Examination of correlates of OUD outcomes in young adults: secondary analysis from the XBOT trial.

Abstract

Background and Objectives:

Opioid use disorder treatment outcomes are poorer for young adults than older adults. Developmental differences are broadly implicated, but particular vulnerability factor interactions are poorly understood. This study sought to identify moderators of OUD relapse between age-groups

Methods:

This secondary analysis compared young adults (18-25) to older adults (26+) from a comparative effectiveness trial (“XBOT”) that randomized participants to extended-release naltrexone or sublingual buprenorphine-naloxone. We explored the relationship between twenty-five pre-specified patient baseline characteristics and relapse to regular opioid use by age-group and treatment condition, using logistic regression.

Results:

Young adults (n=111) had higher rates of 24-week relapse than older adults (n=459) (70.3% vs 58.8%) and differed on a number of specific characteristics, including more smokers, more intravenous opioid use, more cannabis use. No significant moderators predicted relapse, in either 3-way or 2-way interactions.

Conclusions and Scientific Significance:

No baseline factors were identified as moderating the relationship between age group and opioid relapse, nor any interactions between baseline characteristics, age group, and treatment condition to predict opioid relapse. Poorer treatment outcomes for young adults are likely associated with multiple developmental vulnerabilities rather than any single predominant factor. Although not reaching significance, several characteristics (using heroin, smoking tobacco, high levels of depression/anxiety, or treatment because of family/friends) showed higher odds ratio point-estimates for relapse in young adults than older adults. This is the first study to explore moderators of worse OUD treatment outcomes in young adults, highlighting the need to identify predictor variables that could inform treatment enhancements.

Clinical trial registry for the parent study:

ClinicalTrials.gov NCT02032433

Introduction

Young adults, usually defined as ages 18-25 (or 26), in the United States are disproportionately affected by the current opioid crisis. Approximately 1.1% of young adults (392,000) had opioid use disorder (OUD) in 2016, with the highest per capita rates of misuse of both prescription opioids and heroin [1]. Further, approximately two-thirds of all overdose deaths among young adults in recent years involved opioids. With the recent availability of high purity heroin and especially of very high potency, illicitly manufactured fentanyl analogs, the rates of increases in overdose deaths have outstripped the rates of increase in use and OUD [2].

Medications for opioid use disorder (MOUD) have demonstrated effectiveness and are the standard of care for adults with OUD. The XBOT study [ref], the parent study that was the source of our secondary analysis, was a large comparative effectiveness trial that randomized participants with OUD to 24-weeks of treatment with either daily sublingual buprenorphine-naloxone (BUP-NX) or monthly injectable extended-release naltrexone (XR-NTX). The main findings included: that more participants successfully initiated BUP-NX than XR-NTX; that in the intention-to treat (ITT) analysis of all participants (whether or not they successfully initiated medication), 24-week opioid relapse rates were lower in the BUP-NX group; and that in the per-protocol analysis of those participants who successfully initiated either medication, 24-week opioid relapse rates were not significantly different between the groups.

XBOT included a fairly large sub-group of young adult participants. But despite the growing body of evidence that MOUD is effective for and should be considered first-line for youth, [3, 4, 5,6], youth tend to have poorer engagement in and response to MOUD [7][8]. We recently conducted a secondary analysis of the XBOT study, comparing the relapse rates of young adults (18-25) to older adults. In the ITT sample (n=570, all randomized participants), a main effect of age group was found, with higher relapse rates among young adults overall at 24 weeks (70.3%) compared to older adults (58.8%; OR= 1.72, 95% CI=[1.08, 2.70], p=.02). No significant main effect of age group was found on induction status, and no significant interactions were found between treatment assignment and age group on opioid relapse. That is, age group was not associated with likelihood of induction, nor was there evidence that the comparative response to medications (BUP-NX vs XR-NTX) in terms of relapse differed by age group.[9]

Speculation about reasons for overall poorer outcomes in youth have included features of their developmental vulnerability, including: lack of economic and social independence, early onset of SUD, limited engagement in clinical care or low motivation to change [10][11][12], subjective sense of invincibility, immature executive function, high rates of psychiatric comorbidity [13], OUD and MOUD-related stigma, misinformation about OUD risk and the potential benefits of MOUD, difficulties with sustaining medication adherence, and insurance and regulatory restrictions [14][15][16][17][18][19][20]. Assessment of effectiveness of treatments for OUD in this vulnerable population, considering such factors as accessibility, acceptability, initiation of MOUD, and retention, is a top public health priority. But no study that we know of has directly documented specific factors associated with differential treatment response by age group.

We therefore conducted a secondary analysis of the XBOT trial, to explore the potential moderating effect of age group on the relationship between various baseline demographic and clinical characteristics and opioid relapse. Our hope was that although the XBOT trial was not designed to specifically elucidate developmental differences, we might be able to identify among the broad range of assessments included in the data-set, certain hypothesized characteristics relevant to young adult vulnerability. Identification in this exploratory way of such baseline characteristics that might contribute to young adults’ poorer treatment outcomes, could provide clues for more refined future exploration of particular factors that could serve as prognostic markers, and perhaps eventually as therapeutic targets for intervention. Three-way interactions between these characteristics, medication group (BUP-NX vs XR-NTX), and age group were also explored, to see whether any of the baseline characteristics might contribute to differential treatment response to one of the medications or the other according to age group.

Methods

Brief Characteristics of Parent Study

The methods and design [21][22][23] of the parent multi-site trial are presented elsewhere. In brief, participants (ages 18 and over) seeking acute care for OUD were recruited during an index residential treatment episode from the routine patient flow at eight specialty SUD treatment sites during 2014-16. Participants were randomized in a 1:1 ratio to either daily sublingual BUP-NX or monthly injectable XR-NTX, having agreed that they would accept either as a randomized assignment. Participants were inducted onto the assigned medication and followed weekly in outpatient medication treatment for 24 weeks. This secondary analysis included all randomized participants from the ITT sample (N=570), including those who did not start the assigned medication (17%), discontinued the assigned medication (63%), or met relapse criteria (61%). All sites obtained local Institutional Review Board approval and all participants provided written informed consent.

Baseline characteristic measures

Participants were categorized into two age groups to assess moderation by dichotomized age: young adults (ages 18-25; N=111 or 19.5%, range 19-25; mean age = 23.1; Males = 59%) vs. older adults (ages 26 and up; N=459 or 80.5%; range 26-67; mean age = 36.5; Male = 73%). This age cut-off was used based on common definitions in the existing literature [24]. The strategy of dichotomizing age groups, as opposed to analyzing age as a continuous variable, was chosen for several reasons: because of the hypothesis that youth developmental vulnerability is a distinct categorical feature of youth rather than a continuous factor that would be persistently impactful throughout the rest of the lifespan; because it followed our prior secondary analysis that used this dichotomous strategy to show that young adults have worse outcomes than older adults; and because another secondary analysis [25] did not find continuous age to be a significant moderator of differential medication response in the overall sample.

Baseline demographic, substance use, mental health and other clinical measures of interest (variables) were preselected as potential factors moderated by age. Although the parent study assessments and measures were not originally designed to capture developmental or age differences, we selected several candidate characteristics from available baseline assessments that we hypothesized might reflect such differences based on the existing literature and clinical experience. Examples include: cannabis use (known to be more common in youth), being in treatment because of friends/family (hypothesized to reflect external motivation and family leverage, characteristic of youth), Stroop and Trails scores (measured as Color-Word and Interference score for Stroop and Trails B-Trails A difference for Trails, which are measures of executive function, with youth typically preforming worse), depressive symptoms (thought to be more persistent at presentation to treatment in youth [26]). Other baseline characteristics were chosen because they were found to be differentially distributed between the age groups. Examples include: current heroin use, smoking and past successful treatment. One characteristic, homelessness, was selected because it was found in a previous analysis to moderate outcome for the sample as a whole [26]. Finally, other baseline characteristics were more exploratory and chosen as known correlates of severity or prognosis for OUD in general, with the hypothesis that they might moderate outcome in either age group, without prior evidence that they should be different between the age groups. Details of the measures have been reported in the main findings of the parent study [21].

Outcome Measure

As in the parent study, relapse (yes/no) was defined as 4 or more consecutive weeks of any non-study opioid use (by urine toxicology, self-report, or failure to provide a urine sample); or 7 or more consecutive days of self-reported non-study opioid use, occurring at any point after day 20 post-randomization over 24-week follow-up. Self-reported substance use was collected using the Timeline Follow Back [27] method. Urine toxicology was conducted weekly and tested for opioids (buprenorphine, methadone, morphine [heroin, codeine, morphine], oxycodone). Fentanyl was not tested because at the time it was not yet highly prevalent, and there was not yet a CLIA-waived screening immunoassay. Other outcomes were considered, (e.g. time to relapse, opioid negative urine tests, self-reported opioid use, craving, etc.) but the primary relapse outcome was chosen as clinically meaningful and best suited to an initial exploratory analysis.

Statistical Analyses

Descriptive summaries of baseline characteristics between age and treatment groups were calculated using means and standard deviations for continuous measures and proportions for categorical measures. For each baseline characteristic, a logistic regression model was fit to estimate the odds of relapse over the 24-week trial (relapsed vs not relapsed). Each model included a 3-way interaction between baseline characteristic, treatment assignment (BUP-NX vs XR-NTX), and age group (young adult vs older adult), controlling for site as a random effect. When the 3-way interaction was not significant, it was omitted from the model and only the 2-way interaction between the baseline characteristic and age was assessed while adjusting for treatment. The effects of the baseline measure on relapse within each age group were computed as adjusted odds-ratios (aORs) and their 95% confidence intervals. All statistical tests were two-sided with a 5% significance level. Due to these analyses being fundamentally exploratory in nature, no family-wise error adjustments were used to correct for multiple comparisons.

Results

The randomized ITT sample consisted of 111 (19.5%) young adults ages 18-25 (range 19-25; mean age = 23.1) and 459 (80.5%) older adults ages 26 and up (range 26-67; mean age = 36.5). Descriptive summaries for the baseline characteristics by age group and treatment group are presented in Table 1. Participants were predominantly male (young adults vs older adults = 59% vs 73%), white (78% vs 73%), injection opioid users (77% vs 65%), and unemployed (64% vs 63%). Other notable differences included rates of past 30-day cannabis use (72% vs 49%), age of onset of opioid use (age 17 vs age 22), duration of opioid use (6 years vs 14 years), current smoker (96% vs 86%), having any friends/family with heroin/other opioid problems (72% vs 59%), scores on the Stroop Color-Word test (45 vs 41) and Stroop Interference test (3.6 vs 0.9). Of the young adults, 78 out of 111 (70%) relapsed to regular opioid use by 24-weeks, compared to 270 out of 459 older adults (59%).

Table 1.

Descriptive summaries of participant baseline clinical characteristics by age and treatment group among the ITT sample (N=570)

	By Age Group and Treatment								By Age Group
	Younger (≤25 years)				Older (>25 years)				Younger (≤25 years) (N=111)		Older (>25 years)		Difference between groups
	BUP-NX (N=62)		XR-NTX (N=49)		BUP-NX (N=225)		XR-NTX (N=234)
Baseline Measure	N	% or M (SD)	N	% or M (SD)	N	% or M (SD)	N	% or M (SD)	N	% or M (SD)	N	% or M (SD)	p-value*
Demographics
Gender (% male)	35	56.5%	31	63.3%	171	76.0%	164	70.1%	66	59.5%	335	73.0%	0.005
Substance Use
Any current heroin use (% yes)	58	93.5%	41	83.7%	193	85.8%	207	88.5%	99	89.2%	400	87.1%	0.559
Any current intravenous use (% yes)	49	79.0%	36	73.5%	147	65.3%	153	65.4%	85	76.6%	300	65.4%	0.024
Cigarette smoker													0.007
Non-smoker	3	4.9%	1	2.0%	26	11.7%	40	17.2%	4	3.6%	66	14.5%
Some days	8	13.1%	3	6.1%	23	10.3%	10	4.3%	11	10.0%	33	7.2%
Every day	50	82.0%	45	91.8%	174	78.0%	183	78.5%	95	86.4%	357	78.3%
Any current cocaine/crack use (% yes)	22	35.5%	20	40.8%	106	47.1%	92	39.3%	42	37.8%	198	43.1%	0.310
Any current cannabis use (% yes)	43	69.4%	36	73.5%	109	48.4%	117	50.0%	79	71.2%	226	49.2%	<.001
Past Treatment History
Current treatment is first treatment in lifetime (% yes)	22	35.5%	21	42.9%	87	38.7%	79	33.8%	43	38.7%	166	36.2%	0.614
Past methadone or buprenorphine treatment Successful (% yes)	14	22.6%	13	26.5%	72	32.0%	102	43.6%	27	24.3%	174	37.9%	0.007
Past naltrexone treatment successful (% yes)	6	9.7%	3	6.1%	5	2.2%	10	4.3%	9	8.1%	15	3.3%	0.023
Opioid Withdrawal
Subjective Opioid Withdrawal Scale (range: 0-64)	62	13.6 (11.1)	49	16.7 (14.0)	225	16.1 (13.6)	234	15.3 (13.3)	111	15.0 (12.5)	459	15.7 (13.4)	0.610
Other Psychiatric Symptoms/Disorders
Hamilton Depression scale													0.585
None	35	56.5%	26	53.1%	102	45.5%	129	55.1%	61	55.0%	231	50.4%
Mild	21	33.9%	17	34.7%	84	37.5%	79	33.8%	38	34.2%	163	35.6%
Moderate/Severe	6	9.7%	6	12.2%	38	17.0%	26	11.1%	12	10.8%	64	14.0%
Anxiety-depression moderate or extreme (EuroQoL)													0.003
None	15	24.2%	7	14.3%	72	32.0%	85	36.3%	22	19.8%	157	34.2%
Moderate/Extreme	47	75.8%	42	85.7%	153	68.0%	149	63.7%	89	80.2%	302	65.8%
Suicidal thoughts (30 days prior to admission) (% yes)	6	9.7%	1	2.0%	8	3.6%	16	6.8%	7	6.3%	24	5.2%	0.653
Social/Living Situation
Current homelessness (% yes)	14	22.6%	8	16.3%	55	24.4%	66	28.2%	22	19.8%	121	26.4%	0.154
Any friends/family with heroin/other opioid problems (% yes)	47	75.8%	33	67.3%	124	55.6%	142	61.5%	80	72.1%	266	58.6%	0.009
Legal Issues
Current probation/parole (% yes)	16	25.8%	9	18.4%	34	15.1%	33	14.2%	25	22.5%	67	14.6%	0.043
Motivation for Participating/ Attitudes Regarding Medication
Participating to avoid relapse													0.035
Disagree	1	1.6%	0	0.0%	13	5.8%	8	3.4%	1	0.9%	21	4.6%
Neutral	3	4.8%	0	0.0%	2	0.9%	1	0.4%	3	2.7%	3	0.7%
Agree	58	93.5%	49	100.0%	210	93.3%	224	96.1%	107	96.4%	434	94.8%
Participating because of family/friends													0.676
Disagree	46	74.2%	27	55.1%	142	63.1%	159	68.2%	73	65.8%	301	65.7%
Neutral	6	9.7%	14	28.6%	44	19.6%	51	21.9%	20	18.0%	95	20.7%
Agree	10	16.1%	8	16.3%	39	17.3%	23	9.9%	18	16.2%	62	13.5%
Prefer to receive BUP-NX													0.367
Disagree	12	19.4%	12	24.5%	61	27.1%	56	24.0%	24	21.6%	117	25.5%
Neutral	26	41.9%	18	36.7%	96	42.7%	99	42.5%	44	39.6%	195	42.6%
Agree	24	38.7%	19	38.8%	68	30.2%	78	33.5%	43	38.7%	146	31.9%
Prefer to receive XR-NTX													0.777
Disagree	13	21.0%	13	26.5%	49	21.8%	54	23.2%	26	23.4%	103	22.5%
Neutral	32	51.6%	18	36.7%	104	46.2%	119	51.1%	50	45.0%	223	48.7%
Agree	17	27.4%	18	36.7%	72	32.0%	60	25.8%	35	31.5%	132	28.8%
Receiving a preferred medication (% yes)	24	38.7%	18	36.7%	68	30.2%	60	25.8%	42	37.8%	128	27.9%	0.041
Cognition
Trails Difference (B-A)	62	45.7 (24.1)	49	47.4 (24.0)	225	49.0 (25.8)	234	48.6 (24.2)	111	46.4 (24.0)	459	48.8 (25.0)	0.370
Stroop Color-Word raw score	62	46.2 (8.7)	49	44.2 (8.7)	223	41.0 (9.9)	230	40.4 (10.5)	111	45.3 (8.7)	453	40.7 (10.2)	<.001
Stroop Interference raw score	62	4.3 (5.9)	49	2.6 (7.3)	223	0.9 (7.3)	230	0.8 (8.5)	111	3.6 (6.6)	453	0.9 (8.0)	0.001

^*Differences between age groups are assessed using chi-square test for categorical measures and t-test for continuous measures

Moderation model results for the 2-way interactions (age by baseline characteristic) are presented in Table 2. The 2-way interaction models also show estimates of the adjusted odds of relapse, with the effect of the baseline characteristic on relapse estimated separately among young adults and among older adults while controlling for treatment assignment. When assessing two-way interactions, there were no significant differential effects of baseline characteristics on relapse between age groups. Figure 1 depicts the odds ratios and confidence intervals for each baseline characteristic as a predictor of relapse, with side-by-side comparison of the age groups, ranked by magnitude of the aOR. Although none of the aORs are significant (95% confidence interval crossing 1), several have notable larger estimated values for young adults than for older adults, as well as estimated aORs in the opposite direction (e.g., smoking tobacco, heroin use, and anxiety/depression). Additionally, in all cases, the confidence intervals are wider for young adults. There were no significant 3-way interactions in regression moderation models for any of the baseline characteristics by age and treatment group (not shown).

Table 2.

Model results of logistic regression including 2-way age by baseline characteristic interaction estimating the odds of relapse. Within each model, the effect of the baseline characteristic on relapse is estimated separately among young adults and among older adults.

	Overall Effecta of 2-way Age*Baseline Measure Interaction		Effect of baseline measure among young adults (≤25 years)			Effect of baseline measure among older adults (>25 years)
Baseline characteristic measure	F(df)	p-value	aORb	95% CI	p-value	aORb	95% CI	p-value
Demographics
Gender (ref=Female)	F(1, 558)=0.29	0.592	1.24	(0.54, 2.86)	0.611	0.96	(0.62, 1.48)	0.855
Substance Use
Any current heroin use (ref=No)	F(1, 558)=0.54	0.461	2.78	(0.80, 9.63)	0.107	1.66	(0.93, 2.97)	0.084
Any current intravenous use (ref=No)	F(1, 558)=0.09	0.765	0.91	(0.34, 2.46)	0.855	1.07	(0.72, 1.60)	0.730
Cigarette smoker (ref=Non-smoker)	F(2, 552)=1.40	0.247
Every day			2.08	(0.27, 15.92)	0.478	0.64	(0.36, 1.14)	0.130
Some days			3.89	(0.31, 48.10)	0.290	0.41	(0.17, 0.98)	0.046
Any current cocaine/crack use (ref=No)	F(1, 558)=0.07	0.786	0.96	(0.41, 2.23)	0.917	0.84	(0.57, 1.24)	0.381
Any current cannabis use (ref=No)	F(1, 558)=0.17	0.683	1.06	(0.43, 2.62)	0.899	0.86	(0.59, 1.27)	0.455
Past Treatment History
Current treatment is first treatment in lifetime (ref=No)	F(1, 558)=0.11	0.740	1.13	(0.48, 2.65)	0.784	0.96	(0.64, 1.44)	0.848
Past methadone or buprenorphine treatment successful (ref=No)	F(1, 558)=0.45	0.504	0.92	(0.35, 2.41)	0.870	1.32	(0.88, 1.97)	0.181
Past naltrexone treatment successful	NA*
Opioid Withdrawal
Subjective Opioid Withdrawal Scale (range: 0–64)	F(1, 558)=0.13	0.719	1.14	(0.72, 1.80)	0.584	1.04	(0.85, 1.27)	0.718
Other Psychiatric Symptoms/Disorders
Hamilton Depression scale (ref=None)	F(2, 555)=0.07	0.929
Mild			1.19	(0.47, 3.00)	0.713	1.03	(0.66, 1.60)	0.912
Moderate/Severe			1.34	(0.31, 5.81)	0.694	1.05	(0.55, 1.99)	0.890
Anxiety-depression moderate or extreme (EuroQoL) (ref=None)	F(1, 558)=1.99	0.159	1.95	(0.72, 5.29)	0.191	0.90	(0.60, 1.36)	0.620
Suicidal thoughts (ref=No)	F(1, 558)=0.37	0.546	0.61	(0.13, 2.98)	0.545	1.07	(0.45, 2.53)	0.879
Social/Living Situation
Current homelessness (ref=No)	F(1, 558)=0.32	0.575	1.25	(0.44, 3.58)	0.678	0.90	(0.58, 1.40)	0.642
Any friends/family with heroin/other opioid problems (ref=No)	F(1, 553)=0.69	0.407	0.50	(0.18, 1.39)	0.185	0.80	(0.54, 1.18)	0.256
Legal Issues
Current probation/parole (ref=No)	F(1, 557)=0.01	0.933	1.68	(0.59, 4.78)	0.327	1.60	(0.91, 2.82)	0.103
Motivation for Participating/ Attitudes Regarding Medication
Participating to avoid relapse	NA*
Participating because of family/friends (ref=Disagree)	F(2, 555)=1.56	0.212
Agree			3.34	(0.70, 15.97)	0.131	0.90	(0.51, 1.59)	0.722
Neutral			0.63	(0.22, 1.81)	0.394	0.90	(0.56, 1.44)	0.651
Prefer to receive BUP-NX (ref=Disagree)	F(2, 555)=1.96	0.142
Agree			1.06	(0.30, 3.66)	0.932	1.27	(0.76, 2.11)	0.360
Neutral			0.37	(0.11, 1.19)	0.095	1.05	(0.65, 1.69)	0.833
Prefer to receive XR-NTX (ref=Disagree)	F(2, 555)=0.56	0.573
Agree			1.24	(0.38, 4.06)	0.725	0.72	(0.41, 1.24)	0.231
Neutral			0.67	(0.23, 1.93)	0.459	0.67	(0.41, 1.10)	0.116
Receiving a preferred medication (ref=No)	F(1, 557)=0.03	0.870	0.97	(0.42, 2.26)	0.944	0.90	(0.59, 1.37)	0.613
Cognition
Trails Difference (B-A)	F(1, 558)=0.09	0.763	1.01	(0.66, 1.54)	0.976	1.08	(0.89, 1.31)	0.432
Stroop Color-Word raw scorec	F(1, 552)=0.05	0.825	1.01	(0.62, 1.64)	0.966	1.07	(0.88, 1.30)	0.484
Stroop Interference raw scorec	F(1, 552)=0.07	0.792	1.04	(0.64, 1.71)	0.862	0.97	(0.81, 1.18)	0.780

^*NA indicates that model did not converge due to zero cells

^aOverall effect is a test of the type 3 overall effect of age by baseline measure adjusting for treatment, which tests whether the effect of the baseline measure on the odds of relapse significantly differs between young adults and older adults

^baOR; adjusting for treatment assignment

^cAnalyses using t-scores instead of raw scores for Stroop tests produced similar results

Fig 1.

Model estimated adjusted odds-ratios (aOR) and 95% confidence intervals, expressing the odds of Relapse as a function of baseline characteristics, comparing younger (N=111) vs older (N=459) age groups, from logistic regression models including 2- way age by baseline measure interactions. Baseline measures are sorted by highest aOR to lowest aOR within the younger group with multilevel variables tracked together

* ORs are presented on x-axis with a natural log scale.

Discussion

The parent XBOT study showed overall 6-month relapse rates that are roughly comparable to other studies in the literature across the range of available MOUDs, but relapse rates in the sub-group of young adults were considerably worse, also consistent with previous findings. We conducted a secondary analysis of a comparative effectiveness trial of medication treatment for opioid use disorder to evaluate age as a moderator of the relationship between baseline characteristics and relapse, as well as of possible differential response to treatment between the age groups. There were no significant moderator effects found in 3-way interactions between baseline characteristics, medication group and age group, nor any significant moderators in 2-way interactions between baseline characteristics and age group. In other words, younger versus older age group was not associated with differences in the effect of baseline characteristics on the risk of relapse between medication treatment (BUP-NX vs XR-NTX), nor collapsed across medications.

The result that no significantly different effects of baseline characteristics between age groups were identified, despite the striking difference in relapse rates between the age groups, 70.3% among young adults vs. 58.8% among older adults [9], suggests the conclusion that poorer treatment outcomes for young adults is likely associated with more complex interactions between multiple developmental vulnerabilities, rather than any single predominant factor. And important factors may not have been captured. As previously noted, the parent study assessments and measures were not originally designed to capture developmental or age differences. Notably absent, for example, were measures of impulsivity, or robust measures of peer influence or social support.

Several of the candidate baseline characteristic odds ratios had notably larger estimated point estimates of the associations with relapsefor young adults than for older adults, although confidence limits were wide and none of these odds ratios were statistically different between groups. Characteristics with notable non-significant differences include: current heroin use (aOR=2.78; 95% CI: [0.80, 9.63] for young adults compared to aOR=1.66 [0.93, 2.97] for older adults), cigarette use during some days (aOR: 3.89 [0.31, 48.10] vs 0.41 [0.17, 0.98]) and everyday (aOR: 2.08 [0.27, 15.92] vs 0.64 [0.36, 1.14]), participating because of family/friend (aOR: 3.34 [0.70, 15.97] vs 0.90 [0.51, 1.59]), moderate/extreme anxiety-depression (via EuroQoL scale aOR: 1.95 [0.72, 5.29] vs 0.90 [0.60, 1.36]), and friends/family with an opioid problem (aOR: 0.50 [0.18, 1.39] vs 0.80 [0.54, 1.18]). It is interesting to speculate whether these suggest themes that might be fruitful targets of future investigation. Wider confidence intervals among the young adults likely reflects the substantially smaller size of that group. Lack of statistical significance may have been due to lack of power, since although the total sample of young adults was not small, individual cell counts often contained very few subjects. It is possible that some of these candidate variables would reach significance in a sample that included more young adults. It is also possible that the chosen baseline characteristics did not adequately define or distinguish the age groups, or that the relatively small mean age difference between the groups (13.4 years) did not allow for sufficient separation of these characteristics (or may have included confounding “borderline” cases especially at the lower age margins of the older adults), even though the groups did separate in relapse outcomes.

Strengths of this study include: exploration of predictors of treatment outcome in the largest sample of young adults with OUD to date and the only exploration of moderators of comparative treatment effectiveness in young adults with OUD. Limitations include: secondary analysis, assessments and measurements not chosen in the parent study to elucidate the differences between age groups, possible lack of generalizability to other treatment contexts, very small size for several subgroup cells as subdivided for relapse status and presence/absence of baseline characteristics, and mismatch in sample sizes between age groups with probable lack of power to detect differences.

Future research should explore these notable characteristics and other candidates in larger samples of young adults, and should utilize assessments and measurements specifically designed to address hypotheses related to developmental vulnerability. Such variables could include measures of impulsiveness, emotion regulation, excitement-seeking [24], perceptions of harm [28], further exploration of sources of motivation [19], therapeutic alliance and treatment engagement [7, 8, 10], perceptions and use of social supports [11] [12], deviant peer affiliations, psychiatric co-morbidity [25], parental and other caregiver status. An area of particular interest would be more detailed exploration of cognitive differences in relation to treatment outcomes, especially in executive function [32] [33] [34] [35], since the developmental literature clearly points to immaturity in youth, some studies have found deficits in youth SUD populations [36] [37] [38], and some have shown associations with poorer treatment outcomes [39]. If the trends toward greater relapse among those in the younger group using heroin or smoking tobacco, or having high levels of depression/anxiety, are supported in larger samples, it might suggest that young people are more vulnerable to certain elements of severity and co-morbidity. If the trend for greater rates of relapse among those in the younger group endorsing being in treatment because of friends or family is supported in larger samples, it might suggest that the perception of external motivation is not as effective for young people, , or simply that they are a less motivated group. Clinical experience suggests that pressure from family is what often brings younger people into treatment, but it could be that perception of external motivation needs to be stronger in young people to be impactful. In these cases, continued extra external pressure or incentives to sustain treatment may be needed, or perhaps more focus on peer / family influences [29, 30, 31]. Additional areas for future research should include mediator analyses, trajectory analyses, and times-series analyses looking for within-treatment variables as predictors. It may also be useful to undertake other approaches to analyzing age, such as treating it as a continuous variable to maximize power, or trichotomizing the sample and examining the polar groups to maximize group age differences and eliminate “border” cases. Prospective studies of treatment-matching strategies and hypotheses, based on medication choice, help-seeking attitudes, illness severity, co-morbidity, and/or other clinical characteristics, should also be undertaken.

Poorer OUD treatment outcomes in young adults remains a significant problem for the addiction field, especially as so many youth are affected by the opioid epidemic, and as most individuals with OUD have onset in youth. Many aspects of developmental vulnerability likely confer risk, and are associated with problems in treatment engagement, medication adherence and treatment retention. It is essential that we continue our efforts to identify the particular barriers to success in this critical target population which may lead to design and implementation of interventions to overcome them.

Abbreviations:

BUP-NX

Buprenorphine naloxone

ITT

Intent to treat

MOUD

Medications for opioid use disorder

OUD

Opioid use disorder

SUD

Substance use disorder

XR-NTX

Extended release naltrexone