INCREASING PREFERENCE FOR FENTANYL AMONG A COHORT OF PEOPLE WHO USE OPIOIDS IN VANCOUVER, CANADA, 2017–2018

Abstract

Background:

Despite increasing prevalence of illicit fentanyl use in the US and Canada, preference for fentanyl over other illicit opioids has not been fully characterized. Therefore, we sought to describe changes in illicit opioid preferences over time among people who inject drugs (PWID).

Methods:

Data were obtained from two prospective cohort studies between 2017 and 2018. Trends in opioid preference over time were examined using bivariable generalized estimating equation (GEE) analysis. Multivariable models were used to identify factors associated with fentanyl preference.

Results:

Among 732 eligible participants, including 425 (58%) males, the prevalence of preference for fentanyl increased from 4.4% in 2017 to 6.6% in 2018 (Odds Ratio [OR] = 1.27, 95% Confidence Interval [CI]: 1.05–1.52). In a multivariable analysis, younger age (Adjusted Odds Ratio [AOR] = 0.94, 95% CI: 0.92–0.96) and daily crystal methamphetamine injection (AOR = 1.68, 95% CI: 1.01–2.78) were independently associated with preference for fentanyl. The most common reasons for preferring fentanyl included “better high than other opioids” (45%), and “lasts longer than heroin” (27%).

Conclusions:

The current study has demonstrated that preference for fentanyl has been increasing over time among our sample of PWID who use opioids. Further work is needed to clarify risk factors surrounding transitions to illicit fentanyl.

INTRODUCTION

Drug overdose is now a leading cause of death in the United States and Canada, with the rise in mortality being driven by illicit use of synthetic opioids, including fentanyl and its analogues.^1,2 Life expectancies in the United States and Canada are falling for the first time in more than four decades, a phenomenon that is being driven by opioid overdoses.^3,4 The public health burden of the opioid crisis is particularly high due to disproportionate involvement of young and middle-aged individuals in fatal overdose cases, leading to large numbers of years of life lost.⁵ National mortality data from the United States have shown a rapid increase in fentanyl as the most common drug detected in overdose deaths since 2016, with over half of overdose deaths now involving synthetic opioids.^2,6 Urine toxicology studies conducted in our setting of Vancouver, Canada, showed an increase of fentanyl-positive samples from 19% in 2016 to 50% in 2017 among people who reported injection drug use;^7,8 by 2017, among a small sample of 229 participants who reported using opioids from the illicit market, urine samples had become ubiquitously positive for fentanyl.⁹ The higher potency of fentanyl and its analogues makes them easier to traffic and significantly increases overdose risk due to a small margin of error in dosage preparation.¹⁰

The introduction of fentanyl into the drug supply has impacted drug preferences and perceptions of overdose risk among people who use drugs (PWUD). Previous qualitative studies have shown that PWUD differentiate between heroin, fentanyl, and mixed products, and identify the presence of fentanyl based on a number of factors, including perceived potency, speed of onset, taste, color, and appearance.^11–13 Previous studies have also revealed a high level of awareness among PWUD that the presence of fentanyl increases overdose risk,¹⁴ and that fentanyl overdose has certain distinguishing characteristics, including increased naloxone requirements for overdose reversal.¹³ Ad hoc precautionary measures employed by PWUD have included using smaller test doses, observing effects of the drug on others with perceived higher tolerance, and collecting information from others who had purchased fentanyl from the same source.¹⁵ However, previous studies have shown that the ability of PWUD to detect exposure to fentanyl without formal testing is inconsistent, with low concordance in previous studies, and exposure to fentanyl generally being underestimated.^9,12,16 Also, there is emerging evidence to suggest that some PWUD prefer fentanyl to other illicit opioids.^12,17

While previous studies have shown an elevated risk of overdose among PWUD with unintentional fentanyl exposure,^18,19 understanding changes in drug preference is relevant to tailoring current evidence-based interventions for individuals with intentional illicit fentanyl use, including medicated treatments for opioid use disorder and other harm reduction-based approaches such as naloxone distribution and supervised injection facilities. Therefore, the objectives of the current study are to describe changes in opioid preferences over time, and to further characterize a population of PWUD who prefer fentanyl by exploring self-reported reasons for fentanyl preference.

In previous studies, younger age, male gender, crystal methamphetamine injection, heroin injection, homelessness, and previous non-fatal overdose have been identified as risk factors of overdose.^5,20–23 As such, we hypothesized that these factors could be associated with preference for higher potency opioids. In addition, given that methadone and buprenorphine treatment have been shown to be protective against overdose, we hypothesized that lack of access to any medication for opioid use disorder (MOUD) or other healthcare services would also be markers of risk.²⁴ Further, because the prevalence of illicit fentanyl in our setting increased significantly in 2016,²⁵ we were also curious as to whether individuals who began injecting opioids prior to illicit fentanyl being widely available would be less likely to prefer it as compared to individuals who did not begin injecting opioids until after illicit fentanyl became widely available.

METHODS

Data were obtained from the AIDS Care Cohort to evaluate Exposure to Survival Services (ACCESS) and the Vancouver Injection Drug Users Study (VIDUS), both of which are ongoing open prospective cohort studies of PWUD in Vancouver, Canada. Participants are recruited through self-referral, word-of-mouth and street outreach. These studies are described in detail elsewhere.^26,27 Participants are eligible for recruitment if they are at least 18 years of age, reside in Greater Vancouver, and have used illicit drugs other than or in addition to cannabis (which was illegal in this setting until October 2018) in the 30 days prior to baseline interview. For enrollment in VIDUS, participants must also report injection drug use in the past 30 days, and be HIV negative at baseline. ACCESS participants must be HIV positive at baseline. All participants provide written informed consent at the time of enrollment. Following recruitment and semi-annually thereafter, participants complete an interviewer-administered questionnaire to elicit information including substance use patterns, socio-demographic information, and other exposures. ACCESS and VIDUS use a harmonized questionnaire that is prospectively updated to address emerging issues. At each study visit, participants also provided urine samples for a multi-panel qualitative urine drug screen using BTNX Rapid ResponseTM Multi-Drug Test Panel (Markham, ON, Canada). This rapid chromatographic immunoassay screened for fentanyl (fentanyl, 100, and norfentanyl, 20).⁷ The ACCESS and VIDUS studies have been approved by the University of British Columbia/Providence Health Care Research Ethics Board. For the current study, participants were eligible if they completed at least one study visit between June 1, 2017 and November 30, 2018, had a history of injection drug use, and selected a valid response to the outcome measure (excluding those who responded “don’t use opioids”).

For these analyses, the primary outcome of interest was opioid preference, based on responses to the following question: “If you could get good quality street heroin, real prescription opioid pills, medical-grade heroin or fentanyl, which drug would you prefer?” Good quality street heroin referred to illicitly obtained heroin that was not contaminated with fentanyl. At the time of this study, medical-grade heroin was available in the context of a small prescribed injectable diacetylmorphine program.²⁸ Participants who reported preferring fentanyl were subsequently asked why, with potential responses recorded on the questionnaire as “lasts longer than heroin”, “better high than other opioids”, “cheaper than other opioids”, “more easily available”, “easier to make money selling to others”, “easier to withdraw from”, or “other” for which responses were then transcribed.

The outcome variable (preference for fentanyl) was then dichotomized as reporting a preference for fentanyl vs. all other opioids (i.e., illicit heroin, prescription opioid pills and medical-grade heroin). The explanatory variables that we hypothesized would be associated with preference for fentanyl included: age (per year increase), initiation of injection drug use post 2016 (when drug overdoses related to illicit fentanyl began to increase significantly in our setting) vs. prior to 2016, race (white vs. other), gender (male vs. other), at least daily crystal methamphetamine injection (yes vs. no), at least daily heroin injection (yes vs. no), receiving MOUD (yes vs. no), reporting inability to access healthcare or social services (yes vs. no), homelessness (yes vs. no), recent overdose (yes vs. no), HIV positive (yes vs. no). HIV status was included as a covariate because our data was comprised of two cohorts, one of which enrolled HIV-positive individuals. All variables except for HIV serostatus were self-reported. All behavioral variables referred to the six-month period prior to each interview and were treated as time-updated covariates.

First, trends in opioid preference over time were examined using bivariable generalized estimating equation (GEE) analysis based on responses to the above question over the study period. We used descriptive statistics to examine the reasons for preferring fentanyl. For reasons for preferring fentanyl, observations were restricted to the most recent observation for each individual. Baseline sample characteristics were compared between those who reported preference for fentanyl and other opioids, using the Pearson’s Chi-squared test for categorical variables and the Mann-Whitney test for continuous variables.

In order to identify factors associated with fentanyl preference, we fit a multivariable model with an a priori-defined, manual backward model selection procedure based on examination of Akaike Information Criterion (AIC). In brief, we constructed a full model including all variables that were associated with the outcome at p < 0.10 in bivariable analyses. After examining the AIC of the model, we removed the variable with the largest p-value and built a reduced model. We continued this iterative process until we research the lowest AIC score. Further, to explore whether individuals who did not report a preference for fentanyl were exposed to fentanyl, we examined descriptive data regarding fentanyl-positive urine drug screens as well as self-reported suspected exposure to fentanyl stratified by opioid preferences in each follow-up period. Because data was stratified by follow-up period, each individual would have contributed only one observation per follow-up period. All p-values were two-sided. All statistical analyses were preformed using the SAS version 9.4.

RESULTS

A total of 732 participants who used opioids were eligible for the present analyses. Table 1 represents baseline characteristics of the sample stratified by preference for fentanyl vs. all other opioids. As shown, 425 (58.1%) self-identified as male, 317 (43.3%) as white, and the median age was 47.4 years (1st to 3rd quartile [Q1–Q3]: 41.0 − 37.9 − 54.8). A total of 706 (96.4%) participants reported having initiated injection drug use prior to 2016, and the median number of years since first injection was 25.1 years (Q1–Q3: 16.8 − 34.2). There were no significant differences in other baseline sample characteristics among participants who initiated drug injection prior to versus after 2016 (data not shown).

Table 1.

Baseline Characteristics of 732 Participants who Reported an Opioid Preference, Stratified by Preference for Fentanyl.

Demographic Characteristic	Preference for:		p-Value
	Fentanyl n (%) 70 (9.6%)	Other opioids n (%) 662 (90.4%)
Age
Median (1st to 3rd quartile)	38.1 (34.5–48)	48.1 (39.2–55.2)	<0.001
Gender
Male	36 (51.4)	389 (58.8)	0.237
Non-male	34 (48.6)	273 (41.2)
Race
White	36 (51.4)	281 (42.4)	0.137
Non-white	33 (47.1)	375 (56.6)
Crystal methamphetamine injection*
≥Daily	19 (27.1)	77 (11.6)	<0.001
<Daily	51 (72.9)	585 (88.4)
Heroin injection*
≥Daily	36 (51.4)	238 (36.0)	0.011
<Daily	34 (48.6)	424 (64.0)
Non-fatal overdose*
Yes	12 (17.1)	155 (23.4)	0.230
No	58 (82.9)	505 (76.3)
MOUD*
Yes	41 (58.6)	405 (61.2)	0.660
No	29 (41.4)	256 (38.7)
Inability to access services*
Yes	20 (28.6)	135 (20.4)	0.116
No	50 (71.4)	524 (79.2)
Homelessness*
Yes	18 (25.7)	115 (17.4)	0.085
No	52 (74.3)	547 (82.6)
Initiated injecting after 2016
Yes	5 (7.1)	21 (3.2)	0.093
No	65 (92.9)	641 (96.8)
HIV
Yes	22 (31.4)	231 (34.9)	0.093
No	48 (68.6)	431 (65.1)

^*All behavioral variables refer to the six months prior to interview

The 732 participants contributed 1508 observations to the present analyses. Across all study observations, 511 (33.9%) preferred street heroin, 89 (5.9%) preferred prescription opioid pills, 747 (49.5%) preferred medical-grade heroin, and 161 (10.7%) preferred fentanyl. Figure 1 represents trends in preference for street heroin, opioid pills, medical-grade heroin, or fentanyl over the study period. While preference for street heroin (Odds Ratio [OR] = 1.41, 95% Confidence Interval [CI]: 1.24–1.59) and fentanyl (OR = 1.27, 95% CI: 1.05–1.52) increased over the study period, preference for prescription opioids (OR = 0.75, 95% CI: 0.59–0.95) and medical grade heroin (OR = 0.72, 95% CI: 0.64–0.80) decreased.

Figure 1.

Trends in preference for street heroin, prescription opioid pills, medical-grade heroin, or fentanyl among 732 people who use opioids in Vancouver, Canada, 2017–2018 (n = 1508).

Table 2 shows the analyses of factors associated with preference for fentanyl. Following the manual backward model selection procedure, age (AOR = 0.94, 95%, CI: 0.92–0.96) and at least daily crystal methamphetamine injection (AOR = 1.68, 95%, CI: 1.01–2.78) were independently and positively associated with preference for fentanyl.

Table 2.

Univariate and Backward Selection Multivariable Analysis of Factors Associated With Fentanyl Preference Among 732 Participants Who Used Opioids.

Characteristic	Univariate model		Backward selection multivariable model
	Odds ratio (95% CI)	p-Value	Odds ratio (95% CI)	p-Value
Age
(Per year increase)	0.94 (0.92–0.96)	<0.001	0.94 (0.92–0.96)	<0.001
Gender
(Male vs. non-male)	0.68 (0.46–0.99)	0.045
Race
(White vs. non-white)	1.18 (0.80–1.73)	0.408
Crystal methamphetamine injection* use injection
(≥Daily vs. <daily)	2.24 (1.38–3.64)	0.001	1.68 (1.01–2.78)	0.044
Heroin injection*
(≥Daily vs. <daily)	1.33 (0.94–1.90)	0.107
Non-fatal overdose*
(Yes vs. no)	0.85 (0.54–1.33)	0.466
MOUD*
(Yes vs. no)	0.81 (0.55–1.17)	0.259
Inability to access services* community health or social services in the last 6 months
(Yes vs. no)	1.10 (0.73–1.64)	0.661
Homelessness*
(Yes vs. no)	1.68 (1.11–2.54)	0.014
Initiated injecting after 2016
(Yes vs. no)	2.35 (1.01–5.45)	0.046
HIV
(Yes vs. no)	0.96 (0.64–1.43)	0.831

GEE: generalized estimating equations; CI: confidence interval.

^*All behavioral variables refer to the six months prior to interview.

Among 84 participants who reported preferring fentanyl at the most recent interview, 114 reasons for this preference were reported as participants were permitted to provide more than one reason for their preference. The most common responses provided included “better high than other opioids” (51, 44.7%), “lasts longer than heroin” (31, 27.2%), “addicted to fentanyl now” (10, 8.8%), and “more easily available” (6, 5.3%). Reasons for preferring fentanyl are illustrated in Figure 2. Figure 3 shows frequency of fentanyl-positive urine drug screens and frequency of self-reported suspected exposure to fentanyl, stratified by follow-up period. Many individuals who did not report a preference for fentanyl suspected that they had been exposed to it and tested positive on urine drug screens.

Figure 2.

Reasons for reporting a preference for fentanyl (n = 114).

Figure 3.

Fentanyl-positive urine drug tests (n = 1250) and self-reported suspected fentanyl exposure (n = 1506) stratified by opioid preference and follow-up period.

DISCUSSION

We demonstrated that preference for fentanyl was increasing between 2017 and 2018 among our cohorts of PWUD who used opioids. In a multivariable analysis, younger age and daily crystal methamphetamine injection remained independently associated with preference for fentanyl. Most commonly reported reasons for preferring fentanyl included more euphoria, longer effects, and development of high opioid tolerance.

Our findings are consistent with a recent cross-sectional study documenting that approximately one quarter of 308 PWUD surveyed in three cities in the United States expressed a preference for fentanyl over other illicit opioids.²⁹ Our longitudinal study has built on this finding by demonstrating that preference for fentanyl has become more common over time, and by exploring associations with preference for fentanyl. In our cohorts of adult PWID, the majority of participants had established injecting careers with a median of >20 years since first injection. We had postulated initiation of drug injection after 2016, when fentanyl became common in the local illicit opioid supply, may be associated with increased familiarity with fentanyl and reduced familiarity with previously available forms of “street heroin.” The results of our GEE analysis suggest that fentanyl preference is associated with younger age rather than initiation of injecting after 2016. In our sample, age was collinear with length of injecting career, limiting our ability to clarify whether fentanyl preference was associated specifically with younger age or shorter injecting career. Interestingly, an association between younger age and preference for fentanyl has been reported in other settings.²⁹ The small number of participants in the current study who initiated injection drug use after 2016 (n = 26, 3.5%) limit the interpretation of this variable.

In addition, this study has demonstrated that preference for fentanyl was associated with frequent crystal methamphetamine injection. This highlights a group of high-risk individuals with preference for high potency opioids in the context of high intensity stimulant use that has been described in previous studies.²⁰ The phenomenon of concomitant use of fentanyl and crystal methamphetamine has been documented in relation to increasing prevalence of both substances in recent years.^30,31 In addition, toxicity data from both the United States and British Columbia have revealed increasing rates of co-detection of crystal methamphetamine and fentanyl in fatal overdose cases.^32,33 Interestingly, Morales et al. did not find an association between preference for fentanyl and crystal methamphetamine use in their study of people who use illicit opioids on the east coast of the United States, perhaps as a result of relatively lower prevalence of crystal methamphetamine in that setting as opposed to the west coast of North America.^29,30

It is possible that a portion of PWUD who seek out illicit heroin in our setting may be unknowingly consuming fentanyl. A recent study of drug checking services offered within low-barrier harm reduction settings revealed that of 907 samples identified as heroin by participants, only 160 (17.6%) contained heroin, while 822 (90.6%) contained fentanyl.³⁴ Given that fentanyl is now present in a large majority of illicit heroin that is available in our setting, the observed increase in preference for illicit heroin may reflect instances when PWUD are seeking the embodied effects of fentanyl, but use the terminology “street heroin.” It should be noted that for the purposes of this study, interviewers clarified that the terminology ‘street heroin’ was meant to represent “good quality street heroin,” not contaminated by fentanyl. Regardless, given the self-reported nature of this variable, the terminology would be open to interpretation and influenced by the lexicon of local drug markets.

The availability of medical-grade heroin in our setting is related to the existence of a prescribed injectable diacetylmorphine program that is unique in North America and was established in 2009 in the context of a randomized controlled trial, with ongoing, though limited, access following the completion a second randomized controlled trial in 2016.^35,36 We postulate that the decreasing preference for medical-grade heroin over time may be representing PWUD who had initial interest in the diacetylmorphine program and either did not find it met their needs or were placed on a waitlist for access, though further research is needed to clarify this.

This study has some limitations. Firstly, the study questionnaire did not differentiate between medical-grade fentanyl and illicit fentanyl. For the purposes of this study, the authors have made the assumption that the majority of participants were referring to illicit fentanyl, an assumption we believe is reasonable given that participation in this cohort study is based on illicit drug use, and that no prescription fentanyl program for opioid use disorder existed in our setting at the time of data collection. Second, our findings may not be generalizable to other settings where illicit fentanyl is less common. This questionnaire did not collect information on reasons for not preferring fentanyl, or for preferring opioids other than fentanyl, which limited our ability to explore opioid preference more broadly in this population. In addition, study participants were recruited via community-based approaches including snowball sampling which may limit generalizability. Lastly, the nature of self-reported observations is that they can be influenced by reporting bias, though this type of data has been used in previous cohort studies and has been shown to be valid in studies involving PWID.³⁷

In the current study, we demonstrated that preference for fentanyl increased between 2017 and 2018 among our cohorts of PWID who used opioids. Preference for fentanyl was associated with younger age and daily crystal methamphetamine use. Gaining an understanding of changes in opioid preferences among PWUD will allow healthcare providers to anticipate changes in service needs and better tailor services to high-risk individuals, including those with a preference for fentanyl and products containing fentanyl. There will continue to be increased need for evidence-based overdose prevention interventions, including expanded access to supervised consumption facilities, naloxone distribution, and MOUD, while accounting for increasing potency of the illicit drug supply. Further work is needed to clarify the relationship between the risk factors surrounding transitions to fentanyl, as well as to better understand how PWID who prefer fentanyl identify, quantify, and reduce overdose risk and how they experience current harm reduction and treatment services.