CHANGES IN DRUG USE BEHAVIORS COINCIDING WITH THE EMERGENCE OF ILLICIT FENTANYL AMONG PEOPLE WHO USE DRUGS IN VANCOUVER, CANADA

Abstract

Background:

With the emergence of illicitly-manufactured fentanyl, drug overdose deaths have risen in unprecedented numbers. In this context, there is an urgent need to characterize potential changes in drug use behaviors among people who use drugs (PWUD).

Objective:

To examine changes in drug use behaviors following the emergence of illicit fentanyl among people who use drugs (PWUD).

Methods:

Data for this cross-sectional analysis was derived from three prospective cohorts of PWUD between December 2016 and May 2017 in Vancouver, Canada. Multivariable logistic regression was used to determine factors associated with self-reported behavior changes (binary variable “yes” or “no”) following the emergence of illicit fentanyl.

Results:

Among 999 participants [363 (36.3%) females], 388 (38.8%) reported some behavior change. The remaining 611 (61.2%) reported no change in behavior; 240 (39.3%) of these individuals had recently been exposed to fentanyl. In multivariable analyses, factors independently associated with behavior change included recent non-fatal overdose (Adjusted Odds Ratio [AOR] = 2.28), active injection drug use (AOR = 1.96), being on opioid agonist therapy (AOR = 1.80), and urine drug screen positive for fentanyl (AOR = 1.45), (all p< .05).

Conclusion:

The majority of PWUD in our sample did not change their drug use behavior despite a high prevalence of fentanyl exposure, indicating a need for targeted behavior change messaging and overdose prevention efforts such as naloxone and addiction treatment for this sub-population of PWUD. Further, the high fentanyl exposure observed in our sample suggests a need to address upstream structural factors shaping the overdose risk in addition to individual behavioral change.

INTRODUCTION

Since late 2013, there has been a public health crisis across the United States and Canada due to an unprecedented number of accidental drug overdose deaths (1). In more recent years, this has been driven by illicitly-manufactured fentanyl and its analogues (hereinafter referred to as “fentanyl”), which are potent synthetic opioids penetrating the illicit drug supply. In the United States and Canada, life expectancy has decreased, and this is thought to be attributable to opioid overdose deaths (2). In the province of British Columbia (BC), one of the hardest hit areas in Canada, a public health emergency was declared in response to the ongoing opioid overdose crisis in April 2016 (3). Despite this, the number of overdose deaths has continued to rise in BC, with fentanyl being detected in 87% of fatal drug overdoses in 2018 (3). While fentanyl is commonly detected in heroin (4), current data show other substances, including cocaine and methamphetamines, may also contain fentanyl, which may be contributing to the increase in overdose deaths (5,6).

Knowledge of the risk of fentanyl appears to be ubiquitous among people who use drugs (PWUD) in Vancouver, BC, as shown by a recent study where out of 1166 PWUD, 93.6% had the knowledge about overdose risks associated with fentanyl (7). However, there are a limited number of studies that examine how the emergence of fentanyl has impacted drug use behaviors. In a mixed-methods study from Rhode Island, participants who used illicit opioids and prescription opioids non-medically reported adjusting their drug use behaviors out of fear of overdose from fentanyl, including testing the dose, using only a trusted source and snorting instead of injecting (8). Another qualitative study from New York City demonstrated PWUD were adapting their drug use practices due to the increased presence of fentanyl in the drug supply, by not using alone, having naloxone available and using a consistent dealer (9). However, other studies have suggested that many PWUD are unaware that they were consuming fentanyl (10,11), and therefore may not adapt drug use behaviors to reduce the risk of harm. In a recent quantitative study of opioid users with suspected fentanyl exposure in three cities in the US, 39% of their sample (n = 196) reported engaging in harm reduction behaviors (e.g., using less of the drug, etc.) to mitigate the risk of overdose (12). In order to understand the extent to which harm reduction behaviors are being adopted by PWUD across North America, it is important to replicate the study using bigger samples and in other settings. Therefore, utilizing large prospective cohort studies of community recruited PWUD in Vancouver, BC, we sought to examine if and how drug use behaviors have changed following the emergence of fentanyl and identify correlates of self-reported changes in drug use behaviors.

METHODS

Study procedures

The Vancouver Injection Drug Users Study (VIDUS), the AIDS Care Cohort to evaluate Exposure to Survival Services (ACCESS), and the At-Risk Youth Study (ARYS) are ongoing open prospective cohorts of PWUD recruited through self-referral and street outreach in Vancouver, Canada. Detailed sampling and recruitment procedures for these cohorts have been described elsewhere (13). Briefly, VIDUS enrolls HIV-negative adults (≥18 years of age) who reported injecting an illicit drug at least once in the month preceding enrollment; ACCESS enrolls HIV-infected adults who reported using an illicit drug other than or in addition to cannabis (which was illegal during the study period) in the previous month; and ARYS enrolls street-involved youth aged 14–26 years who used illicit drugs other than or in addition to cannabis in the month prior to enrollment. For all cohorts, other eligibility criteria include residing in the greater Vancouver region, understanding English and providing written informed consent. The study instruments and follow-up procedures for each study are harmonized to allow for combined analyzes. At baseline and semiannually thereafter, participants complete an interviewer-administered questionnaire eliciting socio-demographic data as well as information pertaining to drug use behaviors, risk behaviors, and health care utilization. Nurses collect blood samples for HIV and HCV serology, provide basic medical care and arrange referrals to appropriate health care services if required. Urine samples are also collected for drug testing. A multi-panel qualitative urine drug screen (UDS), BTNX Rapid Response^™ Multi-Drug Test Panel (Markham, ON, Canada), is utilized to detect fentanyl. While detection times for the substances vary depending on many factors, including routes of administration, frequencies of use, and individual metabolism rates, the BTNX fentanyl test panel is commonly believed to detect exposure to fentanyl within a maximum of past three days (14). The UDS was administered either before or after completing an interviewer-administered questionnaire. When the UDS was administered before the interview, the UDS results were not shown to the participants until the interview was completed in order to avoid reporting bias.

Participants receive a 40 USD (CDN) honorarium for each study visit. The University of British Columbia/Providence Health Care Research Ethics Board provided ethics approval for all studies.

Study sample and primary outcome measure

The participants included in the present cross-sectional analyses were those who completed a study visit between December 1, 2016 and May 31, 2017, and reported using any illicit drugs (other than or in addition to cannabis) in the preceding six months. The sample was further restricted to those who reported ever using opioids. The primary outcome of interest was a self-reported change in drug use behavior in the previous six months corresponding with the emergence of fentanyl. This was derived from a question: “Has the emergence of fentanyl changed how you use drugs in any of the following ways?” Participants were provided with a list of options including, “more likely to use with others”, “more likely to carry naloxone”, “more likely to use in public” “other (i.e., an open-ended option)” and “drug use has not changed,” among other responses. They were able to choose any response or add their own response. Based on this question, we created a dichotomous variable (no change vs. any changes in drug use behavior). While the majority of the reported behavior changes appeared to have been made to prevent overdoses (see Table 1), a small number of participants (n = 15) reported having initiated more risky behaviors, including using drugs alone more often and seeking out fentanyl. These observations were removed from the analysis as this is a distinct group that needs to be considered separately, however, the number was too small to conduct any statistical analyzes.

Table 1.

Reported changes in drug use behavior following the emergence of fentanyl in PWUD in Vancouver, Canada (n = 411)

Behavior change	N	%
More likely to use with others	264	32.4
Inject slowly/taste drug first	150	18.4
More likely to carry naloxone or use where naloxone is available	149	18.3
Using less often and or smaller amount each time	93	11.4
More likely to use supervised injection facility	46	5.6
Use known source/dealer	40	4.9
Stop using opioids	25	3.1
Use drugs unlikely to contain fentanyl	23	2.8
More likely to use in public where people are around	16	1.9
Stopped injecting	9	1.1

PWUD: people who use drugs.

Study variables

Informed by recent literature regarding PWUD’s perceptions and use of fentanyl (15) and correlates of fentanyl exposure, as well as the risk environment framework (16,17), we selected a set of explanatory variables (8,18) that may be associated with the primary outcome. Socio-demographic data included age (per year older); sex (male vs. female); education (high school completion or higher vs. less than high school); ancestry (White vs. nonwhite); homelessness (yes vs. no); and relationship status (married/common law/regular partner vs. other). Drug use behavior variables were dichotomous (yes vs. no) and included: injection drug use; ≥ daily cocaine use; ≥ daily heroin use; ≥ daily crystal methamphetamine use; ≥ daily prescription illicit opioid use; ≥ daily crack smoking; exclusive stimulant use (powder/crack cocaine or crystal methamphetamine but not any opioids); ≥ daily alcohol use; ≥ daily heroin and stimulant use, ≥ daily heroin and alcohol use; ≥ daily heroin and benzodiazepine use; UDS positive for fentanyl; and having a non-fatal drug overdose. We also used dichotomous variables pertaining to health status and healthcare access (yes vs. no), including: receiving opioid agonist therapy (OAT); and ever diagnosed with a mental health disorder. All behavioral variables referred to the previous six months unless otherwise specified.

Statistical analysis

First, we used the logistic regression to examine bivariate associations between the explanatory variables and the primary outcome variable. Then, we used an a priori-defined backward model selection procedure based on examination of Akaike Information Criterion (AIC) to fit a multivariable model (19). In brief, we constructed a full model including all variables that were associated with the outcome at p < .05 in bivariate 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 reached the lowest AIC score. All p-values were two-sided. All statistical analyses were performed using RStudio, version 0.99.892 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

There were a total of 999 eligible participants, of whom 363 (36.3%) were female. In the sample, 551 (55.2%) reported they were of the white ancestry, and the median age was 46 (quartile [Q]1–Q3: 32.7–54.5) years. Of the 999 participants, 388 (38.8%) reported some behavior changes following the emergence of fentanyl, while 611 (61.2%) reported no change. As shown in Table1, the most commonly reported behavior changes included being more likely to use with others (32.4%), injecting slowly or tasting the drug first (18.4%) and carrying naloxone or using where naloxone is available (18.3%). Overall, 482 (48.2%) tested positive for fentanyl, including 240 (39.3%) of the 611 individuals reporting no behavior change. Baseline characteristics of the sample are further presented in Table 2.

Table 2.

Sample characteristics stratified by self-reported behavior changes following the emergence of illicit fentanyl among PWUD in Vancouver, Canada from December 2016-May 2017 (n = 999)

Characteristic	Total (%) (n = 999)	Change in drug use behavior
		No (%) 611 (61.8)	Yes (%) 388 (38.2)
Age (med, IQR)	46.0 (32.7–54.5)	46.1 (33.4–54.8)	45.9 (31.6–54.3)
Male	636 (63.7)	409 (66.9)	227 (58.5)
White	551 (55.2)	331 (54.2)	220 (56.7)
≥ High school education	444 (44.4)	266 (43.5)	178 (45.9)
Homelessness*	177 (17.7)	96 (15.7)	81 (20.9)
In a stable relationship*#	344 (34.4)	218 (35.7)	126 (32.5)
Injection drug use*	667 (66.8)	349 (57.1)	318 (82.0)
Daily prescription opioid use*†	27 (2.7)	12 (2.0)	15 (3.9)
Daily heroin use*†	247 (24.7)	122 (20.0)	125 (32.2)
Daily cocaine use*†	46 (4.6)	21 (3.4)	25 (6.4)
Daily meth use*†	164 (16.4)	97 (15.9)	67 (17.3)
Daily crack use*†	80 (8.0)	46 (7.5)	34 (8.8)
Exclusive stimulant use*†	309 (30.9)	225 (36.8)	84 (21.6)
Daily heroin & alcohol use*†	13 (1.3)	6 (1.0)	7 (1.8)
Daily heroin &benzodiazepine use*†	7 (0.7)	3 (0.5)	4 (1.0)
Daily heroin and stimulant use*†	80 (8.0)	42 (6.9)	38 (9.8)
Daily alcohol use*	86 (8.6)	60 (9.8)	26 (6.7)
Daily benzodiazepine use*†	51 (5.1)	25 (4.1)	26 (6.7)
UDS positive for fentanyl	482 (48.2)	240 (39.3)	242 (62.4)
Overdose*	169 (16.9)	65 (10.6)	104 (26.8)
Opioid agonist therapy*	519 (52.0)	271 (44.4)	248 (63.9)
Ever been diagnosed with a mental health disorder	691 (69.2)	413 (67.6)	278 (71.6)

Group 1: no drug use behavior change; Group 2: drug use behavior change;

^*Denotes activities in the previous six months.

^†Refers to any route of consumption (i.e., sniffing, snorting, smoking or injecting)

^#Legally married/common law/regular partner UDS: urine drug screen, PWUD: people who use drugs; UDS: urine drug screen IQR: interquartile range, CI: Confidence Interval

Table 3 shows the results of the bivariate and multivariable logistic regression analyses. As shown, in the multivariable analysis, factors that were independently associated with self-reported behavior change included: active injection drug use (Adjusted Odds Ratio [AOR] =1.96, 95% Confidence Interval [CI]: 1.37–2.80), recent overdose (AOR = 2.28, 95% CI: 1.56–3.33), UDS positive for fentanyl (AOR = 1.45, 95% CI: 1.04–2.03), and being enrolled in OAT (AOR = 1.80, 95% CI: 1.34–2.43).

Table 3.

Bivariate and multivariable logistic regression analysis of factors associated with self-reported behavior changes following the emergence of illicit fentanyl among PWUD in Vancouver, BC from December 2016-May 2017 (n = 999)

Characteristic	Odds Ratio
	Unadjusted (95% CI)	Adjusted (95% CI)
Age	0.99 (0.98–1.00)
Male	0.70 (0.54–0.91)	0.76 (0.56–1.02)
White	1.11 (0.86–1.43)
≥High school education	1.12 (0.86–1.45)
Homelessness*	1.42 (1.02–1.96)	1.26 (0.87–1.82)
In a stable relationship*#	0.87 (0.66–1.14)
Injection Drug use*	3.41 (2.53–4.65)	1.96 (1.37–2.80)
Daily prescription opioid*†	2.01 (0.93–4.42)
Daily heroin use*†	1.91 (1.42–2.55)	0.89 (0.62–1.28)
Daily cocaine use*†	1.93 (1.07–3.54)	1.48 (0.77–2.84)
Daily meth use*†	1.11 (0.78–1.55)
Daily crack use*†	1.18 (0.74–1.87)
Exclusive stimulant use*†	0.47 (0.35–0.63)	0.83 (0.58–1.20)
Daily heroin & alcohol use*†	1.85 (0.61–5.79)
Daily heroin &benzodiazepine use*†	2.11 (0.46–10.76)
Daily heroin and stimulant use*†	1.47 (0.93–2.33)
Daily alcohol use*	0.66 (0.40–1.05)
Daily benzodiazepine use*†	1.68 (0.96–2.97)
UDS positive for fentanyl	2.47 (1.89–3.24)	1.45 (1.04–2.03)
Overdose*	3.10 (2.21–4.38)	2.28 (1.56–3.33)
Opioid agonist therapy*	2.22 (1.71–2.89)	1.80 (1.34–2.43)
Ever been diagnosed with mental health disorder	1.21 (0.92–1.60)

Group 1: no drug use behavior change; Group 2: drug use behaviors change;

^*Denotes activities in the previous six months.

^†Refers to any route of consumption (i.e., sniffing, snorting, smoking or injecting)

^#Legally married/common law/regular partner; UDS: urine drug screen, PWUD: people who use drugs; IQR: interquartile range, CI: Confidence Interval

DISCUSSION

Less than half of our sample of PWUD in Vancouver reported some behavior changes since the emergence of fentanyl in the illicit drug supply, in an effort to prevent overdose. Compared to those who reported no behavior change, PWUD who reported behavior changes were more likely to be on OAT, more likely to have experienced a non-fatal overdose, have a UDS positive for fentanyl and use injection drugs. Among these individuals, as many as a quarter reported a non-fatal overdose in the past six months and almost two-thirds tested positive for fentanyl. Much of the literature has shown that despite knowledge of fentanyl, some PWUD did not change their drug use behaviors (8,9,12). With the availability of urine drug test results, this study extends existing literature to show that among those who reported no behavior change, more than a third tested positive for fentanyl.

We found that some self-reported behavior changes may not necessarily be preventive against overdose. For example, consistent with previous qualitative and mixed-methods studies in the US (8,9), some individuals reported purchasing drugs from a reliable source. There is an underlying assumption that the source is aware of all the contents of the illicit substance or that they get a consistent supply (20). In a recent qualitative study (20), participants reported a high level of trust in their regular dealers based on length of relationship, consistency of supply and communication. However, this method of protective behavior may leave PWUD open to some risk (20). In this regard, engaging dealers and drug users in drug checking programs may be of use. During the study period, drug checking was provided only as a pilot service at a single site (ie., a supervised injection facility). While we did not capture the use of this service in our cohorts, Karamouzian et al. (2018) showed that less than 1% of people visiting the supervised injection facility were using the service (21). The pilot nature of this service was also reflected in our findings that no one in our study sample mentioned use of drug checking service, which would have been reported in the “other response” section, as a change in their behavior. Therefore, making these programs more accessible is important as they may help inform what dealers are selling and users are buying (20). Taken together with these previous study findings and implications (8,9), our study findings also indicate a need to explore the potential of drug checking programs to reduce overdose risks by educating drug dealers.

Also consistent with previous research (12), other common behavior changes reported by our participants included using a less risky modality of drug use, including using slowly, using a smaller amount and smoking instead of injecting. Data from the Center for Disease Control in the United States reported that deaths with carfentanil detection increased 94% between July 2016 and June 2017, from 421 to 815 in 10 different states (22). In 2019 in British Columbia, upwards of 16% of all fatal overdose deaths were attributable to carfentanil (23). Carfentanil is 10,000 times more potent than morphine. Although, smoking is a safer method of using opioids than injecting (24), the risk of overdose from smoking carfentanil is still quite high as this level of potency was designed to be used in large animals and not humans (25). While point-of-care drug checking that is currently available for use is not able to discriminate carfentanil from fentanyl, such technology, if made accessible, could help inform PWUD. While the present study was unable to discern the effects of each of the different behavior changes, our findings that some of the reported behavior changes were not fully preventive against overdose may partially explain the extremely high prevalence of recent non-fatal overdose (26.8%) despite adapting their behavior to mitigate the risk of overdose. Also, they may suggest the limited effects of individual behavior changes in the context of highly toxic illicit drug supply and point to a need to consider a broader set of interventions to address the social and structural factors that shape PWUD’s behavior, such as homelessness, violence in the unregulated drug market, and criminalization of illicit drug use (16,26,27).

People who did change their behaviors following the emergence of fentanyl were also more likely to be on OAT in the past six months. This is in line with a body of literature indicating that engagement in OAT promotes safer use of drugs (28,29). This may also be partly explained by the fact that those on OAT in Vancouver have their urine tested for fentanyl and are often made aware of the results. Therefore, knowing they are fentanyl positive may result in drug use behavior change. While OAT is also known to prevent overdose among traditional users of opioids (e.g., heroin) (30), it is unknown whether OAT confers the same preventive benefits to those using fentanyl. Future research should investigate the relationship between OAT engagement and the risk of overdose in more depth and determine how OAT can be optimized in the context of the ongoing opioid overdose crisis.

Although PWUD who did not change their behavior were less likely to have had a recent overdose, it is concerning that as many as one-third of these individuals tested positive for fentanyl. It has been shown in the literature that those without a history of overdose may have a false optimism about their risk compared to others who had experienced an overdose (31,32). Accordingly, people may not adjust their behaviors to prevent overdose if they feel they are not at risk (7). In the context of the overdose crisis, illicit opioids are of an unknown potency, and many overdoses have been attributable to highly potent synthetic opioids including fentanyl and carfentanil (22). That being said, it is also important to note some individuals may be seeking out fentanyl and using it based on their tolerance. Therefore, testing positive for fentanyl may not indicate the same level of risk for each individual. It is important to note that given that this cross-sectional study cannot establish temporality of the association between non-fatal overdose and behavior change, it is also possible that non-fatal overdose may have occurred first, and subsequently leading people to change their behavior versus the behavior change not being sufficient to prevent an overdose.

While up to 40% of those who reported no behavior change exclusively used stimulants, previous research drawing on different sources of data in Vancouver suggests that approximately 5% of stimulants might contain fentanyl (18,21). Therefore, people who exclusively use stimulants are not protected from opioid overdose. The high prevalence of fentanyl exposure among those with no behavior change also indicates that the risk of a fatal overdose may be high among this sub-population of PWUD. Therefore, there is an urgent need to provide more targeted overdose prevention education to close the gap between a perceived and actual overdose risk. There is also a need for innovation in treatment options for stimulant use disorder given the limited effectiveness of existing evidence-based treatments in this area (33,34).

Our study has some limitations. There may be some reporting bias in the self-reported data. Testing for fentanyl using a UDS rapid test may have resulted in false negatives if predetermined cutoffs were not met or false positives with other substances that cross-react with fentanyl on this test strip. Also, we did not utilize a random sample, which may limit the generalizability of the results. As with all observational research, confounding bias cannot be excluded even though we sought to address it through the multivariable model fitting procedures.

CONCLUSION

In summary, less than half of our sample of PWUD in Vancouver reported some behavior changes following the emergence of fentanyl in an effort to prevent an overdose. While these individuals were more likely to be on OAT, they were also more likely to have had a recent non-fatal overdose, test positive for fentanyl and engage in injection drug use, indicating that their risk of overdose remains high. Of concern, some of the reported behavior changes may not be preventive against an overdose. These findings indicate the limited effects of individual behavior changes in the context of highly toxic illicit drug supply and the need for targeted behavior change messaging. Also, they point to a need to consider a broader set of structural interventions as part of the overdose prevention efforts such as those addressing homelessness, violence in the unregulated drug market, and criminalization of illicit drug use.