Cessation of Self-Reported Opioid Use and Impacts on Co-Occurring Health Conditions

Abstract

Background:

Among veterans in care reporting opioid use, we investigated the association between ceasing opioid use on subsequent reduction in report of other substance use and improvements in pain, anxiety, and depression.

Methods:

Using Veterans Aging Cohort Study survey data collected between 2003 to 2015, we emulated a hypothetical randomized trial (target trial) of ceasing self-reported use of prescription opioids and/or heroin, and outcomes including unhealthy alcohol use, smoking, cannabis use, cocaine use, pain, and anxiety and depressive symptoms. Among those with baseline opioid use, we compared participants who stopped reporting opioid use at the first follow-up (approximately 1 year after baseline) with those who did not. We fit logistic regression models to estimate associations with change in each outcome at the second follow-up (approximately 2 years after baseline) among participants with that condition at baseline. We examined two sets of adjusted models that varied temporality assumptions.

Results:

Among 2473 participants reporting opioid use, 872 did not report use, 606 reported use, and 995 were missing data on use at the first follow-up. Ceasing opioid use was associated with no longer reporting cannabis (adjusted odds ratio [AOR]=1.82, 95% confidence interval [CI] 1.10, 3.03) and cocaine use (AOR=1.93, 95% CI 1.16, 3.20), and improvements in pain (AOR=1.53, 95% CI 1.05, 2.24) and anxiety (AOR=1.56, 95% CI 1.01, 2.41) symptoms.

Conclusion:

Cessation of opioid misuse may be associated with subsequent cessation of other substances and reduction in pain and anxiety symptoms, which supports efforts to screen and provide evidence-based intervention where appropriate.

1. Introduction

The opioid overdose crisis persists as a major public health issue in the United States (US). Currently, overdose deaths are surging because fentanyl and other synthetic opioids are replacing or supplementing heroin and other illicit drugs (Centers for Disease Control and Prevention, 2020). Since 1999, over half a million people have died from an opioid-related overdose (Centers for Disease Control and Prevention, 2020). In addition to mortality, misuse of prescription opioids, illicit opioid use such as heroin, and opioid use disorder (OUD) are associated with significant morbidity including co-occurring use of other substances such as tobacco, alcohol, and other drugs (Azagba et al., 2019; Centers for Disease Control and Prevention, 2020; Grigsby & Howard, 2019; Witkiewitz & Vowles, 2018); mental health disorders such as depression and anxiety (Martins et al., 2012; Martins et al., 2009); infectious conditions, such as HIV and Hepatitis C virus (HCV), endocarditis (Hodder et al., 2021; Holtzman et al., 2021); adverse maternal and neonatal outcomes (Corsi et al., 2020; Maeda et al., 2014); and social consequences such as homelessness (Marshall et al., 2019). In this context, reducing prescription opioid misuse and use of illicit opioids, henceforth referred to as “opioid misuse”, and addressing OUD may subsequently reduce adverse impacts on health and wellbeing.

Screening for and responding to opioid misuse as appropriate is an important component of the public health response (Donroe et al., 2020; Leshner & Dzau, 2019; Schwartz et al., 2017; U.S. Preventive Services Task Force, 2020). This may include modifying a patient’s opioid prescriptions for greater safety, providing the naloxone to reduce risk of fatal overdose, or provision of medication for opioid use disorder (MOUD), such as methadone and buprenorphine. Moreover, integrating screening and intervention for opioid misuse with that of co-occurring health conditions—such as other unhealthy substance use, chronic pain, and mental health disorders—may be advantageous. Screening for one condition may reveal the presence of a co-occurring condition that would otherwise be undetected if that condition itself is not routinely screened for, and treating one condition may improve co-occurring conditions even if treatment was not targeted to those conditions (Caniglia et al., 2021; Caniglia et al., 2020; Khan et al., 2020). For example, treating OUD with medications could also treat a potentially co-occurring alcohol use disorder (Sudakin, 2016; Xu et al., 2021). Similarly, behavioral health treatment approaches for addressing opioid misuse may also address other unhealthy substance use, chronic pain, and co-morbid mental health conditions (Drake et al., 2004; Ehde et al., 2014; McHugh et al., 2010).

While research has shown that treating OUD can reduce overall opioid misuse and negative effects on health (Leshner & Dzau, 2019), there has been relatively little examination of whether cessation of any opioid misuse results in improvement in other domains of health. Most research in this area has focused on viral infections among people who inject drugs and have OUD, and has demonstrated that treatment with MOUD is associated with decreased risk of HIV and HCV, likely by reducing overall opioid use as well as unsterile injection (Metzger et al., 2015; Metzger et al., 2010; Metzger & Zhang, 2010; Springer et al., 2015; Woody et al., 2014). The sparse research also suggests that after initiating treatment for OUD, individuals are more likely to access health care (Huang et al., 2021), which may facilitate identification and treatment of otherwise unidentified and unaddressed health conditions (Rowe et al., 2012). Moreover, treating OUD is associated with overall improved quality of life (Mitchell et al., 2015), which may in turn affect co-occurring health conditions. Therefore, screening for and addressing the potential for opioid misuse, even if not yet at the threshold of OUD, could potentially result in improvements across a range of co-occurring health conditions, but to our knowledge, no studies have specifically assessed this.

To address this gap, we emulated a hypothetical randomized trial – a target trial (Hernan, 2021; Hernan & Robins, 2016; Hernan et al., 2016) – to evaluate whether ceasing opioid misuse, where we use self-reported prescription opioid use or heroin use as a proxy indicator for opioid misuse, may reduce unhealthy use of other substances (alcohol, stimulants, and tobacco) and improve pain, anxiety, and depression among those who reported opioid use using observational data from a sample of patients receiving medical primary care.

2. Materials and Methods

2.1. Study Sample

We used observational data from the Veterans Aging Cohort Study (VACS), which is an ongoing prospective cohort study conducted among veterans receiving care at nine Veterans Health Administration (VHA) centers: Atlanta, Baltimore, the Bronx, Dallas, Houston, Los Angeles, Manhattan/Brooklyn, Pittsburgh, and Washington, DC. VACS includes clinical, administrative, and survey data collected from approximately 3,500 participants with HIV and 3,500 controls without HIV that are frequency matched by age, race, gender, and site (Justice, Landefeld, et al., 2001). Survey data captures information including HIV-related risk factors, substance use, pain interference symptoms, and symptoms of mental disorders. Among participants enrolled in Atlanta, the Bronx, Houston, Los Angeles, Manhattan/Brooklyn, and Pittsburgh, surveys were administered approximately annually between 2003 to 2015; annual surveys were conducted between 2004 to 2015 among participants in Baltimore and Washington, DC. Institutional Reviews Boards at the VHA study sites, Yale University, approved all study activities; the secondary analysis of de-identified data conducted by New York University is considered non-human subjects. The analytic sample in the current secondary data analysis was restricted to participants who reported opioid use at one or more study visits (see Section 2.3.2 below).

2.2. Measures

2.2.1. Self-Reported Opioid Use.

At each annual survey, consistent with items in the National Survey of Drug Use and Health, participants were given a list of substances and asked to select the frequency of use in the past 12 months; response options included never used, no use in the past year, less than once a month, 1–3 times a month, 1–3 times a week, 4–6 times a week, and every day. We measured self-reported opioid use as endorsement of prescription opioids (e.g., Oxycontin, Vicodin, Percocet) and/or heroin use in the past year. We defined cessation of self-reported opioid use as no longer reporting any use of opioids (e.g., prescription or illicit) at the follow-up visit immediately following the initial self-report of use (see Section 2.3.2). This combined indicator of opioid use regardless of reason or source has been used previously in other VACS studies (Adams et al., 2021; Banerjee et al., 2016; Caniglia et al., 2020; Edelman et al., 2020) and since non-medical use of prescribed opioids is a risk factor for transitioning to illicit opioid use among VACS participants (Banerjee et al., 2016), our combined measure captures those who reported prescription opioid use at baseline and may now be using illicit opioids by the next follow-up survey.

2.2.2. Other Substance Use.

We measured past-year unhealthy alcohol use using the alcohol use disorder identification test (AUDIT), which is a screening instrument that assesses hazardous or harmful drinking using a 10-item questionnaire (Saunders et al., 1993; World Health Organization et al., 2001), with responses for each item scored from 0 to 4. Summing each item results in a total score ranging from 0 to 40; those who reported no alcohol use were scored as 0. Scores ranging from 8 to 14 suggest hazardous or harmful alcohol consumption and a score of ≥15 indicates likelihood of alcohol dependence. We dichotomized unhealthy alcohol use as AUDIT scores ≥8 (Gache et al., 2005). Participants were asked if they currently smoked cigarettes, and we categorized those responding into “currently smokes” versus those responding “never smoked” and “formerly smoked.” Past year cannabis and cocaine use (yes/no) were self-reported at each survey.

2.2.3. Pain Interference Symptoms.

We measured current pain interference symptoms using one item from the Health Survey Short-Form 12 that asked the participants how much pain had interfered with their normal work in the past month; we categorized those who responded “moderately,” “quite a bit,” or “extremely” as having moderate or severe pain, and those reporting “not at all” and “a little bit” as having little to no pain. (Becker et al., 2009; Novak et al., 2009; Stevens et al., 2020; Ware et al., 1996).

2.2.4. Mental Health Disorder Symptoms.

Participants were asked if they had felt nervous or anxious in the past four weeks and those reporting that this symptom “bothers me a little” or greater were categorized as having anxiety symptoms (Justice, Holmes, et al., 2001). We measured depressive symptoms using Patient Health Questionnaire (PHQ-9), with scores of ≥10 categorized as current depression (Kroenke et al., 2001).

2.3. Analysis

2.3.1. Target trial framework.

We applied a framework to emulate a hypothetical randomized trial to evaluate whether cessation of self-reported opioid use could lead to improvement or resolution of other substance use (e.g., alcohol, tobacco, cannabis, and cocaine), pain, and depressive and anxiety symptoms among individuals with opioid use. When it is not feasible to conduct a randomized trial, one can use the target trial framework to outline the key components of the hypothetical randomized trial, such as eligibility criteria, treatment strategies, start and end of follow-up, outcomes, causal contrasts of interest, and analysis plan, and describe how one can attempt to emulate these components using observational data (Hernan & Robins, 2016; Hernan et al., 2016). This process helps researchers avoid some of the common biases often found in observational studies’ design and analysis, for example, by aligning specification of eligibility, treatment assignment, and the start of follow-up (Hernan et al., 2016).

2.3.2. Eligibility criteria, treatment strategies, and start of follow-up.

In our analysis, we included individuals who indicated opioid use at one or more visits. We defined baseline (i.e., the start of follow up) as the initial survey during the study period at which the individual indicated past year opioid use. The two treatment strategies that we would compare in our target trial are: 1) resolve opioid use within the next year and 2) do not resolve opioid use within the next year. To emulate these treatment strategies, we compared 1) participants who reported no opioid use at the first follow-up visit (approximately one year after their baseline visit; i.e., resolved opioid use) and 2) participants who reported continued opioid use at the first follow-up after baseline. Those who had missing opioid use data at the first follow-up or who did not attend the first follow-up were classified as having an unknown opioid use status (see Section 2.3.4).

2.3.3. Outcome definitions and end of follow-up.

The primary outcomes were reporting reduction of alcohol use (AUDIT score to <8), stopping smoking, discontinuation of reporting cannabis and cocaine use (no use in the past year), improvement of moderate or severe pain (answering “not at all” or “a little bit” to the same pain question), improvement of depressive symptoms (defined as PHQ-9 ≤9), resolution of anxiety symptoms (answering “I do not have this symptom” to the same question) at the second follow-up (approximately two years after baseline), among those who had that condition. Using VACS survey data collected between 2003 to 2015, individuals were followed from baseline (i.e., the first survey on which one reported opioid use) until the second follow-up post baseline, the administrative end of follow-up, death, or censoring (defined as having an unknown opioid use status at the first follow-up). Hence, eligible participants for the analysis contributed data from three consecutive surveys.

2.3.4. Statistical analyses.

We fit separate logistic regression models to estimate odds ratios (ORs) for resolution of each outcome at the second follow-up comparing participants who stopped self-reported opioid use to those who did not stop. For each outcome, the analyses were restricted to individuals who had that condition at baseline. Therefore, the number of participants included in each analysis varied by outcome (alcohol use n=150; smoking n=553; cannabis n=290; cocaine n=274; pain interference symptoms n=565; anxiety n=454; depression n=300).

Eligibility criteria for our target trial specified that participants had opioid use data reported at baseline, and therefore those with missing data on baseline opioid use were not included in our analysis. Potential selection bias due to censoring those who did not have opioid use measured at the first follow-up after baseline and confounding by co-occurring condition status and demographic factors at baseline were accounted for by estimating two sets of inverse probability weights (Hernán & Robins, 2020). First, we fit a logistic regression model for having opioid use measured at the first follow-up versus missing measurement of opioid use at the first follow-up, conditional on baseline measures of HIV status, race, education, income, smoking status, AUDIT score, pain, depressive, and anxiety symptoms, and past-year cannabis, cocaine, and other stimulant use. These weights are used to account for measured differences between those who did and did not have opioid use measured at the first follow-up. Second, to account for measured differences between those who did and did not resolve opioid use (e.g., potential confounding), we fit a logistic regression model for no longer reporting opioid use versus reporting opioid use at the first follow-up, conditional on having opioid use measured at the first follow-up and the baseline variables mentioned above. All uncensored individuals (i.e., those with opioid use status measured at the first follow-up) received a weight that was inversely proportional to the product of their conditional probability of having opioid use measured and having the opioid use status that they reported. The weights were stabilized and truncated at the 99th percentile. Logistic regression models to estimate associations between cessation of reported opioid use and each outcome were then fit in the pseudo-population created by the inverse probability weights, in which the exposure (i.e., opioid use) is independent of the measure confounders (Hernán & Robins, 2020).

The exact time of cessation of opioid use was unknown among individuals who reported use at baseline and no longer reported opioid use at the first follow-up. This complicates confounding adjustments as it is unknown whether changes in the covariates of other substance use, pain, and mental health symptoms occurred between baseline and the first follow-up are caused by stopping opioid use or are causes of stopping opioid use. Assuming the former is true, those covariates are mediators and should be excluded from the model to avoid adjusting for factors that are on the causal pathway between opioid use and subsequent changes in other substance use, pain, and mental health symptoms, which could potentially underestimate effects (Figure 1a). Assuming the latter is true, those factors are confounders and should be included in the model to control for their potential confounding effects (Figure 1b). Since we are unable to know which of these scenarios is the truth, we examined two sets of adjusted models, as we have done in similar analyses (Caniglia et al., 2020). In the primary set of models, we assumed changes in opioid use status between baseline and the first follow-up occurred before changes in other substance use, pain, and mental health disorder symptoms measured at the first follow-up, and excluded these variables from the model for the weights (i.e., adjusted; Figure 1a). In a secondary set of models, we assumed that changes in opioid use status occurred after changes in the co-occurring conditions measured at the first follow-up, and included these variables in the model for the weights (i.e., adjusted: time-varying; Figure 1b).

Figure 1. Causal DAG for estimating the effect of cessation of opioid misuse (A) on co-occurring conditions (Y) under various assumptions about the causal structure of the data.

Figure 1a depicts assumptions for the adjusted models, in which potentially time-varying covariates measures at the first follow-up visit are excluded from the weight model; baseline covariates are included in the weight model.

Figure 1b.

depicts assumptions for the adjusted: time-varying models, in which potentially time-varying covariates measures at the first follow-up visit are included from the weight model; baseline covariates are included in the weight model.

V₀ Baseline covariates (e.g., age, race, HIV status, education, income; past year unhealthy alcohol use, cannabis, cocaine, other stimulants; and current pain symptoms, anxiety symptoms, depressive symptoms)

A1 Past year opioid misuse measured at first follow-up visit

L₁ Past year unhealthy alcohol use, cannabis, cocaine, other stimulants; and current pain symptoms, anxiety symptoms, depressive symptoms measured first follow-up visit

Y2 Outcome measured at second follow-up visit (for all outcomes)

3. Results

A total of 2,473 participants reported opioid use at baseline and were included in the analyses. Given VACS’ study design, half of participants had an HIV diagnosis (Table 1). The majority of the sample was male. Seventy-five percent were Black, approximately 90% had at least a high school education, and 60% reported an annual household income of less than $12,000. Co-occurring substance use was frequent, with approximately 27% with unhealthy alcohol use in the past year, 74% current smoking, 50% with past year cannabis use, 32% with past year cocaine use, and 12% with past year use of other stimulants. Almost half of the sample experienced moderate to severe pain in the past month. Approximately one-third of the sample experienced depressive symptoms, and over half experienced anxiety symptoms.

Table 1.

Baseline characteristics of included individuals, overall and by self-reported opioid use at the next survey, VACS

Baseline characteristic Number (%)	All individuals n=2473	No opioid use at next survey n=872	Any opioid use at next survey n=606	No opioid measure at next survey n=995
HIV status
Positive	1054 (56.9)	315 (60.7)	406 (60.3)	333 (40.5)
Negative	798 (43.1)	204 (39.3)	267 (39.7)	327 (49.6)
Race
African-American	1389 (75.0)	414 (79.8)	539 (80.1)	436 (66.1)
Other	463 (25.0)	105 (20.2)	134 (19.9)	224 (33.9)
Biological Sex
Male	2336 (94.5)	828 (95.0)	574 (94.7)	934 (93.9)
Female	137 (5.54)	44 (5.05)	32 (5.28)	61 (6.13)
Highest educational attainment
Less than high school	136 (7.34)	34 (6.6)	51 (7.6)	51 (7.7)
High school or more	1687 (91.1)	479 (92.3)	607 (90.2)	601 (91.1)
Missing	29 (1.6)	6 (1.2)	15 (2.2)	8 (1.2)
Annual household income
<$12,000	1126 (60.8)	327 (63.0)	433 (64.3)	366 (55.5)
≥12,000	661 (35.7)	175 (33.7)	217 (32.2)	269 (40.8)
Missing	65 (3.5)	17 (3.3)	23 (3.4)	25 (3.8)
Current (past year) unhealthy alcohol use
Yes	504 (27.2)	157 (30.3)	215 (32.0)	132 (20.0)
No	932 (50.3)	301 (58.0)	362 (53.8)	269 (40.8)
Missing	416 (22.5)	61 (11.8)	96 (14.3)	259 (39.2)
Current smoking
Yes	1362 (73.5)	369 (71.1)	523 (77.7)	470 (71.2)
No	490 (26.5)	150 (28.9)	150 (22.3)	190 (28.8)
Current (past year) cannabis use
Yes	935 (50.5)	253 (48.8)	331 (49.2)	351 (53.2)
No	879 (47.5)	253 (48.8)	325 (48.3)	301 (45.6)
Missing	38 (2.0)	13 (2.50)	17 (2.5)	8 (1.2)
Current (past year) other stimulant use
Yes	222 (12.0)	63 (12.1)	64 (9.5)	95 (14.4)
No	1575 (85.0)	435 (83.8)	590 (87.7)	550 (83.3)
Missing	55 (3.0)	21 (4.05)	19 (2.8)	15 (2.3)
Current (past year) cocaine use
Yes	597 (32.2)	169 (32.6)	203 (30.2)	225 (34.1)
No	1219 (65.8)	340 (65.5)	453 (67.3)	426 (64.6)
Missing	36 (1.9)	10 (1.9)	17 (2.5)	9 (1.4)
Current (past month) moderate or severe pain interference symptoms
Yes	812 (43.8)	227 (43.7)	295 (43.8)	290 (43.9)
No	1012 (54.6)	286 (55.1)	364 (54.1)	362 (54.9)
Missing	28 (1.5)	6 (1.16)	14 (2.1)	8 (1.2)
Current (past 2 weeks) depressive symptoms
Yes	611 (33.0)	180 (34.7)	233 (34.6)	198 (30.0)
No	1215 (65.6)	337 (64.9)	425 (63.2)	453 (68.6)
Missing	26 (1.4)	2 (0.39)	15 (2.2)	9 (1.4)
Current (past 4 weeks) anxiety symptoms
Yes	994 (53.7)	265 (51.1)	354 (52.6)	375 (56.8)
No	662 (35.8)	194 (37.4)	240 (35.7)	228 (34.6)
Missing	196 (10.6)	60 (11.6)	79 (11.7)	57 (8.6)

Among the 2,473 who reported opioid use, 872 (35.3%) reported no opioid use at the first follow-up (i.e., opioid use cessation), 606 (24.5%) reported opioid use, and 995 (40.2%) did not have data regarding opioid use; among those without data regarding opioid use, 921 (92.6%) did not participate in that follow-up, and 74 (7.4%) participated, but were missing data on opioid use (Table 1). Baseline sociodemographic characteristics, use of other substances, pain, and mental health symptoms were similar between those who reported opioid use at the first follow-up and those who did not, except use of other stimulants which appeared slightly higher among those who did not report opioid use (12.1% vs 9.5%). Those who did not have data regarding opioid use at the first follow-up did not appear to differ from those who did (Table 1).

In analyses that excluded potentially time-varying covariates that were measured at the first follow-up (i.e., adjusted models), ceasing opioid use did not appear to be associated with reporting no unhealthy alcohol or tobacco use at the second follow-up among those who had used those substances at baseline (Table 2). Ceasing opioid use was associated with approximately twice the odds of reporting no past year cannabis [adjusted odds ratio (AOR 1.82, 95% CI: 1.10, 3.03) and cocaine use (AOR 1.93, 95% CI: 1.16, 3.20) at the second follow-up among those who had used those substances at baseline. Among those who reported moderate to severe pain at baseline, ceasing opioid use was associated with approximately 50% increase in the odds of improving pain at the second follow-up (AOR 1.53, 95% CI: 1.05, 2.24). The odds of improving anxiety symptoms were nearly 60% higher for those who stopped opioid use compared to those who did not (AOR 1.56, 95%CI: 1.01, 2.41). While ceasing opioid use was strongly associated with improved depression symptoms in the unadjusted model (OR 2.93, 95% CI: 1.64, 5.24), the association was essentially null after adjustment (AOR 0.93, 95% CI: 0.57, 1.51).

Table 2:

Odds ratios and 95% confidence intervals for each condition improving or resolving at the second follow-up after baseline comparing those who stopped opioid use at the first follow-up after baseline with those who do not stop opioid use, under different assumptions about the causal structure of the data

Condition improves	Analysis	Odds ratios (95% CIs)
Past year unhealthy alcohol use	Unadjusted	1.16 (0.60, 2.24)
	Adjusted*	1.10 (0.56, 2.14)
	Adjusted: Time-Varying **	0.93 (0.47, 1.86)
Current smoking	Unadjusted	1.17 (0.75, 1.82)
	Adjusted*	1.19 (0.75, 2.89)
	Adjusted: Time-Varying **	0.91 (0.56, 1.49)
Past year cannabis use	Unadjusted	1.86 (1.14, 3.02)
	Adjusted*	1.82 (1.10, 3.03)
	Adjusted: Time-Varying **	1.23 (0.73, 2.09)
Past year cocaine use	Unadjusted	2.03 (1.25, 3.30)
	Adjusted*	1.93 (1.16, 3.20)
	Adjusted: Time-Varying **	1.30 (0.77, 2.20)
Current moderate or severe pain interference symptoms	Unadjusted	1.47 (1.02, 2,11)
	Adjusted*	1.53 (1.05, 2.24)
	Adjusted: Time-Varying **	1.23 (0.83, 1.82)
Current anxiety symptoms	Unadjusted	1.68 (1.10, 2.56)
	Adjusted*	1.56 (1.01, 2.41)
	Adjusted: Time-Varying **	1.48 (0.94, 2.33)
Current depressive symptoms	Unadjusted	2.93 (1.64, 5.24)
	Adjusted*	0.93 (0.57, 1.51)
	Adjusted: Time-Varying **	0.79 (0.47, 1.31)

^*Adjusted: All time-varying covariates measured at 1-year (past-year cannabis, cocaine, other stimulant, opioid use, unhealthy alcohol use, depressive symptoms, anxiety symptoms, and moderate or severe pain interference symptoms) are excluded from the model calculating the weights, based on the assumption that opioid use stopped prior to changes in the covariates; baseline covariates are included in the model calculating the weights.

^**Adjusted: Time-Varying: All time-varying covariates measured at 1-year are included in the model calculating the weights, based on the assumption that opioid use stopped after changes in the covariates; baseline covariates are included in the model calculating the weights.

In the secondary adjusted models in which we assumed that the changes in the other substance use, mental health, and pain covariates occurred before changes in opioid use and thus included them in the models estimating the inverse probability weights (i.e., adjusted: time-varying models), we found that the odds ratios were attenuated to various extents relative to the adjustment set in which those covariates were excluded (Table 2). The essentially null and non-significant associations between reporting no opioid use and unhealthy alcohol use, smoking, and depressive symptoms that were observed in the primary set of adjusted models were similar in the secondary set of adjusted models., The association between reporting no opioid use and anxiety symptoms was slightly attenuated and the confidence interval less precise (AOR 1.48, 95% CI: 0.94, 2.33) when including the time-varying covariates in the model. However, the results for reporting no cannabis and cocaine use and pain symptoms appeared to demonstrate greater attenuation and loss of precision with inclusion of time-varying covariates, which could suggest that these covariates measured at the first follow-up may be mediators of the relationship.

4. Discussion

In this sample of patients receiving medical care, we emulated a target trial of cessation of opioid use, in which we used self-reported use of prescription opioids and/or heroin as a proxy for misuse, on the outcomes of co-occurring unhealthy substance use and symptoms of pain and mental health disorders. We found that stopping opioid use was associated with subsequently reporting no cannabis and cocaine use, and with improvements in pain and anxiety symptoms. However, we did not find evidence of associations with unhealthy alcohol and tobacco use or improvement in depressive symptoms. Our results suggest that addressing opioid misuse in general medical settings, for instance via screening and evidence-based intervention, may reduce drivers of morbidity and mortality. Moreover, our results seem to contradict concerns that reduced prescription and/or illicit opioid use will increase pain or transition to unhealthy use of other substances, as we did not find any evidence that this is the case. Although our measure of self-reported opioid use may not capture individuals with OUD, considering that we found that reducing a potentially lower threshold of opioid use had significant associations with improving other domains of health, our findings inform the role of screening for and addressing opioid misuse and potential comorbid health conditions across clinical settings.

While we do not know the mechanism by which opioid use stopped, we found that cessation reporting of opioid use was associated with improvements in anxiety and pain symptoms, and findings were relatively robust to varying the assumptions about the potential time-varying covariates. These findings are aligned with prior studies that MOUD may ameliorate these symptoms among people with OUD (Ahmadi & Jahromi, 2017; Daitch et al., 2014; Pade et al., 2012) and prior studies suggest that treatment with evidence-based approaches is important to prevent recurrence of opioid misuse among individuals with OUD (Strain et al., 2022). Hence, our results may have important implications for treatment. OUD and anxiety disorders frequently co-occur (Conway et al., 2006), and people with both conditions often have increased risk of not receiving treatment for OUD (Lejuez et al., 2008) and for overdose (Lavie et al., 2009). Chronic pain may further complicate these comorbid relationships, given that pain-related anxiety (i.e., worry about experiencing negative effects of pain) is a risk factor for prescription opioid misuse (LaRowe et al., 2018; Rogers et al., 2018). Because the relationships among opioid use, anxiety, and pain appear to be bidirectional (Langdon et al., 2019), our findings showing that no longer reporting opioid use may subsequently improve anxiety and pain indicate the potential promise for screening and treating opioid use in order to disrupt complex, cyclical relationships among these health conditions.

We did not observe that participants no longer reported smoking after no longer reporting opioid use, which may be surprising since the prevalence of tobacco use is quite high among people who misuse opioids and those retained on OUD treatment have increased opportunity to receive effective smoking cessation mediations (Baldassarri et al., 2019; Yee et al., 2018). However, MOUD alone do not appear to result in smoking cessation (David et al., 2006), and our findings may suggest that smoking cessation requires specific attention and treatment (Streck et al., 2022), or a longer period of time is required to reduce smoking after stopping opioid use. This may also be the case for co-occurring unhealthy alcohol use, which is seen in roughly one third of individuals seeking treatment for OUD (Cicero et al., 2020; Rosic et al., 2017; Strain, 2002). Like smoking, specific alcohol treatment may be required in concert with addressing opioid misuse. Naltrexone, which has efficacy for alcohol use disorder and OUD, is more likely to be prescribed among people with both substance use disorders but has poorer retention compared to other medications for OUD (Mintz et al., 2021), suggesting that treatment with opioid agonists should be prioritized and improved interventions are needed for those with co-occurring disorders (Xu et al., 2021). In contrast, we found that reports of cannabis and cocaine use were resolved following cessation of opioid use, although findings were attenuated when including potential time-varying confounders in the models, which may indicate these factors lie on the pathway between ceasing opioid misuse and subsequently stopping use of other substances. People who misuse opioids frequently use cannabis (Cicero et al., 2020; Compton et al., 2021), and opioids and stimulants are often used together to balance effects of both drugs and/or to decrease negative effects of opioids (Leri et al., 2003). It is possible treatments such as contingency management may impact co-occurring cannabis and stimulant use (Cunningham et al., 2013; Griffith et al., 2000), or that when one stops opioid use they also cease use of other substances often used in conjunction.

Although anxiety symptoms improved following opioid use cessation in our sample, depression symptoms did not. One prior study that measured depressive symptoms following entrance into methadone treatment found that symptoms decreased quickly in the early stages of treatment and then leveled off (Wang et al., 2017). However, all participants in that study were receiving opioid agonist treatment and there was no comparison group. Further, those results indicated that decreases in depressive symptoms were greater among females and those who were younger; the participants in our sample are predominantly male and older age, which may be an explanation for the discrepancy.

Historically, screening and treatment of substance use has been restricted to specialty care settings. Considering that the majority of individuals with a substance use disorder are not aware or do not seek treatment (Substance Abuse and Mental Health Services Administration, 2019), incorporating screening and treatment into primary care may facilitate identification of people who may not have been reached otherwise and enables continuity of care in potentially less stigmatized settings (Donroe et al., 2020). Guidance from the US Preventive Service Task Force recommends yearly screening for drug use in adult patients (U.S. Preventive Services Task Force, 2020), although there is debate regarding how effective this would be in patients who are not actively seeking treatment (Saitz, 2020). In addition to expanding screening and treatment for opioid misuse, a complementary approach may be the screening and treatment of co-occurring substance use and mental disorders within opioid treatment and harm reduction service settings (Krawczyk et al., 2017).

Findings of our study must be considered in the context of important limitations. We attempted to estimate the effects of stopping self-reported prescription and/or illicit opioid use, which is a crude measure that will capture both OUD and use that has not progressed to OUD. Further, our measure combines illicit and prescription opioids, which likely have distinct pathways leading to initiation, resolution, and subsequent impacts on other health conditions. This measure of opioid use has been used in numerous other studies (Adams et al., 2021; Caniglia et al., 2020; Edelman et al., 2020), but some participants may report use of opioids as prescribed, although transitioning from prescribed to illicit opioid use has been observed in VACS (Banerjee et al., 2016), and the combined self-reported opioid use measure in this study allows us to capture this. Other substance use measures such as any reported cannabis and cocaine use are also crude and combine both people who use infrequently and recreationally as well as those who are dependent. Importantly, we do not know the mechanism by which opioid use stopped, and although our method to emulate a target trial is agnostic to the intervention that reduced the opioid use (i.e., treatment, cessation through other mechanisms), it is unable to elucidate what interventions may have the greatest impact on our outcomes. We cannot make causal interpretations, considering there is potential for unmeasured confounding and an inability to ascertain the timing of resolution of the various conditions in relation to each other, even with our careful design to emulate a randomized trial. However, the association between the unmeasured confounder and both opioid use and the outcomes would have to be relatively strong to attenuate our more robust findings. For example, we estimated an E-value (i.e., the minimum strength of an association on the ratio scale that the unmeasured confounder would need to have with both exposure and outcome) of approximately 2 would be needed to completely attenuate the association between ceasing opioid use and stopping cocaine use (VanderWeele & Ding, 2017). We do not know the exact timing of when opioid use resolved in relation to potential confounders and the outcomes. While some of our findings were robust to assumptions regarding temporality, others were attenuated. Finally, our findings may not be generalizable. Our analysis was conducted in a matched sample of Veterans with and without HIV who are engaged in care in the VA, and future research in samples with greater gender and racial/ethnic diversity is needed. While we were able to adjust for potential selection bias induced by not having opioid use measured at the first follow-up visit whether due to loss to follow up or non-response on the questionnaire item, data on the outcomes at the second follow-up visit was not complete.

In sum, our findings suggest that ceasing opioid use may be associated with resolving co-occurring cannabis and cocaine use and improving pain and anxiety symptoms among patients in care in the VA. However, varying assumptions regarding the timing of stopping opioid use in relation to other substance use, mental health, and pain covariates attenuated these associations, and may indicate that these factors are potential pathways through which ceasing opioid use affects health outcomes. Future research is needed to ascertain the temporality of the co-occurrence of opioid use, other substance use, mental health, and pain with greater specificity and to determine which specific interventions may be the most beneficial to reduce opioid misuse and the pathways through which opioid use cessation ultimately improves health across populations.

Highlights.

• Ceasing opioid use is associated with subsequently reporting stopping use of cannabis and cocaine

• Symptoms of pain inference and anxiety improved after ceasing opioid use

• Smoking and unhealthy alcohol use were not changed following ceasing opioid use