States (Wilkerson, Kim, Windsor, & Mareiniss, 2016). In 2015, more than two million Americans had a substance use disorder involving the use of prescription opioid medications and nearly 600,000 people had a substance use disorder involving the use of heroin (Substance Abuse and Mental Health Services Administration [SAMHSA], 2017). Of particular concern is the high prevalence of co-occurring opioid use disorder (OUD) and posttraumatic stress disorder (PTSD; Ecker & Hundt, 2018). In a national survey of over 10,000 individuals, Mills, Teesson, Ross, and Peters (2006) reported that as many as 33% of people with OUD met criteria for PTSD. Patients with PTSD experience a complex series of neurobiological events that alter the body’s stress response system and may disrupt their ability to regulate endogenous opioids (Valentino & Van Bockstaele, 2015; van der Kolk, 1994). Accordingly, patients with PTSD may have a higher tolerance to opioids (Valentino & Van Bockstaele, 2015) and may be more prone to misuse them (Danovitch, 2016; Valentino & Van Bockstaele, 2015).

Comorbid OUD may worsen the course of PTSD (Ecker & Hundt, 2018; Hassan, LeFoll, Imtiaz, & Rehm, 2017; Hildebrand, Behrendt, & Hoyer, 2015; Mills, Teesson, Ross, Darke, & Shanahan, 2005) and contribute to increased risk of mortality (Ahmed & Stanciu, 2017; Anestis, Tull, Bagge, & Gratz, 2012; Conner et al., 2013; Shorter, Hsieh, & Kosten, 2015). OUD profoundly disrupts the brain’s reward circuitry system (Volkow & Morales, 2015; Wilcox, Pommy, & Adinoff, 2016). In fact, OUD is strongly associated with all-cause mortality as well as death by overdose, suicide and accidents (Degenhardt et al., 2011). In a meta-analysis of 58 studies evaluating opioid-dependent samples, Degenhardt et al. (2011) reported a pooled standardized mortality ratio of 14.66, with overdose being the most common cause of death (650 deaths per 100,000 person-years) followed by other external causes of death including trauma (250 deaths per 100,000 person-years) and suicide (120 deaths per 100,000 person-years). Mortality associated with OUD may be further exacerbated in patients with PTSD because patients with PTSD may be more likely to act impulsively (Weiss, Tull, Sullivan, Dixon-Gordon, & Gratz, 2015) and to die by suicide or accident (Forehand et al., 2019; Gradus et al., 2010). Given the adverse health effects associated with OUD in patients with PTSD, it is imperative to identify treatments to alleviate symptoms and reduce mortality.

Medication-assisted treatment (MAT) for OUD has become standard of care for OUD (Sofuoglu, DeVito, & Carroll, 2019). Patients are provided opioid agonist (methadone or buprenorphine) or antagonist (naltrexone) treatment. MAT has established efficacy in reducing the use of opioids and improving treatment retention (Sofuoglu et al., 2019). MAT has been shown to decrease overall mortality as well as overdose mortality (Ma et al., 2019; Sordo et al., 2017). While the greater duration of MAT is protective, individuals may be at higher risk for death in the first few weeks of starting and stopping MAT (Degenhardt et al., 2011; Ma et al., 2019).

Although MAT reduces mortality in patients with OUD, there are has been relatively little attention on its role on health outcomes in patients with co-occurring PTSD and OUD. There is some suggestion that buprenorphine may improve PTSD symptoms (Ecker & Hundt, 2018; Seal et al., 2016). To date, no study has assessed the impact of MAT on mortality in patients with PTSD and OUD, yet there is emerging evidence it is being provided with greater regularity in patients with both disorders (Shiner, Leonard Westgate, Bernardy, Schnurr, & Watts, 2017). In a prior study of almost 20,000 veterans with PTSD and comorbid OUD treated in the Department of Veterans Affairs (VA) between 2004 and 2013, Shiner et al. (2017) found that the frequency of buprenorphine prescriptions rose from 2.0% to 22.7% and the frequency of naltrexone prescriptions rose from 2.8% to 8.6%. It is also unclear whether patients with OUD benefit from general substance abuse treatment visits in addition to MAT (Amato, Minozzi, Davoli, & Vecchi, 2011; Carroll & Weiss, 2017; Dugosh et al., 2016; Sordo et al., 2017). Finally, only a few studies have assessed the effect of MAT on suicide (Degenhardt et al., 2009; Molero et al., 2018).

Given the notable concern of elevated mortality risk in patients with PTSD and OUD, it is critical to understand the role that MAT may have in alleviating this risk. Therefore, we extended our prior work in our cohort of almost 20,000 veterans with PTSD and OUD by evaluating the association between MAT and mortality. Our overall goal was to understand whether there is an association between MAT, encompassing both dosing patterns and associated substance abuse therapy, and mortality. Our hypotheses were (1) that greater adherence to MAT would be associated with reduced mortality (including all-cause, external cause, overdose, and suicide mortality), (2) that greater exposure to high risk transition periods during MAT would be associated with increased the risk for death, and (3) that greater exposure to general substance abuse treatment visits would be associated with lower mortality.

Methods

Study population

Our study population included 5,901 VA outpatients who had newly engaged in PTSD treatment, were diagnosed with OUD in the year following the index PTSD diagnosis, and were provided MAT for at least one day in the year following the index PTSD diagnosis. Our study sample was drawn from a cohort that included 20,000 patients with PTSD and OUD who used VA healthcare between fiscal years (FY) 2004 through 2013 (Shiner et al., 2017). If a patient met inclusion criteria multiple times during the 10-year period, we only counted the first treatment episode. We assessed health services use and mortality for one year following the index date. We collected information on preexisting conditions in the year prior to the index episode.

Data sources

We used the VA Corporate Data Warehouse (CDW) to develop our study population. From the CDW, we collected demographic information, utilization, pharmacy, other treatment data as well as diagnostic data. We identified mortality via the VA Vital Status File. We determined cause of death using the VA-Department of Defense (DoD) Suicide Data Repository (SDR; Defense Manpower and Data Center), which links information on VA decedents to the Centers for Disease Control and Prevention (CDC) and National Death Index (NDI). We classified causes of death using the CDC National Center for Health Statistics ICD-10 definitions. The study was approved by the Veterans Institutional Review Board of Northern New England (VINNE).

Study variables

Table 1 describes the key variables of interest. We defined MAT as one or more active days of buprenorphine, naltrexone, or methadone doses. We identified buprenorphine and naltrexone prescriptions and days supplied using the outpatient medication database. Because methadone doses that are dispensed in supervised clinics are commonly not recorded as prescriptions, we used a documented visit to a methadone maintenance clinic (stop code 523) as a proxy for the receipt of methadone MAT. This approach aligns with the VA’s definition of stop code 523 (see Supplemental Online Material). Furthermore, our methods are similar to those of other VA researchers who have examined methadone MAT exposure using VA administrative data (Manhapra, Quinones, & Rosenheck, 2016). Per VA coding guidance, we considered that a single 523 stop code that was found within a week (i.e., Sunday – Saturday) represented 7 days of methadone MAT coverage. In our sample, we did observe that most patients who had a large number of 523 stop codes, had received a code on a nearly daily basis.

Table 1.

Study Variables

Variables of interest	Variable definition
Patient characteristics
Posttraumatic stress disorder with comorbid opioid use disorder	Patients with any opioid abuse or dependence diagnosis in the year following their index PTSD diagnosis. Thus all patients were diagnosed with both OUD and PTSD. Eligible patients were identified using ICD9 codes.
Other Characteristics	Gender, age, race, marital status, psychiatric comorbidities, medical comorbidities
MAT
Buprenorphine	One or more active days of prescriptions for buprenorphine (tablet formulation).
Methadone for MAT	Methadone maintenance clinic visits (stop code 523) was considered to be a proxy for the receipt of methadone MAT.
Naltrexone	One or more active days of prescriptions for naltrexone (long-acting injectable or tablet formulation)
MAT and non-MAT segments
New treatment segment (‘ON’ MAT)	Receipt of one or more days of MAT
Non-treatment segment (‘OFF MAT’)	Seven consecutive days without MAT
MAT initiation period	The first two weeks of a new treatment segment
MAT cessation period	The first two weeks of a non-treatment segment
MAT characteristics
Low adherence to MAT	<30% of the follow-up year was spent on MAT
High adherence to MAT Low exposure to high risk MAT transition periods	≥30% of the follow-up year was spent on MAT <20% of the follow-up year was spent in a MAT initiation or cessation period.
High exposure to high risk MAT transition periods	≥20% of the follow-up year was spent in a MAT initiation or cessation period.
Other interventions for OUD
Days of general substance abuse treatment visits	Includes total number of days of participation in any outpatient visits to substance abuse specialty clinics that does not include MAT.
Days of general substance abuse treatment visits and methadone maintenance visits	Includes the total combined number of days of participation in general substance abuse treatment visits or methadone maintenance clinic visits. In the event that a patient was seen in multiple clinics on the same day, we treated the encounters as one single day of exposure.
Low exposure to general substance abuse treatment visits	<10 days of visits in the follow-up year
High exposure to general substance abuse treatment visits	≥10 days of visits in the follow-up year
Low exposure to general substance abuse treatment visits and methadone maintenance visits	<60 days of visits in the follow-up year
High exposure to general substance abuse treatment visits and methadone maintenance visits	≥60 days of visits in the follow-up year
Outcomes
All-cause mortality	All-cause mortality within the first year of follow-up after entering the study cohort.
Cause of death	Death due to external cause, overdose, suicide, and suicide plus overdose within the first year of follow-up after entering the study cohort. Causes of death were identified using ICD-10 codes.

Note. MAT: medication-assisted treatment; ICD: International Classification of Diseases; OUD: opioid use disorder; PTSD: posttraumatic stress disorder.

During the follow-up period, it was possible that patients could have been trialed on several types of MAT. Consistent with the mechanism of action of medications provided for MAT, it was uncommon that a patient was on more than one type of MAT at any given time.

Quantifying MAT exposure and periods of risk

We included all patients with MAT episodes during the one-year follow-up period. We defined a new MAT segment as the receipt of one or more days of MAT after a period of seven or more days without MAT. As such, we assumed that intermittent MAT was continuous treatment unless the patient went more than seven consecutive days without MAT. We calculated the percentage of days that a person was covered by MAT during the year.

The optimal amount or duration of exposure to MAT for OUD is unknown (National Academies of Sciences, Engineering & Medicine, Health & Medicine Division & Board on Health Sciences Policy, 2018). However, the National Institute on Drug Abuse (2019) has stated that patients require a minimum of 12 months of methadone maintenance treatment. In lieu of the limited evidence and the fact that our cohort included only one year of follow-up, we selected cutoff points for high versus low adherence to MAT based on the data distributions. We selected values that ensured that each subgroup size was between one-half and two-thirds the size of the overall group. We made this decision because this approach would allow us to maximize power given the limited sample size and relatively rare outcomes.

Prior studies have suggested that patients may be at greater risk for death during MAT transition periods including the first two weeks of initiation or cessation of MAT (Ma et al., 2019; Sordo et al., 2017). Therefore, we hypothezised that patients who spent a higher proportion of the year in one of these high risk transition states would be at greater risk for death. We selected cutoff points to compare high versus low exposure to high risk MAT transition states based on this prior research and data distributions.

Defining general substance abuse treatment

We sought to understand the impact of general substance abuse treatments that are provided along with the medications to treat OUD (see Table 1). Our definition of general substance abuse treatment included all available stop codes for specialty substance use disorder care that are used in the VA healthcare system (513, 514, 519, 545, 547, and 560). The stop codes are broadly defined and may include a variety of different behavioral treatments that target a substance use disorder or a substance use disorder with comorbid PTSD. The treatments may be individual or group-based and delivered in-person, at home, over the phone or in the context of an intensive outpatient program (see Supplemental Online Material). The stop codes do not delineate the specific types of treatments offered (e.g., supportive therapy).

We selected cutoff points to compare high versus low exposure to these interventions based on data distributions. Similar to the approach that we took to define cutoff points for the amount of exposure to MAT, we selected cutoff points that broke the sub-groups up into sizes that were between one-half and two-thirds the size of the overall group. Again, this maximized our power in lieu of our limited sample size and relatively rare outcomes.

Outcomes

Our primary outcome assessed mortality including all-cause mortality, death due to external cause, overdose and death by suicide. Because suicide was rare in the cohort and may be misclassified as death due to overdose in the OUD population (Rockett et al., 2018), we reported a combined outcome of death due to overdose or suicide. We were also unable to report on the specific external causes of death (e.g., death to traffic accident) because the number of events by cause of death were too small to allow reporting (Wasserman & Ossiander, 2018).

Statistical analysis

We calculated all-cause mortality as well as death by external cause, overdose plus suicide, overdose, and suicide rates per 100,000 person for the study cohort. We used hazard ratios (HR) and 95% confidence intervals (CI) to compare death rates between patients exposed to the following conditions: (1) high versus low adherence to MAT, (2) high versus low exposure to high-risk MAT transition periods, (3) high versus low exposure to general substance abuse treatment visits, and (4) high versus low exposure to combined general substance abuse treatment visits and methadone maintenance clinic visits.

All models were adjusted for age, gender, marital status, black race and Hispanic ethnicity. Less than 1% of the covariates were missing. We calculated a Pearson’s correlation coefficient to assess whether there was evidence of a correlation between the number of days of general substance abuse treatment visits and MAT. The grand mean was used as a simple imputation when missing. Because mental health conditions are in the causal pathway for suicide (Conner et al., 2013), we did not adjust for these health conditions as doing so might have produced erroneous results.

We performed all analyses using SAS version 9.4 (Carey, NC). We considered findings to be statistically significant if the confidence interval did not cross one or the p-value was less than .05.

Results

Between FY 2004 and 2013, a total of 5,901 patients with newly diagnosed PTSD and OUD were treated with MAT. As shown in Table 2, most patients were men and the average age was 43.3 years (SD=13.8). Unlike methadone and buprenorphine, only 17% (n=994) of patients received naltrexone during the study period. More than half of patients had greater than 100 days of MAT. Approximately 32% of patients participated in general substance abuse treatment visits and methadone maintenance visits. There were 81 deaths due to any-cause within the year following the PTSD index diagnosis (a mortality rate of 1,370 per 100,000 patients). Fifty deaths were due to external causes and 31 were due to overdose or suicide. There were too few suicides (n< 11) to draw any meaningful conclusions about the impact of MAT on suicide by itself so we combined suicide with overdose.

Table 2.

Characteristics of Patients Who Newly Engaged in PTSD Treatment in the VA Healthcare System, Carried a Comorbid Diagnosis of Opioid Use Disorder and Received MAT

	% (n)
Total population	100.0 (5,901)
Baseline characteristics
Malea	93.4 (5,507)
Age, Mean (SD)	43.3 (13.8)
Demographicsa
Black	22.9 (1,339)
Hispanic	6.1 (361)
Marital status
Married	26.0 (1536)
Divorced/separated	43.3 (2554)
Other (never/widow/unknown)	30.7 (1811)
Psychiatric comorbiditiesb
Anxiety Disorders	69.0 (4,072)
Mood Disorders	50.8 (3,000)
Alcohol Dependence/Abuse	38.7 (2,285)
Medical comorbiditiesc
Hepatitis	22.9 (1,349)
Diabetes	10.5 (2,093)
COPD	5.9 (351)
Cancer	0.7 (133)
Treatment in the first year following index date
MATd Buprenorphine	48.0 (2,833)
Methadone	55.9 (3,299)
Naltrexone	16.8 (994)
Days of MAT
1–7 days	9.4 (555)
8–30 days	14.6 (863)
31–100 days	19.0 (1,120)
>100 days	57.0 (3,363)
Days of general substance abuse treatment visits
None	9.5 (563)
1–7 days	17.6 (1,036)
8–30 days	32.6 (1,924)
31–100 days	36.0 (2,127)
>100 days	4.3 (251)
Days of general substance abuse treatment & methadone maintenance visits
None	3.1 (181)
1–7	7.9 (464)
8–30	22.6 (1,332)
31–100	34.7 (2,046)
>100 days	31.8 (1,878)
Overall one-year crude mortality rate	1.4 (81)
One-year crude mortality rate due to external cause	0.8 (50)
One-year crude mortality rate due to overdose or suicide	0.5 (31)

Note. MAT: medication-assisted therapy; PSTD: posttraumatic stress disorder; COPD: chronic obstructive pulmonary disease.

^aThere was <1% missing data for each demographic characteristic.

^bResults shown for psychiatric comorbidities that are commonly observed in patients with PTSD (Brady, Killeen, Brewerton, & Lucerini, 2000).

^cResults shown for medical comorbidities that have been shown to be common causes of death in patients with PTSD (Forehand et al., 2019).

^dWhile patients could have been trialed on several different types of MAT during the follow-up period, it was uncommon that a patient was on more than one type of MAT at any given time.

As shown in Table 3, high adherence to MAT decreased the risk for death due to all causes (HR=0.73, 95% CI [0.47, 1.13]), external causes (HR=0.71, 95% CI [0.38, 1.35]), and overdose or suicide (HR=0.66, 95% CI [0.33, 1.35]). However, none of these findings was statistically significant. Conversely, patients who spent more than 20% of the follow-up period in a high-risk MAT transition period were at increased risk for death due to all causes (HR=1.77, 95% CI [1.15, 2.73]), external causes (HR=2.10, 95% CI [1.12, 3.93]), and overdose or suicide (HR=2.83, 95% CI [1.38, 5.83]).

Table 3.

Mortality Risk in Patients with PTSD and Comorbid OUD Who Receive Medication-Assisted Treatment^a

		All-cause mortality			External cause			Overdose or suicideb
	n	Deaths	Death per 100K	HR [95% CI]	Deaths	Death per 100K	HR [95% CI]	Deaths	Death per 100K	HR [95% CI]
Adherence to MAT
Low (<30% of follow-up year)	2,615	41	1,567.9	-	21	803.1	-	17	650.1	-
High (≥30% of follow-up year)	3,286	40	1,217.3	0.73 [0.47, 1.13]	18	547.8	0.71 [0.38, 1.33]	14	426.0	0.66 [0.33, 1.35]
Exposure to high-risk MAT transition periods
Low (<20% of follow-up year)	3,818	41	1,073.9	-	18	471.5	-	12	314.3	-
High (≥20% of follow-up year)	2,083	40	1,920.3	1.77 [1.15, 2.73]	21	1,008.2	2.10 [1.12, 3.93]	19	912.1	2.83 [1.38, 5.83]
Exposure to general substance abuse treatment visits
Low (<10 days)	1,791	36	2,010.1	-	16	893.4	-	14	781.7	-
High (≥10 days)	4,110	45	1,094.9	0.55 [0.35, 0.87]	23	559.6	0.63 [0.33, 1.20]	17	413.6	0.56 [0.27, 1.14]
Exposure to general substance abuse treatment visits and methadone maintenance visits
Low (<60 days)	3,142	50	1,591.3	-	30	954.8	-	25	795.7	-
High (≥60 days)	2,759	31	1,123.6	0.65 [0.41, 1.04]	S	S	0.39 [0.18, 0.85]	S	S	0.31 [0.12, 0.77]

Note. CI: confidence interval; HR: hazard ratio; K: 1,000; OUD: opioid use disorder; PTSD: posttraumatic stress disorder; MAT: medication assisted treatment; S: suppressed due to small cell sizes (n < 11).

^aModels were adjusted for age, gender, black race, Hispanic ethnicity and marital status

^bDue to small cell sizes (n < 11), we are unable to show the results of overdose or suicide and overdose alone. However, the analysis of overdose alone produced very similar results to that of overdose or suicide.

As shown in Table 3, Patients who had high exposure (≥10 days) to general substance abuse treatment visits had a lower risk for death due to all-cause as well as external cause, and overdose or suicide. However, the findings were only significant in the case of all-cause mortality (HR=0.55, 95% CI [0.35, 0.87]). Patients who had high exposure (≥60 days) to general substance abuse treatment visits and methadone maintenance clinic visits were also less likely to die during the follow-up period. Importantly, the findings were statistically significant in the case of external cause (HR=0.39, 95% CI [0.18, 0.85]) and overdose or suicide (HR=0.31, 95% CI [0.12, 0.77]). The correlation between the number of days of general substance abuse treatment visits and MAT was positive, but quite low (Pearson’s r= 0.11, p < .0001).

Discussion

Our work shows that better adherence and increased contact with substance abuse professionals are promising mechanisms to decrease mortality among veterans with PTSD and OUD receiving MAT. This indicates that clinicians treating patients in this population may save lives by ensuring that these patients have access to continuous medication supplies and maintain regular contact with mental health providers. Clinical processes should be designed to achieve this goal. While our conclusions are limited by small sample size and relatively low mortality rates, all models produced consistent trends.

Notably, our overall result found that MAT may have positive health benefits in patients with PTSD and OUD. While our sample sizes were too small to compare mortality across MAT, our findings suggest that improved adherence to any MAT may be protective against death in patients with PTSD and OUD. Prior studies have also found that a greater duration of MAT in patients with OUD may reduce the risk of death (Ma et al., 2019; Sordo et al., 2017). Furthermore, akin to our findings, other studies have shown that patients with OUD may be at higher risk for death during high-risk MAT transition periods including the first few weeks on and off MAT. However, these prior studies have only been able to evaluate high risk transition periods as it relates to methadone treatment (Ma et al., 2019; Sordo et al., 2017). While our work adds to this literature by also including naltrexone and buprenorphine, replication in larger samples is required to independently examine risk following cessation of buprenorphine and naltrexone, as we could only examine MAT as a class in this relatively small sample. Although there is some evidence that patients with PTSD and substance use disorder may also experience significant improvements in PTSD symptoms with MAT (Seal et al., 2016), we were unable to directly compare our findings with this evidence because our study focused solely on mortality. To the best of our knowledge, no prior studies have evaluated mortality outcomes in patients with PTSD and OUD who receive MAT.

The overall mortality rate in our study population was approximately two times that of the general U.S. population (850 per 100,000 people population) and six times that of U.S. adults between the ages of 40–44 years old (220 per 100,000 population; Xu, Murphy, Kochanek, Bastian, & Arias, 2018). Surprisingly, the all-cause crude mortality rates in our sample (roughly 1,370 per 100,000 person-years) was also higher than those reported by Ma et al. (2019) in their meta-analysis of 21 studies of MAT in opioid users (920 per 100,000 person-years). The discrepancy in findings may be due to several factors. First, our study focused exclusively on patients with PTSD and OUD. Second, we included patients who may have received as little as one dose of MAT during the follow-up period. Studies have suggested that the duration of MAT may contribute to its efficacy in reducing the risk of death (Ma et al., 2019; Sordo et al., 2017). Related to this evidence, we found that greater adherence to MAT may be protective against death by all causes, external causes, and overdose in patients with PTSD and OUD.

While previous studies have shown a positive effect of MAT on all-cause and overdose mortality in patients with OUD (Ma et al., 2019; Sordo et al., 2017), we are unaware of any prior studies that have explored the specific role of MAT in preventing death by external cause in patients with PTSD and OUD. This is an important question to address because patients with PTSD may be at elevated risk for death due to external cause such as an accident. For example, in a separate analysis of the overall parent cohort from which our study sample was drawn, Forehand et al. (2019) reported that among Veterans with PTSD there was a significant increase in mortality from accidental injury (SMR=1.99, 95% CI [1.83, 2.16], p < .001). It is conceivable that patients with PTSD and OUD may benefit from MAT because MAT mitigates their risk for engaging in impulsive behaviors that contribute to their elevated risk for death due to overdose or accident. Prior studies have also reported that buprenorphine may improve PTSD symptoms (Ecker & Hundt, 2018; Seal et al., 2016). This, in turn, may have important implications for suicide prevention in this population. Finally, while our sample sizes were too small to evaluate specific causes of death, Forehand et al. (2019) found that VA users with PTSD are at a two-fold elevated risk for death due to viral hepatitis compared to an age and sex-matched general population sample. Because viral hepatitis is associated with OUD and MAT plays an important role in preventing new (or recurrent) viral hepatitis, MAT could have other relevant health benefits in patients with PTSD (Rich, Bia, Altice, & Feinberg, 2018).

Our study highlights that more exposure to general substance abuse treatment visits may be protective in patients with comorbid PTSD and OUD. While there was a positive correlation between the number of days of general substance abuse treatment visits and MAT characteristics, the correlation was very low, suggesting that general substance abuse treatment may be an independent factor. To the best of our knowledge, the impact of general substance abuse treatment on mortality in patients with OUD has not been rigorously studied. While the role of behavioral interventions such as therapy in improving outcomes in patients treated with buprenorphine remains unclear (Carroll & Weiss, 2017; Sofuoglu et al., 2019), studies have suggested that behavioral therapy plus methadone maintenance can improve health outcomes (Sofuoglu et al., 2019). For example, contingency management in conjunction with methadone maintenance has been shown to be reduce drug use (Sofuoglu et al., 2019). These behavioral changes, in turn, could result in a reduction in mortality risk, whether it be from overdose, suicide, or other external causes. Contingency management may also be useful in combination with naltrexone, where treatment retention is a notable concern (Sofuoglu et al., 2019). Because we did not have granular data on the quality or types of substance abuse treatments that patients received, we are unable to comment more directly on which types of behavioral treatments may impact mortality in patients with PTSD and comorbid OUD. However, our findings highlight a critical need for studies that include information on treatment characteristics such as type of therapy and session length in order to better characterize the relationship between specific behavioral interventions and mortality outcomes in patients with PTSD and comorbid OUD. There is also a need for larger sample sizes in order to evaluate the relationship between MAT and death by suicide. If our findings hold true, this would have important implications for the design of clinical practice guidelines to inform the management of OUD in patients with PTSD.

To the best of our knowledge, we are the first study to characterize the role of MAT on mortality in patients with PTSD and comorbid OUD. Our study population is drawn from one of the largest healthcare systems in the United States. Our analysis made use of a comprehensive dataset that included various treatment characteristics. We used a rigorous process to identify and characterize MAT and adjusted for several potential confounders.

However, our study has several limitations. First, our sample included a small number of deaths because our study focuses on a relatively small population of younger individuals. There were also very few suicide cases. It would be valuable for future studies to include larger, longitudinal samples of all VA users, which represent a more diverse age group and a greater number of suicide cases. Second, we followed patients for one year following their index PTSD diagnosis. While this approach mitigated concerns about the temporal association between MAT and mortality, it reduced our power. Third, our population included primarily male veterans. Our findings may not be generalizable to other populations such as women and non-veterans. Fourth, we were unable to evaluate for differential effects on mortality and adherence across behavioral treatments. We also did not track whether patients received behavioral treatments outside of the VA. Fifth, we did not measure dosage and did not confirm whether patients took the medications that they were provided. There is mixed evidence regarding the impact of methadone dosage on mortality (Ma et al., 2019). Sixth, while some patients could have received methadone for pain (rather than MAT), our more stringent methods likely mitigated this risk. Seventh, we were unable to stratify results based on the type of MAT due to power and sample size. In addition, we defined high versus low treatment exposure based on the amount of exposure that occurred during the year following treatment initiation. Because no prior study has taken this approach to evaluating MAT and there is a lack of robust data to define the optimal cutoff for MAT exposure, it is difficult to directly compare our results with those of prior studies. Reassuringly, however, our results tended to align with trends reported in the literature. Eighth, because of the retrospective nature of our study, we relied on stop codes to define exposure to methadone MAT and general substance abuse treatment. Concerns have been raised regarding the use of Veterans Health Administration location codes to identify patients who have received specialty substance use disorder treatment (Harris, Reeder, Ellerbe, & Bowe, 2010). It is possible that in some situations, a location code may be assigned to a patient who was, in fact, seen for a reason other than opioid substitution treatment. As recommended in the literature (Harris et al., 2010), we took steps to mitigate this risk by coupling the stop code with a diagnostic code to confirm that the encounter was associated with treatment for an OUD. Finally, we are unable to make a definitive conclusion about causation. It is conceivable that patients who show up for visits and MAT might have better social support or other factors such as education that reduce their risk of death which are correlated with treatment conformance.

In conclusion, it appears that better MAT adherence and increased substance abuse treatment contacts may be associated with improved mortality during the initial year of care for patients with PTSD and OUD. However, in the majority of analyses, the findings were not significant. Further validation in larger samples, followed for more extended periods of time, is required to confirm our findings.