The Association Between Self-Reported Anxiety and Retention in Opioid Agonist Therapy: Findings From a Canadian Pragmatic Trial

Abstract

Background

Prescription-type opioid use disorder (POUD) is often accompanied by comorbid anxiety, yet the impact of anxiety on retention in opioid agonist therapy (OAT) is unclear. Therefore, this study investigated whether baseline anxiety severity affects retention in OAT and whether this effect differs by OAT type (methadone maintenance therapy (MMT) vs. buprenorphine/naloxone (BNX)).

Methods

This secondary analysis used data from a pan-Canadian randomized trial comparing flexible take-home dosing BNX and standard supervised MMT for 24 weeks. The study included 268 adults with POUD. Baseline anxiety was assessed using the Beck Anxiety Inventory (BAI), with BAI ≥ 16 indicating moderate-to-severe anxiety. The primary outcomes were retention in assigned and any OAT at week 24. In addition, the impact of anxiety severity on retention was examined, and assigned OAT was considered an effect modifier.

Results

Of the participants, 176 (65%) reported moderate-to-severe baseline anxiety. In adjusted analyses, there was no significant difference in retention between those with BAI ≥ 16 and those with BAI < 16 assigned (29% vs. 28%; odds ratio (OR) = 2.03, 95% confidence interval (CI) = 0.94–4.40; P = 0.07) or any OAT (35% vs. 34%; OR = 1.57, 95% CI = 0.77–3.21; P = 0.21). In addition, there was no significant effect modification by OAT type for retention in assigned (P = 0.41) or any OAT (P = 0.71). In adjusted analyses, greater retention in treatment was associated with BNX (vs. MMT), male gender identity (vs. female, transgender, or other), enrolment in the Quebec study site (vs. other sites), and absence of a positive urine drug screen for stimulants at baseline.

Conclusions

Baseline anxiety severity did not significantly impact retention in OAT for adults with POUD, and there was no significant effect modification by OAT type. However, the overall retention rates were low, highlighting the need to develop new strategies to minimize the risk of attrition from treatment.

Clinical Trial Registration

This study was registered in ClinicalTrials.gov (NCT03033732).

Background

The opioid epidemic has become a significant public health crisis in North America, with opioid-related deaths involving fentanyl now Canada's leading cause of accidental deaths.^1,2 Roughly 13% of Canadians reported using an opioid analgesic in 2018, with 10% engaging in problematic opioid use. ³

Opioid use disorder (OUD) and prescription-type OUD (POUD) have empirically supported pharmacotherapies, including buprenorphine/naloxone (BNX) and methadone maintenance therapy (MMT). ⁴ BNX and MMT are 2 forms of oral opioid agonist therapy (OAT) considered the gold standard treatment for OUD and POUD.^5–8 Despite the efficacy of OAT, pharmacotherapy is underutilized.^7–10 In the United States, OUD rates are approximately 900 per 100,000, while maximum potential BNX and MMT national rates are 420 and 120 per 100,000, respectively. ¹¹ Among OAT initiates, approximately 40% discontinue treatment within the first 6 months, underscoring the importance of identifying modifiable risk factors to improve retention in treatment.^12–15 Psychiatric comorbidity is common among people with OUD and POUD.^16,17 In 1 cross-sectional study of adults with POUD, nearly half (47%) were diagnosed with a comorbid mood or anxiety disorder. ¹⁸ Co-occurring anxiety disorders have been associated with more severe polysubstance use, greater psychosocial adversity, poorer retention in treatment, more psychiatric symptoms, sleep impairment, and greater dose requirements for OAT to achieve stabilization.^18–24 BNX has been hypothesized to have some anxiolytic properties, which appear to be mediated by kappa opioid receptor antagonism at higher doses, which may, in turn, counter OUD-related dysphoria and negativism.^25–29 In 1 randomized controlled trial study, a single high dose of BNX reduced anxiety symptoms in adults with OUD and comorbid generalized anxiety disorder. ³⁰ In another study of dually diagnosed individuals with OUD, BNX appeared more effective for those with comorbid depression and anxiety. ³¹

Whether BNX may have additional efficacy over MMT among the specific subpopulation of patients with POUD and comorbid anxiety is unclear. Therefore, this study aimed to determine the impacts of baseline anxiety symptom severity on OAT retention and whether the type of OAT moderates the association between baseline anxiety severity and retention in treatment among adults with POUD.

Methods

Design

We conducted a secondary analysis of the Optimizing Patient Centred-Care: A Pragmatic Randomized Control Trial Comparing Models of Care in the Management of Prescription Opioid Misuse (OPTIMA), a 24-week pan-Canadian, phase IV pragmatic, open-label randomized noninferiority controlled trial of adults with POUD comparing flexible take-home BNX to daily witnessed MMT across 4 Canadian provinces. ³²

Participants

Eligible OPTIMA participants were clinically stable, treatment-seeking adults (aged 18–64) meeting DSM-5 ³³ criteria for POUD requiring OAT. In this study, we operationalized the POUD diagnosis as a DSM-5 diagnosis for OUD primarily attributed to prescription-type opioids (as opposed to heroin). Ineligible participants were those primarily using heroin, enrolled in OAT treatment during the past 30 days, pregnant, breastfeeding, or planning to conceive. We restricted our analytic sample to all eligible, randomized OPTIMA participants who completed the baseline anxiety self-report data. In addition, we excluded participants with missing data on key outcomes (i.e., retention in OAT) and covariates.

Procedures/Interventions

Participants were randomized 1:1 to supervised MMT or flexible take-home dosing BNX using a stratified permuted block design. OAT dosing and other clinical treatment aspects were pragmatic and left to the clinician's discretion. Most BNX participants received doses starting at 4 mg/1 mg and up to 24 mg/6 mg; MMT participants started at a maximum of 30 mg/day, up to 60–120 mg/day. ⁴ Take-home carries were permitted when participants were deemed clinically stable, and participants could switch OAT following randomization if deemed clinically necessary. Data were obtained from urine drug testing (UDT) and biweekly questionnaires over the 24-week study period, including general health status, sociodemographic details, and patterns of substance use. The institutional review boards at each participating site approved the study. The OPTIMA study was registered with clinicaltrials.gov (NCT03033732). Participants did not receive any psychosocial interventions beyond OAT and did not receive any treatments specifically to manage an anxiety disorder.

Measures

Explanatory. The primary explanatory variable was baseline anxiety symptom severity measured using the Beck Anxiety Inventory (BAI), a 21-item self-report instrument (scores from 0 to 63) based on DSM-III-R criteria for panic disorder. ³⁴ As per prior validation studies, we dichotomized anxiety scores, with BAI ≥ 16 indicating moderate-to-severe and <16 indicating none-to-mild. ³⁵

Outcome. The primary outcome was retention in the assigned OAT, defined as having both an active prescription and a positive UDT for the assigned OAT at week 24 from randomization.

Covariates. We selected covariates previously shown to be associated with the explanatory and outcome variables used in this study among persons with OUD.^32,36–40 Sociodemographic characteristics included age, gender, race/ethnicity, self-reported housing instability, educational attainment, and study site. Regarding the study site, we included which of the 4 study regions (British Columbia, Alberta, Ontario, and Quebec) the participants were seen through. Pain severity and pain interference were assessed using a 1-item pain severity score (sum of Brief Pain Inventory (BPI) ⁴¹ responses to questions 3 through 6) and pain interference measure (sum of responses to BPI question 9). Both BPI scales yield scores ranging from 0 to 10, with higher scores meaning worse symptoms: 1 to 4 indicates “mild” pain, 5 to 6 indicates “moderate” pain, and 7 to 10 indicates “severe” pain.^41,42

Other substance use covariates included DSM-5 severity of OUD, lifetime heroin use, previous OAT, and alcohol intoxication in the past 30 days (using responses from the Addiction Severity Index ⁴³ ). In addition, we analysed the results of baseline UDT for opioids, benzodiazepines, stimulants, and cannabinoids, which offered a measure of each substance use category in the past 30 days before study initiation.

Statistical Analyses

We used R (version 4.0.5; R Foundation for Statistical Computing, Vienne, Austria) for all analyses and considered P-values <0.05 statistically significant. ⁴⁴ We initially summarized selected participant characteristics, stratified by dichotomized baseline anxiety status (≥16 vs. <16 on the BAI). Then, we used the Pearson χ² and Mann–Whitney tests for nonparametric categorical and continuous variables, respectively.

We used logistic regression models to estimate the impact of baseline anxiety on OAT retention. First, we examine bivariable (i.e., unadjusted) associations between the outcome of interest (dichotomized retention in OAT) with our primary predictor variable (baseline anxiety status), the hypothesized effect modifier (type of assigned OAT), and other key covariates. Second, to estimate the independent association between baseline anxiety and retention, we conducted multivariable logistic regression models, which in addition to the primary explanatory variable, also included covariates associated with the outcome in bivariable analyses at P < 0.1. Third, we tested 2-way interactions to investigate if the type of OAT moderated the association between anxiety and retention in OAT. Finally, we repeated the same steps in a sensitivity analysis using retention in any OAT as the outcome (allowing for switches).

Results

Baseline Participant Characteristics

Among adults with POUD receiving OAT (N = 272), we excluded 4 participants (2%) due to the absence of data on retention in treatment, yielding a final analytic sample of 268 (98%) participants. The study sample consisted mostly of men (n = 173, 66%), and 179 (67%) participants were white. The median age was 38 (interquartile range (IQR) 31 to 46). Most of the sample (82%) had completed high school, while nearly half (46%) reported unstable housing. All participants had a moderate-to-severe OUD diagnosis (>5 DSM-5 symptoms in the past 12 months), and around half (56%) had prior OAT. As assessed by UDT, baseline substance use was highly prevalent: 95% for opioids, 67% for stimulants, 45% for cannabinoids, and 14% for benzodiazepines. Moderate-to-severe anxiety was highly prevalent in our sample, with 65% (n = 176 out of 268) having a BAI ≥ 16 (Table 1).

Table 1.

Participant and Covariate Characteristics of Interest, Stratified by Beck Anxiety Inventory Score (≥16 vs. <16).

Variable	Value	Total (n = 268)	Anxiety status		P-value
			High (n = 176)	Low (n = 92)
Assigned OAT	Buprenorphine	138 (51%)	98 (56%)	40 (43%)	0.077
	Methadone	130 (49%)	78 (44%)	52 (57%)
Region	Alberta	78 (29%)	60 (34%)	18 (20%)	0.055
	Quebec	72 (27%)	42 (24%)	30 (33%)
	Ontario	52 (19%)	30 (17%)	22 (24%)
	British Columbia	66 (25%)	44 (25%)	22 (24%)
Lifetime heroin use	Yes	183 (68%)	124 (70%)	59 (64%)	0.359
	No	85 (32%)	52 (30%)	33 (36%)
Age (years)	Q1	31	31	30.75	0.537*
	Median	38	38	37.5
	Q3	46	44	49
Gender identity	Man	173 (65%)	103 (59%)	70 (76%)	0 . 007
	Women, transgender, or other	95 (35%)	73 (41%)	22 (24%)
Race/ethnicity	White	179 (67%)	112 (63%)	67 (73%)	0.164
	Black, indigenous, and people of colour	86 (32%)	62 (35%)	24 (26%)
Current homelessness	Yes	123 (46%)	92 (52%)	31 (34%)	0.006
	No	137 (51%)	79 (45%)	58 (63%)
High school education or higher	Yes	218 (81%)	143 (81%)	75 (82%)	0.947
	No	49 (18%)	33 (19%)	16 (17%)
Severe opioid use disorder	Yes	262 (98%)	173 (98%)	89 (97%)	0.702
	No	6 (2%)	3 (2%)	3 (3%)
Lifetime opioid agonist therapy use	Yes	151 (56%)	102 (58%)	49 (53%)	0.545
	No	117 (44%)	74 (42%)	43 (47%)
Pain severity score	Q1	0	0	0	<0.001*
	Median	3	4	0
	Q3	6	6	4
Pain interference score	Q1	0	0	0	<0.001*
	Median	3	5	0
	Q3	7	7	3
UDT, opioids	Positive	252 (94%)	165 (94%)	87 (95%)	0.867
	Negative	14 (5%)	10 (6%)	4 (4%)
UDT, stimulants	Positive	176 (66%)	123 (70%)	53 (58%)	0.057
	Negative	89 (33%)	51 (29%)	38 (41%)
UDT, cannabinoids	Positive	122 (46%)	75 (43%)	47 (51%)	0.217
	Negative	144 (54%)	100 (57%)	44 (48%)
UDT, benzodiazepines	Positive	39 (15%)	29 (16%)	10 (11%)	0.299
	Negative	227 (85%)	146 (83%)	81 (88%)
Alcohol intoxication in the last 30 days	Yes	44 (16%)	28 (16%)	16 (17%)	0.891
	No	224 (84%)	148 (84%)	76 (83%)

Note. OAT = opioid agonist therapy; UDT = urine drug test. For housing instability, participants were asked to describe their housing stability. Again, we dichotomized “unstable” as responses to “very unstable” or “a little unstable” and the rest as “stable.” For P-values, those with an asterisk (*) were obtained using the Mann–Whitney test, while the remainder used the chi-square association test. Bold indicates that the p-values were statistically significant.

Factors Associated With Baseline Anxiety Status: Bivariable Analyses

In bivariable analyses stratified by baseline anxiety status (Table 1), men were less likely than women to report moderate-to-severe anxiety (OR = 0.44; 95% CI, 0.25 to 0.78), while those with higher baseline anxiety were more likely to report current unstable housing (OR = 2.18; 95% CI, 1.28 to 3.70), higher baseline pain severity scores (OR = 1.27; 95% CI, 1.15 to 1.40), and higher baseline pain interference scores (OR = 1.27; 95% CI, 1.16 to 1.39).

Factors Associated With Retention in Assigned OAT

Bivariable analyses: Overall, there were no differences in retention in assigned OAT between groups stratified by anxiety (29% vs. 28%; OR = 0.88, 95% CI, 0.51 to 1.53; P = 0.76). Also, there was no significant effect modification by type of OAT for retention in assigned (P = 0.41) or any OAT (P = 0.71). Covariates that were significantly associated (P < 0.05) with retention in assigned OAT in bivariable (i.e., unadjusted) analyses were OAT type (buprenorphine vs. methadone; OR = 0.58; 95% CI, 0.34 to 0.99; P = 0.048), study site (Ontario vs. British Columbia; OR = 4.64; 95% CI, 1.63 to 15.31; P = 0.006; and Quebec vs. British Columbia; OR = 26.85; 95% CI, 10.31 to 84.88; P < 0.001), lifetime heroin use (OR = 0.48, 95% CI, 0.28 to 0.86; P = 0.012), gender identity (man vs. woman, transgender, or other; OR = 2.03; 95% CI, 1.14 to 3.74; P = 0.020), ethnicity (white vs. black, indigenous, and people of colour; OR = 2.47; 95% CI, 1.33 to 4.79; P = 0.005), unstable housing (OR = 0.44; 95% CI, 0.25 to 0.77; P = 0.04), OAT history (OR = 0.42; 95% CI, 0.25 to 09.72; P = 0.002), and finally, urine drug screen positivity for stimulants (OR = 0.17; 95% CI, 0.09 to 0.30; P < 0.001).

Multivariable analyses: For multivariable models, we included all covariates significantly associated (P < 0.1) with retention in assigned OAT in bivariable models (Table 2). However, there was no significant association between baseline anxiety status and retention in assigned OAT (OR = 2.03; 95% CI, 0.94 to 4.40; P = 0.073). Covariates that were significantly associated (P < 0.05) with retention in assigned OAT in the multivariable analyses were type of OAT (buprenorphine vs. methadone: OR = 0.38; 95% CI, 0.18 to 0.78; P = 0.008), study site (Quebec vs. British Columbia; OR = 23.72; 95% CI, 6.96 to 80.82; P < 0.001), gender identity (man vs. woman, transgender, or other; OR = 2.81; 95% CI, 1.26 to 6.30; P = 0.012), and urine drug screen positivity for stimulants (OR = 0.40; 95% CI, 0.17 to 0.91; P = 0.028).

Table 2.

Unadjusted and Adjusted Logistic Regression of the Association Between Baseline Anxiety and Retention in Assigned Opioid Agonist Therapy at Week 24 of the OPTIMA Trial.

Variable	Unadjusted		Adjusted
	Odds ratio (95% CI)	P-value	Odds ratio (95% CI)	P-value
Anxiety status (BAI ≥ 16 vs. <16)	0.88 (0.51–1.54)	0.656	2.03 (0.94–4.40)	0.073
Buprenorphine versus methadone	0.58 (0.34–0.99)	0.048	0.38 (0.18–0.78)	0.008
Alberta versus British Columbia	1.64 (0.54–5.59)	0.395	1.72 (0.50–5.89)	0.386
Ontario versus British Columbia	4.64 (1.63–15.31)	0.006	3.22 (0.91–11.41)	0.07
Quebec versus British Columbia	26.84 (10.31–84.88)	<0.001	23.72 (6.96–80.82)	<0.001
Lifetime heroin use (yes vs. no)	0.49 (0.28–0.86)	0.012	0.88 (0.38–2.03)	0.762
Age (per year older)	1.02 (1.00–1.05)	0.086	1.02 (0.99–1.05)	0.273
Gender identity (man vs. woman, transgender, or other)	2.03 (1.14–3.74)	0.020	2.81 (1.26–6.30)	0.012
Ethnicity (white vs. Black, Indigenous, and people of colour)	2.47 (1.33–4.79)	0.005	0.71 (0.30–1.68)	0.438
Housing (unstable vs. stable)	0.44 (0.25–0.77)	0.004	0.60 (.28–1.28)	0.19
Education (high school vs. other)	0.63 (0.33–1.23)	0.167
OUD severity (severe vs. mild-to-moderate)	2.02 (0.32–39.07)	0.524
OAT history (ever vs. never)	0.42 (0.25–0.72)	0.002	1.14 (0.52–2.51)	0.743
BPI, pain severity score (per 1 point higher)	1.02 (0.93–1.12)	0.706
BPI, pain interference score (per 1 point higher)	0.98 (0.91–1.06)	0.680
UDT (positive for opioids)	5.45 (1.06–99.83)	0.105
UDT (positive for stimulants)	0.17 (0.09–0.30)	<0.001	0.40 (0.17–0.91)	0.028
UDT (positive for cannabinoids)	1.11 (0.65–1.89)	0.703
UDT (positive for benzodiazepines)	1.72 (0.83–3.47)	0.133
Alcohol intoxication (in the last 30 days)	1.96 (0.99–3.82)	0.049	1.10 (0.44–2.72)	0.844

Note. BAI = Beck Anxiety Inventory; BPI = brief pain inventory; OAT = opioid agonist therapy; OUD = opioid use disorder; UDT = urine drug test. P-values <0.05 are marked in bold.

Factors Associated With Retention in any OAT

Bivariable analyses: We repeated the previous analyses using retention in any OAT in sensitivity analyses. Again, there were no differences in retention in any OAT between groups stratified by anxiety (35% vs. 34%; OR = 0.88, 95% CI, 0.52 to 1.49; P = 0.74; Table 3). Also, there was no significant effect modification by type of OAT for retention in any OAT (P = 0.71). Covariates that were significantly associated (P < 0.05) with retention in any OAT in bivariable (i.e., unadjusted) analyses were study site (Ontario vs. British Columbia; OR = 3.07; 95% CI, 1.29 to 7.651; P = 0.013; and Quebec vs. British Columbia; OR = 16.18; 95% CI, 7.16 to 39.71; P < 0.001), lifetime heroin use (OR = 0.58, 95% CI, 0.34 to 0.98; P = 0.043), gender identity (man vs. woman, transgender, or other; OR = 2.02; 95% CI, 1.18 to 3.56; P = 0.013), ethnicity (white vs. black, indigenous, and people of colour; OR = 2.09; 95% CI, 1.19 to 3.78; P = 0.012), unstable housing (OR = 0.50; 95% CI, 0.29 to 0.84; P = 0.009), OAT history (OR = 0.54; 95% CI, 0.32 to 0.90; P = 0.017), and finally, urine drug screen positivity for stimulants (OR = 0.18; 95% CI, 0.10 to 0.30; P < 0.001) or benzodiazepines (OR = 2.25; 95% CI, 1.13 to 4.50; P = 0.021).

Table 3.

Unadjusted and Adjusted Logistic Regression of the Association Between Baseline Anxiety and Retention in Any Opioid Agonist Therapy at Week 24 of the OPTIMA Trial.

Variable	Unadjusted		Adjusted
	Odds ratio (95% CI)	P-value	Odds ratio (95% CI)	P-value
Anxiety status (BAI ≥16 vs. <16)	0.88 (0.52–1.50)	0.641	1.57 (0.77–3.21)	0.214
Buprenorphine versus methadone	0.77 (0.46–1.27)	0.302
Alberta versus British Columbia	1.16 (0.47- 2.91)	0.746	1.28 (0.49–3.33)	0.615
Ontario versus British Columbia	3.07 (1.29–7.65)	0.013	2.38 (0.83–6.83)	0.108
Quebec versus British Columbia	16.18 (7.16–39.71)	<0.001	14.93 (5.23–42.61)	<0.001
Lifetime heroin use (yes vs. no)	0.58 (0.34–0.98)	0.043	0.98 (0.46–2.12)	0.966
Age (per year older)	1.02 (0.99–1.04)	0.127
Gender identity (man vs. woman, transgender, or other)	2.02 (1.17–3.56)	0.013	2.31 (1.13–4.71)	0.022
Ethnicity (white vs. Black, Indigenous, and people of colour)	2.09 (1.19–3.78)	0.012	0.67 (0.32–1.40)	0.282
Housing (unstable vs. stable)	0.50 (0.29–0.84)	0.009	0.80 (0.41–1.58)	0.528
Education (high school vs. other)	0.82 (0.43–1.57)	0.534
OUD severity (severe vs. mild-to-moderate)	2.72 (0.43–52.51)	0.364
OAT history (ever vs. never)	0.54 (0.32–0.90)	0.017	1.54 (0.73–3.25)	0.254
BPI, pain severity score (per 1 point higher)	1.04 (0.95–1.14)	0.361
BPI, pain interference score (per 1 point higher)	1.01 (0.93–1.08)	0.885
UDT (positive for opioids)	2.01 (0.61–9.06)	0.293
UDT (positive for stimulants)	0.18 (0.10–0.30)	<0.001	0.31 (0.14–0.68)	0.004
UDT (positive for cannabinoids)	1.19 (0.72–1.98)	0.493
UDT (positive for benzodiazepines)	2.25 (1.13–4.50)	0.021	1.60 (0.67–3.82)	0.286
Alcohol intoxication (in the last 30 days)	1.91 (0.99–3.69)	0.052	0.99 (0.43–2.31)	0.988

Note. BAI = Beck Anxiety Inventory; BPI = brief pain inventory; OAT = opioid agonist therapy; OUD = opioid use disorder; UDT = urine drug test. P-values <0.05 are marked in bold.

Multivariable analyses: For multivariable models, we included all covariates significantly associated (P < 0.1) with retention in any OAT in bivariable models (Table 3). However, there was no significant association between baseline anxiety status and retention in any OAT (OR = 1.57; 95% CI, 0.77 to 3.21; P = 0.214). Covariates that were significantly associated (P < 0.05) with retention in any OAT in the multivariable analyses were study site (Quebec vs. British Columbia; OR = 14.93; 95% CI, 5.23 to 42.61; P < 0.001), gender identity (man vs. woman, transgender, or other; OR = 2.31; 95% CI, 1.13 to 4.71; P = 0.022), and urine drug screen positivity for stimulants (OR = 0.31; 95% CI, 0.14 to 0.68; P = 0.004).

Discussion

To our knowledge, the present study is among the few that have investigated the impact of anxiety on retention in POUD treatment. Our results showed that nearly two-thirds of the sample reported moderate-to-severe baseline anxiety, but higher baseline anxiety severity did not significantly affect retention in OAT, regardless of the type of OAT received. The present study aimed to expand on the initial results of the OPTIMA study,^1,45 which showed that BNX was safe and noninferior to MMT in reducing opioid use, as measured by the proportion of participants with opioid-free urine drug screens, regardless of fentanyl exposure. However, the primary OPTIMA study also reported that participants in the BNX group had a reduced OR of retention in assigned treatment compared to those in the MMT arm.

Relationship Between Anxiety and Retention in Treatment

Previous studies have not found any association between baseline anxiety and OAT retention.^46–48 Therefore, several factors might explain our results. Firstly, baseline anxiety alone may be irrelevant to overall treatment retention and guiding OAT selection. Secondly, poor retention rates in our study might have limited our ability to identify factors associated with retention. Thirdly, anxiety severity at the time of OAT initiation or treatment discontinuation might be more relevant than baseline anxiety. However, this may have been confounded by the potential influence of OAT on anxiety severity. Lastly, anxiety might have motivated participants to seek help and treatment, which might have compensated for the risk of dropout due to unstable mental health.

Furthermore, 1 study suggested distinguishing between state anxiety and trait anxiety to contextualize our findings, given that the BAI is a state-based measure of anxiety. ⁴⁸ State anxiety represents transient anxious reactions to specific adverse situations at specific moments, while trait anxiety describes individual differences related to a tendency to present in an anxious state. ⁴⁸ Ultimately, this distinction could help to understand our findings better, given that our measure of baseline anxiety may capture participants’ transient anxiety at the time of evaluation, which could have resulted from various aetiologies such as high rates of unstable housing, opioid withdrawal, and chronic pain.

Despite the lack of an independent association between baseline anxiety severity and OAT retention, participants may have benefited from the direct anxiolysis associated with receipt of OAT or indirectly through indirect psychosocial interventions offered during follow-up. Additionally, the severity of participants’ baseline anxiety may have exaggerated the perceived benefit of OAT on anxiety.

Clinical and Policy Implications of Findings for Improving Retention in Treatment

Baseline anxiety severity, our primary explanatory variable, was not significantly associated with retention in assigned or any OAT. However, multivariable regression models revealed significant associations between retention in assigned OAT and 4 variables: OAT type, stimulant use, gender identity, and study site. For retention in any OAT, only stimulant use, gender identity, and study site were significantly associated with multivariable analyses. We discuss these findings’ clinical and policy implications below, highlighting the potential for targeted interventions to enhance engagement and retention in OAT for individuals with POUD.

Of these factors, our findings suggest that MMT—a full opioid agonist—appears to be associated with comparatively higher odds of retention in the assigned OAT when compared to BNX—a partial opioid agonist. This discrepancy in retention within assigned OAT may stem from differences in medication efficacy and variations in individual responses and preferences. For example, a recent systematic review indicated that MMT generally yielded better overall treatment retention than sublingual BNX, potentially from client-centred factors, such as preferences in treatment. ⁴⁹ However, it is worth noting that while there were differences in retention in assigned OAT based on OAT type, there were no significant differences in overall retention (i.e., any OAT) between participants receiving BNX or MMT. This latter finding is consistent with the results from the main OPTIMA paper ⁴⁵ and other studies,^50,51 which indicate that individuals with OUD receiving BNX were more inclined to switch rather than discontinue OAT altogether. These findings are relevant because they support the principles of stepped, person-centred care, ⁵² underscoring how the primary objective of OUD treatment is to maintain OAT given its established benefits,^4,13,14 regardless of the specific type, as long as it effectively meets the needs of the individual receiving treatment.

Another modifiable factor was concurrent stimulant use, which our study measured by the presence of urine drug screen positivity for a stimulant upon study entry. Concurrent substance use and comorbid stimulant use disorder, which, if untreated, can impact the effectiveness of OAT and disrupt treatment adherence.^53,54 Therefore, comprehensive treatment and harm reduction interventions are vital for individuals with comorbidities.^20,22 Offering treatment for OUD and comorbid conditions highlights the importance of increasing access to integrated care, such as case management, specialized psychosocial programs, and targeted pharmacotherapeutic interventions where available. ⁵⁵

We also observed a tendency towards higher retention in assigned or any OAT among participants enrolled in the Quebec study sites (vs. British Columbia). However, we stress that the confidence intervals for these associations were wide, suggesting that the accuracy of this finding is likely limited and that there is a degree of uncertainty around these results due to the small sample size for specific study sites. While caution should be exercised in interpreting this finding, if the association holds, the observed discrepancy in retention across sites raises the possibility that variations in health-care delivery, local treatment protocols, access to community support, or other regional factors may influence retention in OAT. To improve treatment delivery for OUD and engagement with OAT, future studies must evaluate and improve treatment delivery for OUD across sites in Canada, which could involve increasing the availability of OAT programs, integrating them into other health-care settings, and increasing the use of virtual health delivery platforms (e.g., telehealth) to help improve access to those living in rural and remote regions with limited access to services.

Finally, we found that gender identity (man vs. woman, transgender, or other) was associated with higher retention in either assigned or any OAT. Ultimately, this finding points to the notion that retention in treatment may have also been generally influenced by various individual, social, and contextual factors, which may explain why gender identity was significantly associated with retention. For example, these differences may be influenced by disparities in health-care access, systemic biases, or social support that may have been different depending on individual gender identities. Recognizing these differences underscores the potential for tailored treatment approaches in OUD that remain mindful of gender identities.

Limitations and Future Directions

Our study is also subject to limitations. First, the study was a secondary analysis and not specifically designed to examine the impact of anxiety on OAT retention. Nevertheless, despite the relatively small sample size, this study provides valuable insights into the relationship between anxiety severity and OAT retention among adults with POUD.

Second, our study relied on self-reported data, which may not always be accurate or reliable. However, self-report measures among people who use drugs (PWUD) are generally reliable and valid.^56–58 Third, our study only assessed anxiety severity at baseline. The rationale for only using baseline anxiety symptoms was to evaluate if they could guide OAT choice at intake, also considering that OAT could influence anxiety over time and that using anxiety scores at other points in the trial could have confounded the aim of the study. Future studies should evaluate whether using a time-dependent measure of anxiety, such as mid-point or end-point anxiety severity scores, yields different results.

Fourth, we only examined retention at 24 weeks, which may not capture longer-term retention rates of the impact of anxiety on retention over time. We hope future studies will examine longer-term retention rates and reasons for attrition. Including more participants with more detailed follow-up data may provide insights into potential risk factors for disengagement from OAT in future studies.

Fifth, while we assessed a range of variables that could impact OAT retention, other potential factors could impact OAT retention, such as social support or access to care, that we could not include in our model due to insufficient data. Furthermore, increased flexibility in the BNX group may have potentially confounded the comparison for this substudy of the OPTIMA trial. However, we believe that the comparison of anxiety between the 2 groups remains valid. Furthermore, our analyses pointed to the lack of effect modification by the type of OAT. However, including a control group receiving standard supervised MMT provides a good comparison group to evaluate the efficacy of flexible take-home dosing BNX. Furthermore, using a randomized trial design helps minimize the impact of unmeasured confounding from these and other factors on our results.

Fifth, we only included Canadian participants with POUD, which may limit the generalizability of our study findings to other countries with different health-care systems and OAT programs. Similarly, we did not include participants who primarily used heroin or other illicit opioids, which could also impact the generalizability of our findings to other populations with OUD. However, the pragmatic design, including participants with varying treatment motivations, access to care, living situations, and comorbidities, was valuable because it represented real-world clinical populations.

Sixth, our study did not assess the impact of co-occurring psychiatric disorders other than anxiety, which could impact OAT retention. We also did not capture psychiatric disorder diagnoses and instead focused on symptoms, which could be associated with several psychiatric conditions. However, using a structured assessment tool to measure anxiety symptoms provides a standardized approach that could be applied to other co-occurring psychiatric disorders in future research examining the impact of other co-occurring psychiatric conditions.

Seventh, while participants did not receive psychosocial interventions or any treatments for anxiety disorders as part of the study, it is possible that study participation carried an indirect anxiolytic effect. Furthermore, participants could have received psychosocial interventions as part of the standard of care from the study site.

Conclusions

The results suggest that baseline anxiety severity does not significantly impact retention in OAT for adults with POUD, regardless of the type of OAT received. However, the low overall retention rates emphasize the need to identify and address factors that may contribute to treatment discontinuation. Strategies to improve retention and minimize the risk of attrition from treatment are needed to optimize treatment outcomes for individuals with POUD.