Men’s Misuse of Prescription Opioids from Early to Middle Adulthood: An Examination of Developmental and Concurrent Prediction Models

Abstract

Objective:

The prevalence of misuse of prescription opioids across adulthood and the associations of such misuse with symptoms of psychopathology and use of other substances were examined for an at-risk community sample of men.

Method:

For a longitudinal study of boys (N = 206) followed to adulthood, misuse of prescription opioids was assessed on 13 occasions from ages 20–21 years to 37–38 years. Prediction of misuse was examined from prospectively assessed risk factors in three models: (a) parental substance use during the men’s adolescence; (b) the men’s own risk behaviors in adolescence—delinquent behavior, depressive symptoms, and use of tobacco, alcohol, marijuana, and opioids; and (c) within- and between-individual effects of the men’s risk behaviors during adulthood.

Results:

Opioid misuse was reported by 29% of men. After accounting for effects of age and considered individually, parent marijuana use and all of the adolescent and adult risk factors (except adolescent depressive symptoms) were significant between-individual predictors of opioid misuse. Furthermore, within-individual prediction was significant for adult delinquency and alcohol use after accounting for increases in opioid misuse with age. When risk factors were tested simultaneously, men’s adult delinquency and use of marijuana and tobacco remained significant between-individual predictors, whereas no parental or adolescent risk factors remained significant in these models.

Conclusion:

Both adolescent and adult risk factors were examined that predicted adult opioid misuse. Preventing adolescent problem behavior and using such histories to inform screening for misuse risk in adulthood may reduce the burden of the opioid crisis.

Whereas the United States represents less than 5% of the world’s population, its residents consume about 80% of the global opioid supply (Manchikanti, Fellows, Ailinani, & Pampati, 2010). According to the Centers for Disease Control (CDC; 2018), drug overdoses killed over 60,000 Americans in 2016, 66% of which involved a prescription or illicit opioid. The CDC’s analysis showed that overall drug overdose death rates increased by 21.5% from 2015 to 2016, and the largest increase in opioid overdose deaths was in men aged 25–44 years. The National Institute of Drug Abuse refers to issues related to opioid abuse nationally as a crisis, and the CDC refers to overdoses as an epidemic. Levy, Breen, Lunstead, and Weitzman (2018) argue that, due to escalating opioid use, it is critical for prevention purposes to understand better the predictors of opioid use, prescription misuse, and opioid abuse. Despite the need to understand the emergence and course of prescription opioid misuse over time for community-based samples, and risk factors related to this, knowledge is generally limited to being based on cross-sectional surveys or more specialized studies—for example, of pain patients or of patients receiving treatment for abuse. Thus, there are substantial gaps in our knowledge of the etiology of misuse of prescription opioids.

The purpose of the present study of a community sample of American men was to increase our understanding of predictors of engagement in prescription opioid misuse (Jordan, Blackburn, Des Jarlais, and Hagan, 2017; also termed opioid misuse herein) across time in adulthood, and of the associations that prior and concurrent symptoms of psychopathology and use of other substances have with opioid misuse in adulthood. To focus on etiological issues related to prescription opioid misuse, the use of illicit opioids (e.g., heroin), which may involve a differing etiology, was excluded in the present study. Men’s reports of opioid misuse (assessed on 13 occasions from approximately ages 20–21 years to 37–38 years) were examined in relation to concurrently assessed risk factors as well as to those assessed prospectively from late childhood and adolescence.

The men in the Oregon Youth Study (OYS) were recruited at ages 9–10 years, along with their parents, and were considered at risk for delinquency due to the neighborhoods in which they lived. Recruitment involved whole classrooms of boys in these neighborhoods (Capaldi, Chamberlain, Fetrow, & Wilson, 1997). Due to this factor and the relatively high study retention (approximately 92% of living participants), including of participants with higher levels of problematic behavior (e.g., crime), relatively high levels of substance use were detected in adolescence and adulthood (Washburn & Capaldi, 2014). OYS men were primarily White and low income, two demographic factors associated with higher rates of opioid use in U.S. adults (Winkelman, Chang, & Binswanger, 2018). Furthermore, as discussed by Jordan et al., (2017), the prevalence of prescription opioid misuse rose considerably from 1994 to 2001 for both adolescents and young adults (ages 18 to 25 years). As OYS men were aged 25 years in approximately 1999–2000, they spent part of their young-adult years during this period of increase; thus, they are expected to have been similarly affected by the forces leading to this rise in misuse. According to Phillips, Ford, and Bonnie (2017), medical prescriptions for opioids started to increase sharply in the mid to late 1990s (National Institute on Drug Abuse, 2014) and nonmedical opioid use or abuse increased sharply subsequently, reaching a peak of 2.7 million new users in 2002 (Kolodny et al., 2015), declining to about 1.8 million new users in 2012 (Substance Abuse and Mental Health Services Administration, 2013). However, the overall number of people continuing to use nonmedically is large. The present study involves data regarding prescription opioid misuse through 2012 to 2013; therefore, our assessments during OYS men’s adult years coincide with the years of peak opioid abuse in the United States.

In the present study, prescription opioid misuse versus either nonuse or prescribed use was examined. Prescription opioid misuse is a significant risk for opioid use disorder, particularly after young adulthood. For most substances, abuse is higher in young adulthood than at older ages (e.g., both heavy episodic drinking and amount of marijuana used decreased across the 20s for the OYS men; Capaldi, Feingold, Kim, Yoerger, & Washburn; 2012; Washburn & Capaldi, 2015). Thus, it seemed possible that the prevalence of opioid misuse would be highest in the early 20s and would decrease across the late 20s and 30s. On the other hand, rates of opioid misuse might increase across these later years, as pain-related prescriptions would likely increase and such prescriptions could be an entry point for use that escalates to misuse (Volkow & McLellan, 2016). Regarding the association of prescription opioid misuse to opioid use disorder, Martins et al. (2017) report that the prevalence of opioid use disorder among individuals reporting past-year prescription opioid misuse was on an upward trajectory to approximately 22% for 26–34-year olds in 2014, but on decreasing or fluctuating trajectories for younger adults (approximately 15%) and adolescents (approximately 13%), respectively.

Risk Factors for Prescription Opioid Misuse

Theories of the development of problem behavior are likely to be relevant to understanding opioid misuse (and abuse). Such theory proceeds from the observation that seemingly different, problems of adolescence and early adulthood—such as delinquency, drug use, school failure, and sexual health risks—frequently co-occur and share some common underpinnings (Jessor, 1991). To aid in the conceptualization of the development of problem behaviors, we have posited a Dynamic Developmental Systems (DDS) approach whereby such behaviors are related to a continuous interplay or feedback among individual vulnerabilities (e.g., genetic, temperament), environmental factors (e.g., poverty, substance availability), and key social influences (e.g., from parents and peers). A focus of the DDS approach in understanding the development and course of substance use is the role of both general pathway risk (i.e., shared underpinnings of problem behavior) and outcome-specific (i.e., substance related) risk (Capaldi, Tiberio, & Kerr, 2018; Kendler, Gardner, & Dick, 2011; Zucker, Boyd, & Howard, 1994). We contend that findings regarding the etiology of use of substances other than opioids can inform our conceptual approach to understanding the etiology of opioid misuse.

It is well established that conduct problems in childhood are a general risk factor for early onset of alcohol, tobacco, and marijuana use (Dishion, Capaldi, & Yoerger, 1999; Zucker, 2008). Depressive symptoms are related to opioid use in adults (Sullivan, Edlund, Zhang, & Unützer, 2006) and may also be a developmental risk factor for substance use. For example, McCarty and colleagues (2013) found that the interaction between growth in depression and conduct disorder symptoms in early to midadolescence uniquely predicted later substance use impairment in Grade 12—in addition to main effects of each. Whether these general pathway risks also apply to adult opioid misuse requires study.

Beyond general risk related to psychopathology symptoms, there may be risks that are more specific to substance use or to opioid misuse in particular; notably parental substance use has been found to predict offspring substance use in adolescence, independent of offspring problem behavior (Nadel & Thornberry, 2017). Influential outcome-specific risk factors for opioid misuse may include parental use of the same substance, as parents may model, supply (knowingly or not), or otherwise transmit specific risk to their children. Indeed, many etiological studies of particular forms of substance use such as alcohol consider parental use of the same substance as an important risk factor (e.g., Bailey, Hill, Oesterle, & Hawkins, 2006; Chassin, Flora, & King, 2004; Kerr, Capaldi, Pears, & Owen, 2012). However, many youth and adults use and abuse multiple substances (Leatherdale, Hammond, & Ahmed, 2009; Stinson et al., 2006) and some parental risk factors may be relevant across drug classes. Thus, risk factors for misuse of a particular substance such as opioids may include parents’ use of a variety of other substances. Indeed, prior studies—including those that also report substance-specific effects—indicate use of a particular substance by the offspring may be predicted by parental use of other substances (Bailey et al., 2006). For example, in a study of OYS men’s offspring, maternal tobacco use predicted child age of onset of alcohol use over and above the effect of maternal alcohol use (Capaldi, Tiberio, Kerr, & Pears, 2016). Thus, in the present investigation, it is valuable to examine prediction to the men’s misuse of opioids from a variety of types of parental substance use, rather than just from parental opioid use. For this reason, the association of parental substance use during the men’s late childhood and adolescence with the men’s misuse of opioids across the early years of adulthood was examined.

Although there is evidence that parental substance use is particularly predictive of onset of substance use in adolescence, this risk may not extend to adulthood. Capaldi, Stoolmiller, Kim, and Yoerger (2009) found for the OYS men in adolescence that parental alcohol use was associated with initial levels of alcohol use at ages 11–14 years, but not with growth in use. Similarly, in a study predicting to use of marijuana and growth in such use across the high school years (ages 14 to 18 years), Washburn and Capaldi (2014) found that parental marijuana use predicted use versus nonuse, but did not predict to growth in use. Thus, in the present study, the prediction of opioid misuse from parental substance use was examined due to the importance of this risk factor to offspring substance use. It was expected however that parental risks from the men’s late childhood and adolescence might not be associated with the men’s misuse of opioids in adulthood, when other influences (e.g., from peers and partners) may become more prominent (Angulski, Armstrong, & Bouffard, 2018) and are more proximal.

Other risks for opioid misuse in adulthood are likely to involve individuals’ prior and concurrent behaviors that reflect both general developmental risk pathways and outcome-specific risks of using other substances. Highly relevant to such pathways, Winkelman et al. (2018) examined the associations of nicotine dependence and alcohol abuse and dependence, as well as criminal justice involvement, with varying levels of opioid use involvement using cross-sectional analysis for a national sample aged 18 to 64 years (the 2015–2016 National Survey on Drug Use and Health). The types of past-year opioid use included no use, prescription use, prescription misuse, opioid use disorder, and heroin use. Overall, in addition to demographic and health risks (chronic condition; disability), individuals who reported any type of opioid use were more likely to have a mental illness or co-occurring substance use. Those reporting prescription opioid misuse were about twice as likely to be dependent on nicotine and three times more likely to be diagnosed with alcohol dependence or abuse than those reporting no opioid use. Furthermore, the odds of criminal justice system involvement either in the past year or prior to the past year was close to four times higher for those who misused prescription opioids compared to those who did not use opioids in the past year. This differential was considerably decreased by controls for other study variables, including mental illness and substance use, but remained significant (also, of note, incarcerated adults were not interviewed). Thus, there is a sound basis for examining general and substance-specific risks for men’s prescription opioid misuse, but almost no information on how these outcomes unfold over time in community populations.

Study Questions

Given the limited knowledge regarding prescription opioid use in community-based samples, the prevalence of prescription opioid use and misuse from ages 20–21 to 37–38 years was described for the OYS men. Second, men’s misuse of opioids (vs. prescribed use or nonuse) was predicted in three separate models from (a) parental substance use during the men’s adolescence; (b) six of the men’s own risk behaviors in adolescence, including delinquent acts, depressive symptoms, use of marijuana, tobacco, alcohol, and opioid use; and (c) time-varying effects of the first five of these risk behaviors during adulthood. The associations of the risk factors with the outcomes first were examined individually, yielding unadjusted estimates. Second, given that all of the candidate risk factors tend to co-occur, risk factors were evaluated in multiple regression models within each developmental risk domain, providing adjusted estimates. It was hypothesized that the outcome-specific risk factors (i.e., parental, adolescent, and adult substance use) and the general developmental risk pathways of delinquency and depressive symptoms would be individually predictive of opioid misuse. In both adolescence and adulthood (the latter at both between- and within-individual levels), delinquency was expected to show a more robust association with opioid misuse than was depressive symptoms. At the between-individual level, the men’s use of substances in adolescence and adulthood were hypothesized to be positively associated with opioid misuse in adulthood and were expected to attenuate the effect of delinquency (and any effect of depressive symptoms); the specific pattern(s) of substance use that would be associated with opioid misuse was exploratory. Finally, it was not clear if at the within-individual level in adulthood the men’s opioid misuse would be associated with greater or lesser delinquent behaviors, depressive symptoms, or use of other substances; thus, no specific predictions were made regarding these issues.

Method

Participants

The OYS started in 1984 and ended in 2013. All families with fourth-grade boys in schools in higher-delinquency neighborhoods (determined by density of adolescent offenders residing in the area) in a medium-sized metropolitan area in the Pacific Northwest were eligible to participate (31 families were ineligible as they could not speak English or were planning to move out of state within 6 months). Families were recruited via an initial letter from the school asking them to withdraw their names if they did not want to be contacted by study staff (very few families withdrew their names). Families then received a letter announcing the study, a phone call to schedule a home visit, and a home visit to explain the study participation and answer questions. The recruitment rate of eligible families was 74% (N = 206; Capaldi & Patterson, 1987). The study involved 29 assessment waves that were yearly through ages 31–32 years, and with additional assessments at ages 35–36 years and 37–38 years. For the present study, parental and adolescent predictors were assessed at the earlier years of the study. The assessment of the men’s adult risk behaviors and opioid use and misuse occurred on 13 occasions across adulthood; yearly from ages 20–21 to 31–32 years, except for no assessment at ages 26–27 years (11 assessments), with two further assessments at ages 35–36, and 37–38 years (participation was 98% at age 20–21 years and 88% of living men at ages 37–38 years). Participants were primarily White (90%) and from lower- and working-class families (75%; Hollingshead, 1975). In the first year of the study, 33% of the families received welfare or food stamps. Regarding family structure, 40% of the families involved two biological parents, 25% were two parent including a stepparent, 30% were single-mother families, and 5% were single-father families. Regarding education, 17% of fathers and 8% of mothers were college graduates.

Procedures

OYS parents and boys/men completed in-person interviews and questionnaires. Adults provided informed consent and all procedures were approved by the Institutional Review Board of the Oregon Social Learning Center. Participants were compensated for their time.

Interviews and questionnaires.

The parent(s) and their sons were interviewed separately, with each interview lasting 45 minutes to 1 hour. The interviewers completed a ratings checklist after each interview.

Measures

Means of risk behaviors were calculated across adolescence or adulthood, serving as the level-2, between-subjects’ predictors in the models.

Parental substance use.

The parents of the boys responded to a questionnaire about their substance use (Oregon Social Learning Center, 1984) biannually at Waves 1, 3, 5, 7, and 9 (W1–W9; male’s ages 9–18 years). At W1, parents reported only on their own use; in the later four waves, they also reported on their partner’s substance use. Parent alcohol, marijuana, tobacco, and other drug use variables were calculated separately. Each parent variable was the mean of values calculated separately for mothers and fathers, which in turn were a mean of self- and partner reports (with the exception of W1) calculated at each wave. Parents’ alcohol use was a mean of measures of frequency, volume, and consequences of use at each wave. Marijuana use denoted frequency, ranging from never to nearly daily use across seven categories. Due to low endorsement, both opioid use (i.e., heroin and morphine) and other illicit drug use (e.g., cocaine, speed, LSD, mushrooms, and angel dust) denoted any such use by one or both parents (coded as yes = 1, no = 0). Finally, tobacco was measured as number of cigarettes used per day. Levels of substance use were relatively high for parents across the years of their son’s adolescence, at on average over 90% for maternal and paternal alcohol use, approximately 40–60% for marijuana use, 12–20% for other drug use, and 5–15% for opioid use.

Men’s adolescent and adult risk behaviors.

Delinquency, depressive symptoms, and use of alcohol, tobacco, and marijuana were assessed in adolescence starting at ages as described for each variable (below) through ages 19–20 years (i.e., up to 9 occasions) and in adulthood (ages 20–21 to 37–38 years; up to 13 occasions). Use of some opioids (heroin, opium) was assessed in adolescence. Use and misuse of prescription opioids was only assessed in adulthood. Assesssments of delinquency and substance use were by interview, and of depressive symptoms was by questionnaire.

Delinquency.

Starting at ages 12–13 years (W4), the boys/men completed a 36-item self-report delinquency schedule regarding engagement in various minor and major index crime activities (Elliot, 1983); thus, the scale was designed to cover a variety of criminal acts (e.g., stealing things worth $5 or less—with similar questions regarding stealing at other value ranges; purposely destroying or damaging property; buying, holding, or selling stolen goods; sold marijuana; stolen money from family members; hit others; attacked someone with the idea of seriously hurting them; burglarized residences or businesses, used force or threat of force in a robbery). Although the term delinquency is not typically applied to adults, this label was retained to reflect the consistency in measurement through adulthood. The 36 items were coded 1 if the behavior occurred or 0 if it did not occur. Cronbach’s standardized alphas (α) were acceptable, with a mean α of .82 (SD = .08, range .62 – .94) over all waves; across adolescent years, mean ([SD] α = .88 [.04], range .83 – .94); ages 20–38 years: mean ([SD] α = .78 [.08], range .62 – .90). Adolescent delinquency was calculated as the mean of the scale over ages 12–13 to 19–20 years (W4-11). All delinquency scales were log transformed to reduce skew.

Depressive symptoms.

Depressive symptoms. The CES-D Depression Questionnaire (Radloff, 1977) was administered at ages 14–15 years (W6) through ages 37–38 years. Depressive symptoms were the sum of the 20 CES-D items. Internal consistency was acceptable across waves (M [SD] α = .90 [.03], range .85 – .95).

Tobacco use.

Use of tobacco and other substances was assessed from ages 11–12 years onward. Boys/men provided self-reports of tobacco use that included the aggregate of cigarettes, pipes, and chewing tobacco (which were converted to mg doses of nicotine per week; see Capaldi et al., 2016).

Alcohol use.

The boys/men provided self-reports for beer, wine, and hard liquor regarding (a) any use in the past year and (b) for users, the number of times used (capped at 365 to reduce skew) and the amount consumed on a typical occasion (i.e., in units of < 1 drink, 1, 2, 3, 4 to 5, 6 or more). Units were equilibrated for alcohol content (Capaldi, Tiberio, Washburn, Yoerger, & Feingold, 2015). Volume of alcohol use was calculated from frequency of use multiplied by the usual amount consumed for beer, wine, and hard liquor, separately; the three values were summed to create the total yearly alcohol volume score.

Marijuana use.

Participants reported the number of times they used marijuana in the past year, and the quantity they usually used per occasion (in joints, tokes, bong hits, grams, and ounces). Units were converted to grams as follows: 1 joint = 1 gram; 1 toke or hit = 1/10 gram; 1 ounce = 28 grams. Estimated grams of marijuana used per year was calculated as the product of these two variables.

Adolescent opioid use.

In adolescence, the boys were queried regarding heroin and opium use, and a binary variable was created indicating endorsement of either kind of use at any wave through ages 19–20 years.

Adult opioid use and misuse.

In the present study, occasions of use of illicit heroin and opium in adulthood were excluded from the analyses to focus on misuse of prescription opioids. From ages 20–21 years on, men were asked about all substance type categories they had used in the previous year and were specifically asked about opioid use. Men were shown a card of common names of opioids and asked to list the types they had used. For each type of opioid used, the men were asked whether they always got the drug from a doctor and, if yes, did they ever take more than prescribed. Obtaining the opioids from a source other than a doctor or taking more than prescribed was considered prescription misuse. Men’s adult prescription opioid use was classified as nonuse, prescribed use, or misuse at each occasion.

Data Analytic Plan

The prevalence of prescription opioid use and misuse at any time in adulthood and the prevalence by age was examined. Model testing proceeded in several steps. First, the association of men’s likelihood of misusing prescription opioids across adulthood (i.e., outcome with age) was examined using hierarchical linear modeling with a logit link to account for the binary outcome (HLM; Raudenbush & Bryk, 2002). This included fitting an unconditional linear model with only age as a level-1 (time-varying) predictor and allowing for fixed intercept and slope terms. In addition, a random intercept term was estimated to test for significant heterogeneity in men’s likelihood of opioid misuse versus prescription use or no use at an average age of 27.7 years (intercept). If significant, the random intercept was retained, otherwise a fixed-effects only model was estimated within an HLM framework to correctly estimate the level-1 standard errors. All developmental risk factors models controlled for men’s age as a time-varying predictor.

To test the hypotheses, the parental, adolescent, and adult risk factors were added as predictors of men’s likelihood of misusing prescription opioids across adulthood at the within or between level. This allowed for the estimation of (a) time-varying (level-1) effects (within individuals across time) for the adult risk factors models—denoting whether the years in which men had higher-than-average risk were associated with concurrent increases in the likelihood of misusing opioids (vs. using as prescribed or no use) relative to years of lower than average risk—and (b) time-invariant (level-2, between-individual) effects for all models—denoting for example, whether men who had used more alcohol in adolescence were more likely to misuse opioids in adulthood relative to men who had used less alcohol in adolescence. To estimate within-individual effects, all time-varying level-1 predictors were group-mean centered around the level-2 means (Raudenbush & Bryk, 2002), which served as the level-2 predictors and were formed by mean aggregation of the variables across adolescence (from ages 11–12 or 14–15 years depending on the measure through ages 19–20 years, i.e., up to 9 occasions) or in adulthood (ages 20–21 to 37–38 years; up to 13 occasions).

Models were estimated in Mplus version 7.4 (Muthén & Muthén, 1998–2012) for the binary outcome of opioid misuse versus non-misuse (prescribed use or nonuse). First, associations between each of the parental, adolescent and adult risk factors and men’s likelihood of endorsing opioid misuse were examined, yielding unadjusted estimates (i.e., ignoring other risk factors except age). Second, within each developmental domain (i.e., parent, adolescent, and adult risk factors) risk factors were considered simultaneously, yielding adjusted estimates within domain (e.g., men’s estimated odds for misusing opioids given greater vs. less alcohol use across adulthood, while controlling for delinquent behavior, depressive symptoms, and tobacco and marijuana use in adulthood, as well as for age). Correlations among the predictor variables within the three developmental domains were also examined. To facilitate interpretation and estimate unadjusted and adjusted effect sizes, results are presented as fully standardized regression coefficients; for multiple regression models, R² are presented. Note that coefficients were standardized such that the variance of the outcome was estimated as a latent continuous variable underlying the observed dichotomous outcome (Winship & Mare, 1984).

Results

Prevalence of Opioid Misuse and Prescribed Use

Of the 206 men, 31.1% (N = 64) did not report use (either misuse or prescribed use) in any of the 19 years of adulthood studied (based on up to 13 assessments); 39.8% (N = 82) reported only prescribed use; and 29.1% (N = 60) reported opioid misuse at some time (possibly in addition to prescribed use). Of those N = 142 men who reported any use (either misuse or prescribed use), 32.4%, 21.8%, 14.1%, 13.4%, and 18.3% reported use at 1, 2, 3, 4, or 5 or more assessments, respectively. Of the N = 82 men who reported only prescribed opioid use, 43.9%, 26.8%, 12.2%, and 17.1% reported use at 1, 2, 3, or 4 or more assessments, respectively. Finally, of the N = 60 men who reported opioid misuse, 48.3%, 18.3%, 16.7%, and 16.7% reported doing so at 1, 2, 3, and 4 or more assessments, respectively. Note that of the N = 60 men who misused prescription opioids, 58.3% also reported prescribed use at one or more occasions; N = 41 of whom reported both misuse and prescribed use in the same year (coded as misuse in these years in the analyses).

Figure 1 depicts the proportion of the men who reported (a) any prescribed opioid use, or (b) any misuse of prescription opioids in the past 12 months at each age range. Note that rates looked relatively similar when the small number of observations from when men reported any heroin or opium use in the past year (about 1% across the entire adult-assessment period [31 of 2517 total person-by-time observations]) were included (not shown). The proportion of men who misused opioids rose to 10% of the sample at ages 27–28 years and then varied from around 6–10% at each assessment through ages 37–38 years.

Figure 1.

Sample Prevalence of men’s prescription opioid use and misuse by assessment age (20-38 years) and year (1996-2013).

Predictions to Men’s Opioid Misuse across Adulthood

Due to space limitations, we limited presentation of the multiple regression results to the level-1 and level-2 standardized regression coefficients and the proportion of variance (model estimated R²) in men’s opioid misuse explained by the risk factors. Full model results, including fit statistics, are available from the authors upon request.

Prediction from age.

First, an unconditional model was estimated containing only age as a level-1 (time-varying) predictor of men’s opioid misuse. Age was group-mean centered (average age of 27.7 years). As results indicated that a random intercept term (denoting variation in men’s likelihood of opioid misuse vs. non-misuse at age 27.7 years) was nonsignificant b[se] = .004[.005], p = .421), a fixed-effects only model was estimated. Men’s risk for misuse versus non-misuse of opioids increased across adulthood; on average, men were 1.12 times significantly more likely (p < .001) to misuse opioids across each year of adulthood. Age accounted for 9% of the within-level variance in men’s opioid misuse (R²[SE] = .09[.03], p = .003). All subsequent models controlled for the linear trends in opioid misuse across adulthood by including age as time-varying (level-1) predictor.

Prediction from parental substance use.

Associations of each type of parental substance use with the men’s opioid misuse across ages 20–21 years to 37–38 years was examined. When considered individually (Table 1, column 1), parental marijuana use was the only significant predictor of men’s opioid misuse. This effect was small and was attenuated after accounting for parents’ other substance use (Table 1, column 2). The parental risk factors were all significantly associated (Table 1, columns 3–5) and together only accounted for 5% of the between-level variance in men’s opioid misuse (R²[SE] = .05[.04], p = .238).

Table 1.

Between-Level Associations of Men’s Adult Opioid Misuse and Parental Risk Factors

	Prediction to men’s adult opioid misuse		Bivariate associations among parental risk factors r(se)
	Unadjusted	Adjusted	Alcohol	Marijuana	Opioids
Parental	B(se)	B(se)
Alcohol	.05(.09)	.00 (.09)	--
Marijuana	.15(.07)*	.14 (.11)	.29(.05)***	--
Opioids	−.06(.09)	−.20 (.12)	.14(.07)*	.41(.07)***	--
Other drugs	.13(.10)	.15 (.15)	.21(.07)**	.63(.04)***	.58(.05)***

Note. Tabled numbers denote standardized estimates.

^***p < .001.

^**p < .01.

^*p < .05.

Prediction from men’s adolescent behaviors.

Shown in Table 2 are the unadjusted (column 1) and adjusted (column 2) associations of men’s adolescent delinquency, depressive symptoms, and substance use (ages 11–12 to 19–20 years) with their opioid misuse versus non-misuse in adulthood. In the unadjusted models (which accounted for age only), men who had had higher levels of problems in each area in adolescence (except depressive symptoms) were significantly more likely to misuse opioids in adulthood than those who had had lower levels. Adolescent delinquency, alcohol use, and marijuana use showed the strongest associations with adult opioid misuse. In the multiple regression model, however, there were no significant independent predictors of men’s opioid misuse. Of note, most of the adolescent risk factors were significantly associated with one another (except for adolescent opioid use and depressive symptoms; Table 2, columns 3–7) and adolescent delinquency, alcohol use, and marijuana use were especially closely related (r = .55 to .66). The adolescent risk factors accounted for a quarter of the between-level variance in men’s opioid misuse (R²[SE] = .25[.08], p < .001).

Table 2.

Between-Level Associations of Men’s Adult Opioid Misuse and Adolescent Risk Factor

	Prediction to adult opioid misuse		Bivariate associations among adolescent risk factors r(se)
Adolescent	Unadjusted B(se)	Adjusted B(se)	Delinquency	Depressive Symptoms	Tobacco	Alcohol	Marijuana
Delinquency	.44(.08)***	.30 (.15)	--
Depressive symptoms	.01(.11)	−.12 (.11)	.35(.06)***	--
Tobacco	.19(.09)*	−.12 (.11)	.52(.05)***	.22(.07)**	--
Alcohol	.45(.09)***	.24 (.13)	.66(.04)***	.15(.07)*	.48(.05)***	--
Marijuana	.35(.07)***	.11 (.10)	.60(.04)***	.18(.06)**	.44(.06)***	.55(.04)***	--
Opioids	.19(.08)*	.04 (.08)	.32(.06)***	.12(.07)	.26(.07)***	.26(.06)***	.40(.08)***

Note. Tabled numbers denote standardized estimates.

^***p < .001.

^**p < .01.

^*p < .05.

Prediction from men’s adult behaviors.

Shown in Table 3 are the associations of the measures of delinquency, depressive symptoms, and substance use during men’s adulthood with their misuse versus non-misuse of opioids across adulthood examined as both time-invariant (level-2, between individuals; Panel A) and time-variant (level-1, within individuals across time, controlling for age; Panel B) effects.

Table 3.

Between-Level (Panel A) and Within-Level (Panel B) Associations of the Men’s Adult Opioid Misuse and Adult Risk Factors

Panel A
	Between-level prediction to adult opioid misuse		Bivariate associations among adult risk factors between men r(se)
Adult	Unadjusted B(se)	Adjusted B(se)	Delinquency	Depressive symptoms	Tobacco	Alcohol
Delinquency	.60 (.06)***	.43 (.09)***	--
Depressive Symptoms	.20 (.09)*	–.08 (.10)	.36(.06)***	--
Tobacco	.23 (.09)*	.17 (.08)*	.16(.07)*	.26(.06)***	--
Alcohol	.44 (.09)***	.15 (.09)	.47(.05)***	.06(.07)	.10(.07)	--
Marijuana	.51 (.06)***	.31 (.07)***	.44(.05)***	.24(.07)***	.16(.07)*	.26(.06)***
Note. Tabled numbers denote standardized estimates. *** p < .001. ** p < .01. * p < .05.
Panel B
	Within-level prediction to adult opioid misuse		Bivariate associations among adult risk factors within men across adulthood r(se)
Adult	Adjusted for age B(se)	Adjusted for age and other risk factors B(se)	Delinquency	Depressive symptoms	Tobacco	Alcohol	Marijuana
Delinquency	.14 (.05)**	.11 (.06)	--
Depressive symptoms	−.004 (.04)	−.03 (.04)	.10(.03)***	--
Tobacco	.04 (.06)	.01 (.06)	.08(.03)**	.04(.03)	--
Alcohol	.14 (.06)*	.10 (.07)	.28(.02)***	.04(.03)	.14(.03)***	--
Marijuana	.08 (.05)	.04 (.05)	.24(.03)***	.08(.03)**	.15(.03)***	.18(.04)***	--
Age	.32 (.05)***	.38 (.05)***	−.36(.03)***	−.01(.03)	−.10(.04)**	−.13(.04)***	−.14(.03)***

Note. Tabled numbers denote standardized estimates.

^***p < .001.

^**p < .01.

^*p < .05

First, for unadjusted predictions to between-individual effects (Table 3, Panel A, column 1), all five predictors showed significant effects. In the model adjusting for all predictors (Table 3, Panel A, column 2), between-level effects of delinquency, tobacco use, and marijuana use in adulthood remained significant predictors of adult opioid misuse versus non-misuse, whereas the effects of alcohol use and depressive symptoms were attenuated. All of the between-level adult risk factors were significantly associated, except for alcohol use with tobacco use and depressive symptoms (Table 3, Panel A, columns 3–6). The adult risk factors accounted for more than half of the between-level variance in men’s opioid misuse (R²[SE] = .53[.08], p < .001).

Unadjusted prediction at the within-person level (Table 3, Panel B, column 1) indicated that age showed a significant, positive association with opioid misuse versus non-misuse. In addition, both delinquency and alcohol use were significantly associated with opioid misuse within individuals; that is, in the years that men showed higher levels of delinquency and alcohol use they were also more likely to misuse opioids. In the model for misuse versus non-misuse adjusting for other predictors (Table 3, Panel B, column 2), however, age was the only significant predictor. Finally, all of the within-level adult risk factors were significantly associated with one another except for depressive symptoms, which was significantly related only to marijuana use (Table 3, Panel A, columns 3–6). On average, men showed significant declines in all of the adult risk factors across adulthood except for depressive symptoms. Age and the adult risk factors accounted for 13% of the within-level variance in men’s opioid misuse (R²(SE) = .13(.04), p < .001).

Discussion

The prevalence of prescription opioid misuse was examined from the early 20s to late 30s for an at-risk community sample of men who had been involved in a longitudinal study since ages 9–10 years. Then factors hypothesized to be associated with an increased risk of opioid misuse during adulthood were examined, with a focus on risk from parental substance use during the men’s adolescence and from the men’s own substance use and psychopathology symptoms during their adolescent and adult years. Both between- and within-individual effects of adulthood risk factors were examined. A substantial proportion of the men (29.1%) misused prescription opioids at some time in adulthood. As hypothesized, opioid misuse in adulthood was predicted from problem behaviors in adolescence—including delinquency, alcohol use, and marijuana use—and was linked to a picture of ongoing antisocial behavior and substance use in adulthood.

Winkelman and colleagues (2018) found for a national sample aged 18–64 years that 63.2% reported no opioid use in the past year, 31.3% reported prescription opioid use, and 5.5% reported opioid misuse, with men being somewhat more likely to misuse opioids than women. Reports of yearly opioid use for the OYS men were between about 20% and 30% from the mid 20s to late 30s, thus rather lower than the findings of the national study. However, it is notable that the national study spanned much older ages, included men and women, and was conducted in 2015–2016 when rates nationally had risen from the years covered in the present study for the OYS men. The proportion of prescription opioid use that was misuse, however, was considerably higher in the present study than in the Winkelman and colleagues’ study—at approximately 35% of users on average in the present study compared with 14% (including prescription misuse and opioid use disorder). The interview response rate for the national study was just under 70%, and it is possible that individuals with problematic drug use were less likely to participate. In addition, findings of that study indicated that lower levels of income and education were associated with a greater likelihood of opioid misuse, as was being White; each characteristics of the present sample.

The longitudinal design of the present study offers an additional viewpoint on the men’s prevalence of opioid use (69%) and misuse (29%) at some time across early to midadulthood, suggesting that a relatively large proportion of men from lower income and at-risk backgrounds have problems with opioid misuse at some time in these years. Furthermore, just over one half of the men who reported misuse did so at two or more occasions across early adulthood, suggesting persistent problems with opioid misuse for some men. Regarding comparability and generalizability of the present study findings, in general, findings of the OYS regarding substance use and related behaviors (e.g., crime) have shown strong comparability to studies with representative samples at similar ages (Capaldi et al., 2009; Washburn & Capaldi, 2014).

Regarding early precursors of risk, parental substance use during the men’s late childhood and adolescence accounted for little of the variance (5%) in men’s opioid misuse; parental marijuana use, although significant in unadjusted analyses, was not independently predictive. Parents’ opioid use in this period also was not associated with their sons’ adult opioid misuse, however, parental use of opioids was relatively low, at 5% to 15% of the parents across waves, which may relate to the lack of association. Although these null or weak effects seem rather surprising, prior OYS studies also have shown that, whereas parental substance use was associated with initiation of substance use in adolescence, other factors were related to subsequent growth in use or the course of use. Taken together with the present findings that men’s substance use in adolescence did predict their later opioid misuse, results may suggest that after substance use is initiated it is other factors, including the use of particular substances and associated problem behaviors, that place men at risk for opioid misuse in adulthood.

The adolescent prediction model indicated the role of both the general pathway risk factor of delinquent (externalizing) behavior and of adolescent engagement in use of substances including opioid use. Adolescent delinquency and substance use (including tobacco, alcohol, marijuana, and opioid use) each showed associations with opioid misuse in adulthood. Although none of these predictors were individually significant in the adjusted model, as a block they explained significant variance (25%) in men’s later opioid misuse. The fact that several associated variables were competing for variance in the multiple regression model, plus the relatively low sample size, may have reduced the predictive power of each individual substance.

Regarding risk for opioid misuse from prior substance use, there is physiological evidence that use of other substances may increase vulnerability to opioid use and misuse. As described by Cruz, Bajo, Schweitzer, and Roberto (2008), most drugs interact with brain receptors (or binding molecules) for specific neurotransmitters and either block or facilitate binding at these receptors. As cannabinoid and opiate substances occur in normal physiology for example, there are receptors which are specific to these substances when exogenous. Alcohol, however, influences the activity of many transmitter systems including endogenous opioids. As discussed by Mendez and Morales-Mulia (2008), ethanol may alter opioidergic transmission at a number of different levels. They present evidence of the role of two kinds of opioid receptors (mu and delta) in ethanol reinforcement and high levels of drinking. This suggests the possibility that heavier drinkers may be susceptible to the reinforcing properties of opioids, which affect the same receptors and systems. Other evidence may implicate prior marijuana use as a risk factor for opioid misuse. For example, although there are cannabinoid-specific receptors, Ghozland and colleagues (Ghozland et al., 2002) found that motivational effects of cannabinoids are mediated by two brain opioid receptors. Thus, these findings indicate that brain responses to substances are relatively complex and there may be cross-over effects from differing domains of substances on differing types of receptors. A further possibility is that there are individual differences in vulnerability of brain reward systems to substance use even prior to any use, relating to vulnerability to alcohol, marijuana, tobacco, and opioid use.

Next, we discuss key findings from the adult prediction model. At the between-individual level, all predictors including depressive symptoms were associated with opioid misuse, but only men’s delinquency and use of marijuana and tobacco were associated with men’s misuse of prescription opioids in the fully adjusted model. Thus, unlike in the adolescent risk factor models, delinquency, marijuana use, and tobacco use each persisted as unique correlates of adult opioid misuse in the adjusted model. This suggests that use of other substances relates to misuse of opioids over and above general pathway risk related to delinquency. It is notable that over one half of the between-individual variance in men’s adult opioid misuse was explained by the concurrent substance use and psychopathology variables considered here.

At the within-individual level, men’s delinquency and alcohol use were associated with their opioid misuse versus prescribed use or nonuse. Thus, men were more likely to misuse opioids in years when they were committing crimes and also drinking more alcohol. This suggests that misuse of opioids was not a substitute for using alcohol but was more likely at periods when the men were vulnerable to higher alcohol use. However, neither alcohol use nor delinquency remained significant in the model adjusted for age and other predictor effects.

Interestingly, adult deviant behavior in the form of delinquent or criminal acts appeared to be closely associated with opioid misuse in adulthood. Adolescent delinquency also showed associations with the adult opioid misuse. We have conceptualized externalizing behavior as a key aspect of general developmental risk in the sense that it is associated with risk for a number of problematic outcomes, including for substance use (Dishion & Patterson, 2006). There are a number of reasons why externalizing behavior is associated with risk for substance use, including underlying poor inhibitory control (Gottfredson & Hirschi, 1990) and association with peers involved in antisocial behavior and substance use (Dishion & Patterson, 2006), including access to suppliers of drugs. Regarding ongoing associations of crime with opioid misuse in adulthood at both the between- and within-individual levels, criminal acts are more deviant in adulthood than in adolescence in the sense that such acts peak in adolescence, at around ages 15–19 years, and then decline rapidly across the adult years (Blumstein, Cohen, Roth, & Visher, 1986). Persistence in deviancy and criminal involvement indicates that those men are not making some of the normative improvements of early adulthood, such as fewer gains in inhibitory control, which usually increases across this period with brain maturation (Luna, Padmanabhan, & O’Hearn, 2010). Continued involvement (or new involvement) in crime in adulthood is also associated with poor adjustment, including poorer employment histories (Wiesner, Capaldi, & Kim, 2011) and disrupted personal relationships, including separations from intimate partners and children (Theobald & Farrington, 2013). These factors may make the men more vulnerable to opioid misuse. A further important consideration is that the men’s involvement in the deviant social networks associated with criminal behavior in adulthood likely gives them greater access to illicit sources of prescription opioids.

Also important were a set of null effects: that in this at-risk sample, at least, depressive symptoms did not play a major role in future or concurrent misuse of opioids. In prior studies, associations have been found between depression or depressive symptoms and opioid use or abuse (Sullivan et al., 2006). It was suggested in Scherrer and colleagues’ (2014) study that use of prescription opioids increases the risk for depression, although they used a retrospective design. A number of prior studies have not taken into account the co-occurrence of depressive symptoms with other risk factors for opioid abuse, including antisocial behavior or crime and use of other substances. In the present study, the men’s depressive symptoms in adulthood were associated with opioid misuse at the between level in the unadjusted model, but not in the adjusted model or at the within level, and furthermore their adolescent depressive symptoms did not predict adult opioid misuse. Thus, overall, we found very limited support for the notion that the men misused prescription opioids in order to alleviate depressive symptoms or to self-medicate. This in turn means we did not find evidence that prevention efforts targeted at men’s depressive symptoms, particularly in adolescence, would be likely to result in substantial reductions in misuse of prescription opiates. However, further research is needed regarding the role of depressive symptoms in opioid misuse, including in women.

Study Strengths and Limitations

Strengths of the present study include the use of a prospective data set involving a well-maintained sample of at-risk boys/men, with regular assessments over approximately a 30-year period. This design meant that behavioral predictors were measured at numerous years during the men’s development, rather than at one point in time, and the men’s opioid use and misuse was examined during a 19-year period of early adulthood. There are however a number of study limitations. The OYS is a smaller community at-risk sample from the Pacific Northwestern United States, involving primarily white non-Latino individuals; thus, generalizability of findings to other areas and ethnic and age groups may be limited. The limited sample size, particularly for examining a low prevalence behavior such as opioid misuse, led to low power for detecting effects. Importantly, as the study only involved men’s opioid misuse, similar studies of women’s misuse would be valuable. Additionally, men’s prevalence of opioid use and misuse at one or more occasions across early adulthood may have been underestimated given that assessments did not occur during some years of the adult period of interest.

Prevention Implications

Overall, findings of the present study indicate that prescription opioid misuse prevention efforts, at least for men, should focus on reducing involvement in delinquent behavior and use of substances in adolescence, particularly alcohol, marijuana, and opioid use. Efforts in adulthood should focus on promoting desistance in these same areas, as well as on prevention of late-starting or adult-onset antisocial behaviors. In addition, there should be educational and prevention efforts with alcohol and marijuana users to address their vulnerability to opioid misuse. For example, additional follow-up efforts focused on preventing prescription misuse could be focused on pain patients who are alcohol or marijuana users, particularly those who use frequently or at high levels. Furthermore, findings suggest that prevention efforts in adulthood tailored for men who have been involved in the criminal justice system, or who self-report criminal acts, could be beneficial.

In conclusion, findings from the study indicated that men who lived in neighborhoods with relatively high levels of delinquency in childhood, which tended to be lower-income neighborhoods, show relatively high levels of misuse of prescription opioids across the earlier years of adulthood, namely ages 20–21 to 37–38 years. Overall findings indicated that whereas parental substance use overall showed only a small association with men’s misuse of opioids (at 5%), the men’s problem behaviors and substance use during adolescence explained considerably more variance (at 25%), and such behaviors during adulthood explained over half (53%) of the variance in men’s misuse of prescription opioids. Criminal acts and use of alcohol and marijuana, in particular, were associated with opioid misuse. Thus, prevention programs for men should be targeted particularly to substance users and those involved in criminal behavior.

Public Health Significance: Misuse of prescription opioids is contributing to a public health crisis. There is little longitudinal data on predictors of such opioid misuse. This study examined prediction from parental substance use, and from adolescent, and adult substance use and psychopathology symptoms for a sample of men at-risk for substance use. Findings indicated that the men’s adolescent and adult criminal acts and use of alcohol and marijuana, in particular, were associated with opioid misuse. Thus, prevention programs for men should be targeted particularly to substance users and those involved in criminal behavior.