The association between cannabis use and risk of non-medical pain reliever misuse onset among young adults in a legal cannabis context

Abstract

BACKGROUND:

Little is known about the prospective relationship between cannabis use and pain reliever misuse. This study examined associations of non-medical and medical cannabis use with onset of non-medical pain reliever misuse among young adults in Washington State (WA), where non-medical cannabis is legal.

METHODS:

Data were from a cohort-sequential study of adults 18–25 residing in WA. Four annual surveys were used from cohorts recruited in 2014, 2015, and 2016. Participants who had not reported non-medical pain reliever misuse at baseline were included in discrete time survival analyses (N = 4,236). Odds ratios (ORs) were estimated for new onset of non-medical pain reliever misuse in any given follow-up year over the course of three years according to baseline non-medical and medical cannabis use.

RESULTS:

When included separately in models, non-medical and medical cannabis use at baseline were associated with increased risk of non-medical pain reliever misuse adjusting for demographic characteristics as well as past year cigarette use and alcohol use (non-medical OR = 5.27; 95% CI: 3.28, 8.48; medical OR = 2.21; 95% CI: 1.39, 3.52). Including both forms of use in the model, associations of non-medical and medical cannabis use with non-medical pain reliever misuse onset remained (non-medical OR = 4.64; 95% CI: 2.88, 7.49; medical OR = 1.65; 95% CI: 1.04, 2.62).

CONCLUSIONS:

Despite claims that cannabis use may reduce opioid use and related harms, findings suggest that cannabis use, including medical use, may not be protective, but instead may increase risk for non-medical pain reliever misuse.

1. INTRODUCTION

In recent years, proponents of cannabis legalization have suggested that cannabis use may mitigate or reduce population-level opioid misuse, dependency, and overdose.^1,2 Reviews and meta-analyses have concluded cannabis use can produce modest reductions for some forms of pain,^3–5 and some studies suggest concurrent use of cannabis and opioids can reduce opioid dose requirements.^6–8 Further, ecological studies (where data are aggregated to and analyzed at a higher level such as the state) have reported reductions in rates of opioid overdose and opioid-related hospitalizations following state legalization of cannabis.⁹ In one frequently-cited state-level ecological study, medical cannabis legalization was associated with a 25% reduction in opioid overdose mortality from 1999 to 2010.¹⁰ However, subsequent state-level studies have found conflicting results. One study extending the analysis period to 2013 showed no effect of state medical cannabis legalization on opioid-related mortality,¹¹ and two other studies using data from later time-periods (2011–2014 and 1999–2017) showed medical cannabis legalization was associated with an over 20% increase in opioid overdose mortality.^12,13 Discrepancy in findings may be due to unmeasured state time-varying factors, including opioid policies implemented during this period (e.g., increase access to opioid dependence treatment) and heterogeneity and changes in cannabis regulations.¹⁴

Because of the potential for ecological fallacy where inference from ecological-level (e.g., state-level) studies is incorrectly applied to the individual-level,^12,15,16 longitudinal research where data are collected and analyzed at the individual-level is necessary to better understand relationships between cannabis use and risk of opioid misuse and related harms. Non-medical pain reliever misuse—use of prescription pain relievers in ways other than prescribed including for non-medical purposes—is an important antecedent of more problematic opioid use and overdose. Thus, as a key step toward other forms of opioid use and related harms, understanding factors, such as cannabis use, that may increase risk of onset of non-medical pain reliever misuse is critical.^17–19 Although prospective research is sparse, one study using nationally representative data and another using a community sample showed associations of non-medical cannabis use with increased likelihood of prescription pain reliever misuse and onset.^20,21 Additional longitudinal studies that investigate the relationship among high-risk groups such as young adults will be important. Compared to other age groups, young adults show the highest prevalence of cannabis use as well as prescription pain reliever misuse.²² Cannabis use has also increased more sharply over the past decade for young adults compared to other age groups.²³ Further, because those using cannabis for medical purposes may have conditions which predispose them to consider misuse of pain relievers,²⁰ studies that can assess both non-medical and medical cannabis use as predictors could be useful to examine their unique associations. Finally, because of claims of benefits of cannabis legalization in reducing population-wide opioid-related harms, it will be important to assess the cannabis and pain reliever misuse relationship in a legal cannabis context. Using data from a longitudinal study of young adults from Washington State (WA), where non-medical cannabis is legal, this study aimed to examine whether non-medical and medical cannabis use are associated with onset of non-medical pain reliever misuse across a three-year period.

2. MATERIAL AND METHODS

2.1. Study Population and Design

Data were from the Young Adult Health Survey (YAHS), a study contracted by the WA Division of Behavioral Health and Recovery to document trends in cannabis and other substance use among young adults. The YAHS utilized a cohort-sequential design whereby a new cohort of approximately 2,000 young adults was recruited each year starting in 2014 and surveyed annually. Young adults 18 to 25 living in WA were recruited through a mailed invitation or from online social media (e.g., Instagram, Craigslist). Additional details on recruitment and data collection are available elsewhere.²⁴

For those responding to letters or online advertisements, eligibility and identity were verified by study staff through a follow-up telephone call. Inclusion criteria included being 18 to 25 years-old and residence in WA. If participants provided consent through a follow-up telephone call, they were directed to the online baseline survey. Each year following, participants were invited to complete online follow-up surveys. Survey completion required approximately 20 minutes, and participants received a $10 e-gift card as compensation. This study used data from baseline and 1-, 2-, and 3-year follow-up surveys from cohorts enrolled in 2014 (n = 2101), 2015 (n = 1677), and 2016 (n = 2611). As context, cannabis for medical use for those with certain conditions was first legalized in 1998, cannabis for non-medical use was legalized in 2012 for individuals ages 21 and older, and the first legal retail cannabis outlets opened in July 2014. The majority of 2014 surveys were conducted prior to July. Results from sensitivity analyses excluding the cohort enrolled in 2014 were similar (see Supplemental Table 1) and, thus, results pooled across survey years are shown.

All measures and procedures were reviewed and approved by the University of Washington Institutional Review Board.

2.2. Measures

Participants were asked at each wave, “How often did you use the following substances (in any form) during the past 12 months?,” to assess use of various substances. Options included “pain relievers to get high” (which we refer to as non-medical pain reliever misuse and consider a proxy for non-medical prescription opioid misuse), “cannabis/marijuana for recreational purposes” (which we refer to as non-medical cannabis use), “cannabis/marijuana for medical purposes,” alcohol, and cigarettes. Ordinal response options ranged from 0 (“never”) to 9 (“every day”). For analyses and ease of interpretation, all substance variables, except alcohol use, were dichotomized as any (1) or no use (0) in the past year. Because of its high prevalence and common co-occurrence with cannabis use,²⁵ alcohol use was categorized as at least weekly use (1) and less than weekly use (0).

Various baseline demographic characteristics were used as covariates in statistical models: sex assigned at birth, race/ethnicity (categorized as non-Hispanic White [reference], non-Hispanic Asian, non-Hispanic other race, and Hispanic any race), geographic region of state (East, Northwest, Southwest), age in years, 4-year college status, and full-time employment status. Sex, race/ethnicity, and geographic region of the state were included because they were incorporated as variables in post-stratification weights. Other covariates were included a priori because of theorized associations with cannabis and pain reliever misuse.^26–28

2.3. Analytic Plan

To compare time until new onset of non-medical pain reliever misuse between those using and not using cannabis, discrete time survival analysis was used as the primary modeling strategy.²⁹ A discrete time approach was used instead of Cox proportional hazards models because exact timing of non-medical pain reliever misuse was not available. Models included data at each follow-up wave only from participants who were “at risk” (i.e., had not yet reported non-medical pain reliever misuse). Thus, those who had already reported past year non-medical pain reliever misuse at baseline were removed from analyses (n = 476, 7.6%). Further, observations from those who reported non-medical pain reliever misuse at a follow-up wave (e.g., follow-up year 1) were removed from all subsequent waves (e.g., years 2 and 3). There may have been cases where pain reliever misuse occurred prior to baseline that were included in these analyses. Thus, analyses capture new onset of pain reliever misuse since the year prior to baseline, or likely onset. For the remainder of the paper, we simply refer to this as “onset.”

Logistic regression models estimated odds ratios (ORs) for the association between baseline cannabis use and onset of pain reliever misuse among those “at risk” during any given year across the three follow-up years. Initially, non-medical cannabis and medical cannabis use were included in separate models along with demographic covariates, cohort (2014, 2015, 2016), and a linear term for follow-up study wave (1 to 3). Additional models further adjusted for baseline alcohol and cigarette use. A final model included both non-medical and medical cannabis use together with all covariates. A non-proportional “hazards” assumption was assessed by including an interaction between the cannabis use predictor and follow-up study wave. No statistically significant interactions were observed, and we only present ORs pooled across the follow-up waves.

There was attrition over time with 63%, 54%, and 51% participating in the 1-, 2-, and 3-year follow-up surveys, respectively. Of 5,774 participants still “at risk” at baseline, 1,538 did not have any available follow-up data (26.6%) and final analyses included 4,236 participants. Missingness appeared to differ by medical cannabis use, cigarette use, male sex, race/ethnicity, and geographic region (Supplemental Table 2). To account for missing data, we created inverse probability of attrition weights, where logistic regression models were used to estimate probability of non-missingness at each follow-up wave according to covariates used in these analyses, which were used as the denominator and the overall proportion of non-missingness at that wave was used as the numerator for the weights.³⁰ Because of differences in the distribution of sex, geographic region, and race/ethnicity between the study sample and the general WA young adult population according to the 2010 Census, we created post-stratification weights to make the sample more representative according to these characteristics.²⁴ The product of the post-stratification weight and the attrition weight to account for missing follow-up data was calculated and applied to analyses.

To explore whether higher frequency of cannabis use showed stronger associations with pain reliever misuse onset, as a post-hoc analysis, past year non-medical cannabis use at baseline was characterized to include four categories: 1) never, 2) once a year to every other month, 3) once a month to more than once a week, and 4) every other day or more. Indicator variables for these categories with ‘never’ as the reference were included into the statistical models replacing the dichotomous any non-medical use variable.

Stata 16.0 (Stata Corporation, College Station, TX) software was used for analyses.

3. RESULTS

Table 1 shows weighted baseline characteristics of the sample. The past-year prevalence was 40.0% for non-medical cannabis use and 10.1% for medical cannabis use. When considering both forms of use, 31.1% reported only non-medical use, 1.2% reported only medical use, and 8.9% reported both non-medical and medical use. The weighted annual incidence of non-medical pain reliever misuse onset was 2.5%, 2.2%, and 1.6% at follow-up years 1 through 3, respectively. The unadjusted cumulative incidence across waves and by baseline non-medical and medical cannabis use are shown in Supplemental Figures 1 to 3.

Table 1.

Weighted^a baseline characteristics of the study sample (N = 4,236)

	% or Mean (SD)
Age	22.7 (2.2)
Female sex	54.5
Race/ethnicity
White, non-Hispanic	66.5
Asian, non-Hispanic	9.0
Other, non-Hispanicb	11.0
Hispanic, any race	13.5
Attending 4-year college	32.9
Employed full-time	34.6
State geographic region
East	24.3
Northwestc	47.2
Southwest	28.6
Cohort
1 (2014)	31.0
2 (2015)	28.2
3 (2016)	40.7
At least weekly alcohol use in past year	29.4
Any past year cigarette use	20.7
Any past year non-medical cannabis use	40.0
Any past year medical cannabis use	10.1
Non-medical and medical cannabis use
Neither	58.8
Non-medical only	31.1
Medical only	1.2
Both non-medical and medical	8.9

^aPost-stratification weights accounting for demographic differences between study sample and general young adult population in WA

^bIncludes Black, American Indian or Alaska Native, Native Hawaiian or Pacific Islander, Multiracial, Other race

^cIncludes the Greater Seattle metropolitan area

Table 2 shows the survival analysis results, adjusting for different sets of covariates. Non-medical cannabis use compared to no use at baseline was associated with over seven times greater odds of onset of non-medical pain reliever misuse over the three-year follow-up after adjusting for demographic characteristics (OR = 7.13; 95% CI: 4.49, 11.31) and over five times greater odds when further adjusting for at least weekly alcohol use and any past year cigarette use at baseline (OR = 5.27; 95% CI: 3.28, 8.48). Post-hoc analyses suggested that associations were stronger at higher frequencies of non-medical cannabis use (OR for at least every other day use = 8.18; 95% CI: 4.17, 16.04; Supplemental Table 3). Although weaker relative to nonmedical use, medical cannabis use was also associated with non-medical pain reliever misuse onset adjusting for demographics (OR = 3.31; 95% CI: 2.17, 5.05) and further adjusting for alcohol use and cigarette use (OR = 2.21; 95% CI: 1.39, 3.52). Including both forms of cannabis use together in the statistical model, non-medical (OR = 4.64; 95% CI: 2.88, 7.49) and medical cannabis use (OR = 1.65; 95% CI: 1.04, 2.62) showed unique statistically significant associations with non-medical pain reliever misuse onset.

Table 2.

Odds ratios for associations of non-medical and medical cannabis use with onset of pain reliever misuse adjusted for covariates

	Model 1	Model 2	Model 3	Model 4	Model 5
Covariate	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Non-medical cannabis	7.13 (4.49, 11.31)	5.27 (3.28, 8.48)	--	--	4.64 (2.88, 7.49)
Medical cannabis	--	--	3.31 (2.17, 5.05)	2.21 (1.39, 3.52)	1.65 (1.04, 2.62)
Cohort	0.75 (0.61, 0.92)	0.78 (0.63, 0.97)	0.80 (0.66, 0.98)	0.85 (0.70, 1.04)	0.79 (0.64, 0.98)
Study wave	0.81 (0.64, 1.02)	0.81 (0.64, 1.02)	0.86 (0.75, 0.99)	0.87 (0.75, 1.00)	0.81 (0.64, 1.02)
Baseline age	0.92 (0.83, 1.01)	0.92 (0.83, 1.01)	0.93 (0.85, 1.01)	0.90 (0.82, 0.99)	0.91 (0.82, 1.01)
Male sex	1.40 (0.97, 2.00)	1.36 (0.95, 1.94)	1.67 (1.21, 2.31)	1.58 (1.14, 2.19)	1.38 (0.96, 1.98)
Region
East (ref)	--	--	--	--	--
Northwest	1.35 (0.82, 2.23)	1.31 (0.79, 2.18)	2.02 (1.24, 3.28)	1.84 (1.13, 3.00)	1.33 (0.80, 2.19)
Southwest	1.76 (1.00, 3.10)	1.67 (0.94, 2.94)	2.18 (1.28, 3.72)	2.06 (1.20, 3.55)	1.64 (0.93, 2.87)
Race/ethnicity
Non-Hispanic	--	--	--	--	--
White (ref)
Non-Hispanic	1.01 (0.57, 1.79)	1.00 (0.56, 1.77)	0.60 (0.34, 1.07)	0.66 (0.37, 1.17)	1.01 (0.56, 1.82)
Asian
Non-Hispanic	0.60 (0.27, 1.36)	0.56 (0.25, 1.27)	0.55 (0.27, 1.13)	0.50 (0.24, 1.02)	0.53 (0.24, 1.18)
Other
Hispanic any race	0.94 (0.47, 1.89)	0.98 (0.49, 1.97)	0.99 (0.53, 1.85)	1.04 (0.55, 1.96)	0.95 (0.47, 1.92)
4-year college	0.56 (0.37, 0.87)	0.60 (0.39, 0.94)	0.75 (0.51, 1.09)	0.75 (0.51, 1.10)	0.64 (0.41, 1.00)
Full-time employed	0.87 (0.56, 1.38)	0.86 (0.55, 1.35)	0.83 (0.55, 1.24)	0.77 (0.52, 1.15)	0.87 (0.55, 1.36)
Weekly alcohol use	--	1.14 (0.76, 1.73)	--	1.44 (0.97, 2.13)	1.24 (0.81, 1.88)
Cigarette use	--	2.06 (1.39, 3.04)	--	3.02 (2.06, 4.42)	1.89 (1.28, 2.81)

4. DISCUSSION

In this study of young adults in WA, both non-medical and medical cannabis use were uniquely associated with elevated risk of non-medical pain reliever misuse onset over the course of three years, controlling for alcohol and tobacco use and other covariates. These associations were observed among cohorts recruited relatively soon after the opening of the first cannabis retail outlets in WA in 2014. Collectively, findings suggest cannabis use may present a risk for initiation of pain reliever misuse, and additionally, this risk may be greatest for those young adults who use cannabis “recreationally,” which has implications for opioid prevention strategies given that onset of pain reliever misuse is an antecedent of other forms of opioid use disorder and related harms.^18,19 This is important given recent findings showing increases in cannabis use among young adults, particularly those ages 21 and older, following non-medical (or “recreational”) cannabis legalization in WA.²⁴ Further, national studies are showing all-time highs in prevalence of cannabis use, including daily use, and all-time lows in perceived harm from use among young adults.²³ Thus, if evidence continues to mount showing individual-level associations between cannabis use and pain reliever misuse and other forms of opioid use, this may have important implications for cannabis policy. This also further compounds pre-existing concerns about harms related to cannabis use itself including increased risk of cannabis use disorder as well as physical, cognitive, and mental health problems.³

This study extends previous work indicating that cannabis use may precede and influence the onset of pain reliever misuse. While a majority of longitudinal studies in this area have focused on general associations between cannabis use and prevalent opioid use, few studies have examined incident use.³¹ Our findings align with a recent systematic review and meta-analysis showing that cannabis use is associated with greater likelihood of transitioning to opioid misuse.³¹ Additionally, studies have rarely distinguished between non-medical and medical cannabis use.^32,33 Our findings are similar to a study using nationally representative data showing that the elevation in likelihood of prescription opioid misuse was greater among those reporting non-medical cannabis use relative to those reporting medical cannabis.³⁴ Studies considering both non-medical and medical cannabis use will continue to be important.³⁵

Findings should be considered in light of limitations. As noted earlier, individuals who engaged in non-medical pain reliever misuse prior to the baseline survey, but not in the 12 months before baseline, may have been misclassified as at risk for onset. However, given the low prevalence of pain reliever misuse in this study and national studies, this would likely only account for a few cases. Next, the non-medical pain reliever misuse measure asked about pain reliever use “to get high” and, thus, other forms of misuse were not captured (e.g., use to treat pain, but not prescribed). The measure also did not explicitly ask about opioid prescription medications. However, it is unlikely that non-opioid pain relievers would be used “to get high.” With regard to measurement of cannabis use, the study did not assess potency or quantity, which is an ongoing challenge for cannabis-related survey research.³⁶ The study was not a random sample of young adults in WA. However, post-stratification weights were used to make the sample more representative of the state population according to key demographic characteristics. Finally, attrition over the course of the follow-up was relatively high in this study. However, our response rates were similar to or better than other large epidemiologic longitudinal studies, such as Monitoring the Future, with similar compensation.³⁷ Further, we accounted for missingness with attrition weights, which should reduce likelihood for bias.^30,37

5. CONCLUSIONS

In this study, both non-medical and medical use of cannabis were associated with an increased risk of onset of non-medical pain reliever misuse during young adulthood, which appear inconsistent with claims that cannabis use may reduce prevalence of opioid use and related harms.¹ Continued research on this topic is needed. Further studies should examine motives for cannabis use, specifically regarding pain relief compared to other motives (e.g., social, enhancement),^38–40 and their role in the relationship between cannabis use and pain reliever misuse. It will also be important to investigate possible differential associations between cannabis use and pain reliever misuse by demographic characteristics including gender and race/ethnicity because of evidence suggesting differences in prevalence of cannabis use and prescription opioid misuse according to these characteristics.²² Finally, to better understand potential substitution effects, future studies could examine the role of cannabis use on sustained prescription opioid use following a new opioid prescription for medical purposes (e.g., pain related to injury).

HIGHLIGHTS.

• Cannabis has been promoted as an alternative to opioids for pain management

• This study examined associations of cannabis use with non-medical pain reliever misuse onset among young adults

• Non-medical and medical cannabis use was associated with increased risk of non-medical pain reliever misuse onset

• Cannabis use may not be protective against pain reliever misuse and related harms