Prevalence and Correlates of Reporting a Gabapentin Prescription for Pain Among People Who Use Opioids: A Longitudinal Community‐Recruited Study

ABSTRACT

Introduction

Gabapentin is widely prescribed off‐label to treat pain. However, there are concerns about the proliferation of medical and non‐medical use, and the potential increased risk of overdose death associated with concurrent opioid use. Despite these considerations, prescription gabapentin use among people who use unregulated drugs is uncharacterised in Canada. Therefore, we investigated trends and factors associated with being prescribed gabapentin for pain among community‐recruited people who use opioids.

Methods

Data were derived from three prospective cohort studies of people who use unregulated drugs in Vancouver, Canada, restricted to periods of opioid use (i.e., unregulated use, non‐medical use, and/or opioid agonist treatment). We characterised factors associated with reporting gabapentin for pain between 2014 and 2020 using multivariable generalised estimating equations. Trends in self‐reported prescription (2014–2020) and non‐medical (2016–2024) gabapentin use were explored.

Results

Between 2014 and 2020, we included 10,999 observations from 2039 participants who had a baseline median age of 38 years and were 62.4% male and 57.5% White. Gabapentin prescriptions for pain were reported by 255 (12.5%) participants. Observations reporting a neuropathic pain diagnosis had 3.54 times the adjusted odds (95% CI 2.81, 4.46) of being prescribed gabapentin. From 2014 to 2020, the prevalence of gabapentin prescriptions rose from 5.4% to 7.8%. Non‐medical gabapentin use remained low.

Discussion and Conclusions

Among people who use opioids, we observed an increase in gabapentin prescriptions for pain between 2014 and 2020. Gabapentin prescriptions were strongly associated with reporting a neuropathic pain diagnosis. Rates of non‐medical gabapentin use were substantially lower than rates reported internationally.

Key Points Summary

• Gabapentin is increasingly prescribed for pain among people who use opioids in the study setting.

• Gabapentin for pain is highly associated with reporting a neuropathic pain diagnosis.

• Non‐medical use (misuse/abuse) of gabapentin remains low in the study setting.

1. Introduction

Gabapentin is an antiepileptic drug approved in Canada in 1994 as an adjunct therapy for epilepsy [1]. One of the most commonly prescribed medications in Canada, estimates suggest that the consumption of daily doses of gabapentin doubled between 2008 and 2018 [2, 3]. This increase is thought to largely be driven by off‐label use (i.e., gabapentin being prescribed for the treatment of a condition that it has not received approval for), which has been estimated to represent over 95% of gabapentin prescriptions among available studies from settings in the United States and Canada [4, 5, 6]. Most commonly prescribed off‐label for the management of chronic pain, especially neuropathic pain, gabapentin is also used across unapproved conditions including migraines, insomnia, withdrawal management, alcohol use disorder, anxiety disorder and bipolar disorder [7, 8]. Steady increases in gabapentin use have also been observed internationally across high‐income countries, with some nations carrying additional indications including postherpetic neuralgia in the United States and peripheral neuropathic pain (e.g., painful diabetic neuropathy) in Europe [3, 9, 10, 11, 12].

The rapid increase in prescribing has been attributed in part to its use as an alternative to opioids for the management of chronic pain in the context of the overdose crisis and growing evidence challenging the safety and efficacy of long‐term opioid therapy for chronic pain [13, 14, 15]. This has been further supported by clinical guidelines presenting gabapentin as a first‐line therapy for broadly indicated neuropathic pain [16, 17, 18], despite the United States Food and Drug Administration's rejection of gabapentin for additional pain conditions in the early 2000s [19]. Since it became available as a generic drug in the early 2000s, the financial incentive to seek approval for additional indications is low. While off‐label pharmaceutical prescribing is not inherently harmful or unusual, it operates at a greater level of clinical uncertainty that makes weighing the potential risks and benefits of treatment more challenging [20]. In the context of gabapentin, there are growing concerns around the rapidly expanded use of the drug in light of the inconsistent quality of evidence supporting its off‐label use, poorly demonstrated efficacy across unapproved indications [8, 13, 19] and emergent safety concerns [21].

Originally considered to have a favourable safety profile, the rapid increase in gabapentin use has been accompanied by an improved understanding of its adverse effects. A 2017 pharmacoepidemiologic study found that moderate‐to‐high‐dose gabapentin increases the likelihood of opioid‐related overdose death by nearly 60% among people concurrently prescribed opioids [22]. Spurred by additional observational studies and adverse event reporting, both Health Canada and the United States Food and Drug Administration issued warnings in 2019 around the increased risk of opioid‐related overdose when co‐prescribing gabapentin alongside opioids and other central nervous system (CNS) depressant drugs [23, 24]. Nevertheless, the persistent need to manage complicated conditions such as chronic pain has likely led to the continued use of gabapentin as a commonly co‐prescribed therapy to other CNS drugs such as opioids and benzodiazepines [25]. There is conflicting evidence on whether the increased risk of opioid‐related overdose death remains among people with opioid use disorder (i.e., people with a higher risk of opioid‐related mortality) [25, 26, 27, 28]. However, the use of gabapentinoids for the treatment of chronic pain was discouraged in the Canadian guidelines for the management of opioid use disorder given persistent concerns [29].

In addition to the elevated risk of overdose, there is a growing body of literature describing gabapentin's potential for non‐medical use. Often described as abuse and misuse, non‐medical use refers to using prescribed medications beyond their intended duration, dosage or purpose or without a personal prescription [30]. Two recent reviews highlight a growing trend in non‐medical use that is in turn linked to harmful outcomes including an increased risk of opioid‐related death and hospitalisation [21, 31]. Motivations for non‐medical use include its euphoric effects and ability to potentiate other drugs (e.g., opioids), as well as to self‐manage withdrawal, anxiety and pain [21]. Numerous studies have identified an association between non‐medical gabapentin use and opioid use disorder, prompting both reviews to highlight the need to exercise caution when prescribing to people with or at risk of developing, opioid use disorder [21, 31].

Despite the higher risk of non‐medical use and the potential increased risk of overdose, we are unaware of any study investigating the prescribing of gabapentin for pain among people who use opioids in the Canadian context. Thus, the study's primary objective was to investigate trends and identify factors associated with self‐reporting a gabapentin prescription for pain among people who use opioids. We additionally explored trends of reported non‐medical gabapentin use and prescription gabapentin use in the context of alcohol use treatment to further characterise the utilisation of gabapentin in the study setting.

2. Methods

2.1. Study Design

For this study, we used data from three open, ongoing and community‐recruited prospective cohort studies of people who use unregulated drugs in Vancouver, Canada, a setting with an ongoing overdose crisis and high rates of chronic pain. These cohorts include 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). The cohorts have been detailed previously [32, 33, 34]. Briefly, since 2005, the cohorts have recruited participants using methods including street outreach, word of mouth and self‐referral. Recruitment and follow‐up activities for VIDUS and ACCESS largely focus on Vancouver's Downtown Eastside, an urban neighbourhood with high rates of polysubstance use, criminalisation and marginalisation, while ARYS operates in the Downtown South, a similar neighbourhood with a substantial population of street‐involved youth. VIDUS is composed of adults at risk of human immunodeficiency virus (HIV) who injected drugs in the month prior to enrolment; ACCESS is composed of people living with HIV who used unregulated drugs (other than or in addition to cannabis) in the month prior to enrolment. ARYS includes street‐involved youth aged 14–26 who used unregulated drugs in the month prior to enrolment. VIDUS and ARYS participants who seroconvert to HIV‐positive status during follow‐up are transferred to the ACCESS cohort. All eligible participants provided written informed consent at enrolment. The cohorts' protocols have been harmonised to allow for pooled analyses.

At baseline and every 6 months thereafter, participants are invited to complete interviewer‐administered questionnaires that cover a range of topics including socio‐demographic characteristics, substance use behaviours, social‐structural exposures, sexual behaviours, use of harm reduction services and care for substance use disorders. Nurse‐administered questionnaires on health status and service use are also conducted at each visit. Participants received a $50 (CAD) honorarium at each study visit during the study period ($40 prior to December 2022). Following institutional guidance, we suspended all in‐person research activities between 17 March and 17 July 2020, due to the emergence of the novel coronavirus and associated infection control measures. From July 2020 to March 2022, we conducted participant interviews over the phone, providing loaner telephones to participants if needed. We resumed all in‐person research activities on 23 March 2022 [35]. All three cohorts have received annual review and approval from the relevant research ethics boards at the University of British Columbia, Simon Fraser University and Providence Health Care.

2.2. Study Sample

In these analyses, we included data from study interviews conducted between 2 June 2014 and 31 May 2024. Observations were eligible for inclusion if participants reported any opioid use in the previous 6 months, defined as reporting unregulated opioid use, non‐medical use of prescribed opioids and/or opioid agonist treatment. This restriction was imposed given the previously identified concerns associated with concurrent opioid and gabapentin use, including increased risk of opioid‐related overdose and non‐medical use of gabapentin. Observations reporting only opioid agonist treatment were included to capture data from participants with an opioid use disorder who were in periods of abstinence from the unregulated opioid supply, as they were understood to be conceptually relevant to this exploratory analysis. For the primary analyses of prescription gabapentin described below, we included data from study interviews conducted between June 2014 and March 2020. For the trend analyses of non‐medical gabapentin use and prescription gabapentin use for alcohol use treatment, we included baseline and follow‐up visits between December 2016 and May 2024. These periods were determined based on the availability of specific questions in the cohort instrument.

2.3. Study Variables

The primary outcome of interest was self‐reporting being prescribed gabapentin for pain at the time of the study interview. The restriction to pain as the indication of interest was imposed as it is a common driver of off‐label gabapentin prescribing, as well as pre‐determined limitations of the cohort instrument (i.e., broadly indicated gabapentin prescriptions were not explored in the survey). Pregabalin, a drug within the gabapentinoid class which has additional indications and different pharmacokinetics, was not included in the present analysis [36]. The exclusion was implemented to simplify the analysis following preliminary findings among the three cohorts (2013–2020) that pregabalin was very rarely reported to be prescribed for pain. This is likely owing to the fact that gabapentin and not pregabalin, is fully covered by the provincial PharmaCare plans that most cohort participants are eligible for (e.g., Income Assistance and First Nations Health Benefits).

We considered the inclusion of explanatory variables based on previous research on pain among people who use unregulated drugs and our experience in the study setting. We included measures of sociodemographic characteristics, specifically: sex assigned at birth (male vs. female); age (in years); and self‐reported ethnicity/ancestry (Indigenous vs. person of colour/other vs. White). Substance use variables referring to behaviours in the previous 6 months were: ≥daily cannabis use (yes vs. no); ≥daily alcohol use (yes vs. no); ≥daily stimulant use (including cocaine, crack cocaine or crystal methamphetamine; yes vs. no); ≥daily sleeping pill use (yes vs. no); ≥daily non‐medical benzodiazepine use (yes vs. no); ≥daily fentanyl or other unregulated opioid use (yes vs. no); and ≥daily non‐medical prescription opioid use (yes vs. no). Data on non‐medical gabapentin use (yes vs. no) and gabapentin prescriptions in the context of alcohol use treatment (yes vs. no) in the 6 months prior to the study visit were also captured. We measured relevant health conditions via: ever been diagnosed with bipolar disorder (yes vs. no); HIV serostatus (positive vs. negative); and ever been diagnosed with a chronic pain condition, specifying subtypes including neuropathic pain (yes vs. no), inflammatory pain (yes vs. no), muscle pain (yes vs. no), bone/mechanical/compressive pain (yes vs. no) and headaches/migraines (yes vs. no). We measured anxiety/depression symptomatology using a sub‐section of the EuroQol EQ‐5D instrument (extreme vs. moderate vs. none) [37]. We asked about engagement in opioid agonist treatment (yes vs. no) and experiencing a non‐fatal overdose (yes vs. no) in the previous 6 months.

2.4. Statistical Analyses

First, we determined the characteristics of the sample at baseline (i.e., earliest interview during the study period), stratified by the outcome and tested for differences using the Mann–Whitney and Pearson's Χ ² tests, as appropriate. When the expected counts included a cell with less than five responses, we used Fisher's exact test. Next, we descriptively presented the trends in the prevalence of gabapentin prescriptions between 2014 and 2020, including among the sub‐section of observations that reported a chronic neuropathic pain condition. To estimate the prevalence rate across the study period, we used generalised estimating equations with a log link and exchangeable correlation structure to estimate the association between study year and the prevalence of gabapentin prescriptions for pain. The model controlled for all factors that were listed in Table 2 that reached significance (p < 0.05) in bivariate analyses.

TABLE 2.

Factors longitudinally associated with reporting gabapentin prescriptions for pain among people who use opioids in Vancouver, Canada (2014–2020). ^a

Characteristic	Unadjusted (n = 10,907–10,999)		Adjusted (n = 10854 c )
	Odds ratio (95% CI)	p‐value	Odds ratio (95% CI)	p‐value
Age
Per 10 years older	1.58 (1.42, 1.76)	< 0.001	1.40 (1.24, 1.59)	< 0.001
Sex
Male vs. female	0.88 (0.66, 1.17)	0.371
Ethnicity/Ancestry
Indigenous vs. White	0.96 (0.71, 1.29)	0.793	1.12 (0.82, 1.51)	0.479
POC/other vs. White	0.38 (0.16, 0.92)	0.033	0.73 (0.31, 1.74)	0.479
HIV serostatus
Positive vs. negative	1.79 (1.35, 2.38)	< 0.001	1.22 (0.91, 1.63)	0.179
Opioid agonist treatment b
Yes vs. no	1.77 (1.39, 2.24)	< 0.001	1.70 (1.31, 2.21)	< 0.001
Daily non‐medical prescription opioid use b
Yes vs. no	0.80 (0.49, 1.32)	0.385
Daily fentanyl/unregulated opioid use b
Yes vs. no	1.00 (0.82, 1.23)	0.972
Daily cannabis use b
Yes vs. no	1.01 (0.78, 1.30)	0.968
Daily stimulant use b , d
Yes vs. no	0.89 (0.75, 1.06)	0.181
Daily sleeping pills use b
Yes vs. no	0.52 (0.08, 3.37)	0.494
Daily non‐medical benzodiazepine use b
Yes vs. no	1.60 (0.72, 3.54)	0.251
Daily alcohol use b
Yes vs. no	1.14 (0.84, 1.55)	0.402
Chronic neuropathic pain
Yes vs. no	3.90 (2.95, 5.15)	< 0.001	3.54 (2.81, 4.46)	< 0.001
Chronic inflammatory pain
Yes vs. no	1.45 (1.17, 1.79)	< 0.001	1.22 (1.02, 1.46)	0.028
Chronic muscle pain
Yes vs. no	0.90 (0.50, 1.62)	0.732
Chronic bone pain
Yes vs. no	1.66 (1.36, 2.03)	< 0.001	1.46 (1.23, 1.74)	< 0.001
Chronic headaches
Yes vs. no	1.30 (0.78, 2.19)	0.314
Anxiety/depression b
Extremely vs. not	1.01 (0.74, 1.39)	0.939
Moderate vs. not	1.09 (0.94, 1.26)	0.246
Bipolar disorder
Yes vs. no	1.60 (1.06, 2.39)	0.024	1.62 (1.07, 2.45)	0.021
Non‐fatal overdose b
Yes vs. no	1.17 (0.97, 1.42)	0.099

Abbreviations: CI, confidence interval; HIV, human immunodeficiency virus; POC, person of colour.

^a Opioid use was defined as unregulated opioid use, non‐medical use of prescribed opioids and/or opioid agonist treatment.

^b In the 6 months prior to the interview date.

^c 145 observations were removed from the final complete‐case model due to missing data.

^d Defined as daily cocaine, crack cocaine, or crystal methamphetamine use.

Finally, we estimated the relationships between the outcome and each explanatory variable of interest. As our data could include serial measures from each participant, we accounted for within‐subject correlations by using generalised estimating equations (GEE) with a logit‐link function and an exchangeable correlation structure. We performed bivariate GEE analyses to determine factors associated with being prescribed gabapentin for pain and fit a complete case multivariable GEE model including all explanatory variables with a significance level of p < 0.05 in bivariate analyses. We calculated generalised variance inflation factors to assess the potential for multicollinearity in the multivariable GEE model [38]. As secondary analyses, we used descriptive statistics to explore the proportion of participants that reported: (i) non‐medical gabapentin use; and (ii) receiving gabapentin for the treatment of alcohol use at each follow‐up visit between 2016 and 2024. All analyses were performed using R (Version 4.2.2, R Foundation for Statistical Computing, Vienna, Austria) [39]. All p‐values were two‐sided and considered significant at p < 0.05.

3. Results

Between June 2014 and March 2020, any opioid use in the last 6 months was reported in 10,999 study visits (68.1% of total visits) by 2039 participants. A minority (7.16%) of participants were included solely based on their engagement with opioid agonist treatment. Of the 2039 participants, the median age was 38 years old at baseline, 1272 (62.4%) were male and 1172 (57.5%) were White. Over the study period, participants contributed an average of five study visits (interquartile range [IQR]: 2–9).

Participant characteristics at baseline are presented in Table 1, stratified by reporting a gabapentin prescription for pain. Of note, 255 (12.5%) participants reported being prescribed gabapentin for pain at least once over the study period. Between 2014 and 2020, there were 700 study visits reporting a gabapentin prescription for pain, with the prevalence increasing from 5.4% to 7.8% among the entire sample and from 27.0% to 56.4% among the subgroup of observations that reported ever being diagnosed with neuropathic pain (n = 907) (Figure 1; Table S1). After controlling for age, ethnicity/ancestry, HIV serostatus, opioid agonist treatment engagement, bipolar disorder and chronic neuropathic, inflammatory and bone pain in the multivariable GEE, there was a 7% (95% confidence interval [CI] 1.02–1.12) higher prevalence of gabapentin prescriptions for pain per year, meaning a cumulative prevalence rate of 1.50 (95% CI 1.13–1.97) over the study period (2014–2020).

TABLE 1.

Baseline characteristics stratified by prescription gabapentin for pain among people who use opioids in Vancouver, Canada (2014–2020). ^a

Characteristic	Total, n = 2039 (100%)	Yes, n = 100 (4.9%)	No, n = 1939 (95.1%)	p‐value
Age (median, IQR)	37.6 (26.7–50.2)	47.0 (37.0–53.1)	36.7 (26.4–49.9)	< 0.001
Sex
Male	1272 (62.4%)	65 (65.0%)	1207 (62.3%)	0.580
Female	767 (37.6%)	35 (35.0%)	732 (37.8%)
Ethnicity/Ancestry
Indigenous	723 (35.5%)	37 (37.0%)	686 (35.4%)	0.408
POC/other	125 (6.1%)	3 (3.0%)	122 (6.3%)
White	1172 (57.5%)	59 (59.0%)	1113 (57.4%)
Missing	19 (0.9%)	1 (1.0%)	18 (0.9%)
HIV serostatus
Positive	539 (26.4%)	45 (45.0%)	494 (25.5%)	< 0.001
Negative	1493 (73.2%)	55 (55.0%)	1438 (74.2%)
Missing	7 (0.3%)	0 (0.0%)	7 (0.4%)
Opioid agonist treatment b
Yes	1169 (57.3%)	76 (76.0%)	1093 (56.4%)	< 0.001
No	865 (42.4%)	24 (24.0%)	841 (43.4%)
Missing	5 (0.3%)	0 (0.0%)	5 (0.3%)
Daily non‐medical prescription opioid use b
Yes	127 (6.2%)	11 (11.0%)	116 (6.0%)	0.043
No	1911 (93.7%)	89 (89.0%)	1822 (94.0%)
Missing	1 (0.1%)	0 (0.0%)	1 (0.1%)
Daily fentanyl/unregulated opioid use b
Yes	726 (35.6%)	27 (27.0%)	699 (36.0%)	0.063
No	1309 (64.2%)	73 (73.0%)	1236 (63.7%)
Missing	4 (0.2%)	0 (0.0%)	4 (0.2%)
Daily cannabis use b
Yes	574 (28.2%)	28 (28.0%)	546 (28.2%)	0.952
No	1457 (71.5%)	72 (72.0%)	1385 (71.4%)
Missing	8 (0.4%)	0 (0.0%)	8 (0.4%)
Daily stimulant use b , c
Yes	735 (36.0%)	34 (34.0%)	701 (36.2%)	0.635
No	1294 (63.5%)	66 (66.0%)	1228 (63.3%)
Missing	10 (0.5%)	0 (0.0%)	10 (0.5%)
Daily sleeping pills use b
Yes	24 (1.2%)	2 (2.0%)	22 (1.1%)	0.332
No	2010 (98.6%)	98 (98.0%)	1912 (98.6%)
Missing	5 (0.3%)	0 (0.0%)	5 (0.3%)
Daily non‐medical benzodiazepine use b
Yes	12 (0.6%)	2 (2.0%)	10 (0.5%)	0.114
No	2026 (99.4%)	98 (98.0%)	1928 (99.4%)
Missing	1 (0.1%)	0 (0.0%)	1 (0.1%)
Daily alcohol use b
Yes	181 (8.9%)	8 (8.0%)	173 (8.9%)	0.748
No	1855 (91.0%)	92 (92.0%)	1763 (90.9%)
Missing	3 (0.2%)	0 (0.0%)	3 (0.2%)
Chronic neuropathic pain
Yes	154 (7.6%)	41 (41.0%)	113 (5.8%)	< 0.001
No	1875 (92.0%)	59 (59.0%)	1816 (93.7%)
Missing	10 (0.5%)	0 (0.0%)	10 (0.5%)
Chronic inflammatory pain
Yes	263 (12.9%)	28 (28.0%)	235 (12.1%)	< 0.001
No	1766 (86.6%)	72 (72.0%)	1694 (87.4%)
Missing	10 (0.5%)	0 (0.0%)	10 (0.5%)
Chronic muscle pain
Yes	53 (2.6%)	1 (1.0%)	52 (2.7%)	0.516
No	1976 (96.9%)	99 (99.0%)	1877 (96.8%)
Missing	10 (0.5%)	0 (0.0%)	10 (0.5%)
Chronic bone pain
Yes	342 (16.8%)	32 (32.0%)	310 (16.0%)	< 0.001
No	1687 (82.7%)	68 (68.0%)	1619 (83.5%)
Missing	10 (0.5%)	0 (0.0%)	10 (0.5%)
Chronic headaches
Yes	53 (2.6%)	7 (7.0%)	46 (2.4%)	0.014
No	1976 (96.9%)	93 (93.0%)	1883 (97.1%)
Missing	10 (0.5%)	0 (0.0%)	10 (0.5%)
Anxiety/depression b
Extremely	182 (8.9%)	15 (15.0%)	167 (8.6%)	0.003
Moderately	725 (35.6%)	44 (44.0%)	681 (35.1%)
Not	1102 (54.0%)	38 (38.0%)	1064 (54.9%)
Missing	30 (1.5%)	3 (3.0%)	27 (1.4%)
Bipolar disorder
Yes	248 (12.2%)	17 (17.0%)	231 (11.9%)	0.130
No	1790 (87.8%)	83 (83.0%)	1707 (88.0%)
Missing	1 (0.1%)	0 (0.0%)	1 (0.1%)
Non‐fatal overdose b
Yes	383 (18.8%)	19 (19.0%)	364 (18.8%)	0.962
No	1652 (81.0%)	81 (81.0%)	1571 (81.0%)
Missing	4 (0.2%)	0 (0.0%)	4 (0.2%)

Note: p‐values were calculated using Mann–Whitney U test for the continuous variable age and Pearson's chi‐square test for binary and categorical variables. When counts were expected < 5, Fisher's exact test was reported (i.e., daily sleeping pills use, daily non‐medical benzodiazepine use, chronic muscle pain and chronic headaches).

Abbreviations: HIV, human immunodeficiency virus; IQR, interquartile range; POC, person of colour.

^a Opioid use was defined as unregulated opioid use, non‐medical use of prescribed opioids and/or opioid agonist treatment.

^b In the 6 months prior to the interview date.

^c Defined as daily cocaine, crack cocaine or crystal methamphetamine use.

FIGURE 1.

Prevalence of self‐reported prescription gabapentin for pain among people who use opioids, June 2014 to March 2020. Complete data available in Supporting Information (Table S1).

Bivariate and multivariable factors associated with reporting a gabapentin prescription for pain are presented in Table 2. Variables that were significantly associated with prescription gabapentin in multivariable GEE analyses included: older age (adjusted odds ratio [aOR] = 1.40, 95% CI 1.24–1.59); being engaged in opioid agonist treatment in the last six months (aOR = 1.70, 95% CI 1.31–2.21); ever being diagnosed with a neuropathic pain condition (aOR = 3.54, 95% CI 2.81–4.46), inflammatory pain condition (aOR = 1.22, 95% CI 1.02–1.46) and bone/mechanical/compressive pain condition (aOR = 1.46, 95% CI 1.23–1.74); and ever being diagnosed with bipolar disorder (aOR = 1.62, 95% CI 1.07–2.45). None of the variables assessing substance use behaviours were statistically significant in the GEE models. The study sample had minimal missingness, with 10,854 (98.7%) observations included in the final model. The multivariable GEE had no evidence of multicollinearity with all generalised variance inflation factors ≤ 1.11.

In our analyses of non‐medical gabapentin use and gabapentin prescriptions for alcohol treatment, we included data from 1831 participants that provided 11,281 interviews conducted between December 2016 and May 2024. Overall, 56 (3.1%) participants reported non‐medical gabapentin use at least once during the study period and the prevalence across follow‐up periods remained consistently less than 1.4% (median [IQR]: 0.58% [0.21, 0.74%]) (Figure S1). Receiving prescription gabapentin for the treatment of alcohol use in the last 6 months was reported at least once across the study period by 48 (2.6%) participants or by 1.3% of study visits or less at each six‐month follow‐up period (median [IQR]: 0.74% [0.56, 0.96%]) (Figure S2).

4. Discussion

In our analyses including 2039 participants observed over a seven‐year study period, we observed a steady increase in gabapentin prescriptions for pain, with 255 (12.5%) people who used opioids reporting a prescription for gabapentin during one or more study visits. Gabapentin prescriptions for pain were significantly associated with reporting a chronic neuropathic pain condition, with this subgroup experiencing an apparent twofold increase in the prevalence of gabapentin prescriptions between 2014 and 2020. Results from secondary analyses indicated that gabapentin prescriptions for the treatment of alcohol use disorder and non‐medical gabapentin use were uncommon, with both reported by only a small number of participants between 2016 and 2024.

The increasing rate of gabapentin prescriptions for pain identified in the present study is consistent with other studies describing the increasing proliferation of gabapentin as an off‐label pharmaceutical intervention [4, 13]. This is further reflective of national prescribing trends, with Canada experiencing an estimated 11% average annual increase of gabapentinoid daily doses between 2008 and 2018 [3]. We identified only one previous Canadian study investigating gabapentin prescriptions among people who use drugs. In a cohort including 9964 people who experienced a fatal or non‐fatal overdose in British Columbia between 2015 and 2016, 25% of men and 31% of women had a gabapentinoid prescription within the 5 years prior to the index event [40]. The higher rates of gabapentinoid dispensations in this cohort (compared to the 12.5% prevalence across our entire study period) are likely explained by the inclusion of gabapentinoid prescriptions for all indications (i.e., beyond pain), the grouping of gabapentin and pregabalin and differences in the sociodemographic makeup of the cohorts. A second study in the United States among buprenorphine‐engaged people who experienced a drug‐related poisoning found that 12.7% of people held a gabapentin prescription in the 90 days following buprenorphine initiation [26].

We identified a strong association between neuropathic pain and gabapentin, with participants with a diagnosis over 3.5 times more likely to report receiving a prescription for pain. This was also observed in Ellis et al., where neuropathic pain was associated with a 68% increased risk of having a gabapentin prescription within 90 days of buprenorphine initiation [26]. The observed independent association with inflammatory and bone‐related chronic pain may be evidence of gabapentin similarly being used off‐label in an attempt to address these non‐specific conditions, despite limited evidence of efficacy, or possible misclassification.

Reflective of other studies among members of similar populations, the odds of being prescribed gabapentin for pain were age‐dependent, increasing by 40% for every 10 years of age among the present sample of people who use opioids [26]. Such an association is expected as rates of chronic pain increase with age. Further, these findings are aligned with a previous analysis among a wider sample of cohort participants which found that older adults were significantly less likely to be denied prescription pain medication [41]. The independent association between gabapentin and being engaged in opioid agonist treatment may be attributed to a number of factors. Among the present sample, participants engaged in opioid agonist therapy may have a longer duration or higher intensity of both regulated and unregulated opioid use that could be associated with higher rates of opioid‐induced hyperalgesia (i.e., a hypersensitivity to painful stimuli) [42]. Gabapentin may be used as a non‐opioid analgesic within this context, though further research is needed to explore this possible association. Opioid agonist treatment engagement may also reflect a level of stability that is conducive to being prescribed additional long‐term pharmaceuticals, further facilitated by the regular contact with healthcare providers [43].

The weak association observed between prescription gabapentin and bipolar disorder may in part be due to the latter having previously been considered an unapproved indication for gabapentin. While gabapentin observations in this analysis were specifically identified in relation to pain, it is reasonable to assume that prescriptions are occasionally given in an effort to address multiple concurrent conditions. Future research is needed to confirm whether this association continues to hold as this off‐label indication has lost popularity given minimal evidence of efficacy [5, 44]. Participants reporting bipolar disorder may also be more likely to have anxiety disorders or other mental health conditions that are not explored in this analysis, but are occasionally treated off‐label with gabapentin, especially in instances of co‐occurring pain [45].

Finally, none of the variables assessing patterns of unregulated substance use (e.g., ≥daily fentanyl/unregulated opioid use, ≥daily non‐medical prescription opioid use) were significantly associated with having a gabapentin prescription in the multivariable model. While these null findings should be interpreted with caution, they are worth noting given the potential risks associated with gabapentin‐opioid interactions. The lack of association may be due to non‐disclosure of substance use between study participants and prescribing clinicians, though it is likely also related to the overreliance on biomedical treatments for pain and the lack of access to non‐pharmaceutical therapies in our study context [46]. Additional rigorous studies are needed to understand whether gabapentin prescriptions increase the risk of opioid‐related overdose death among people who primarily use unregulated opioids and other CNS depressant drugs.

To our knowledge, this is the first study reporting on non‐medical gabapentin use among people who use unregulated drugs in Canada. Our findings suggest that use among people who use opioids in Vancouver is substantially lower than rates reported in international observational studies. Among European studies identified in a 2021 review, rates of non‐medical gabapentinoid use ranged between 10% and 21% among people accessing opioid agonist treatment, although follow‐up periods varied and one additional Swiss study found zero non‐medical use [21]. Similarly, literature from the United States has estimated that 9%–43% of people who use unregulated drugs have used gabapentin non‐medically, with significant regional variation [47, 48, 49, 50]. However, inconsistencies in study populations, follow‐up periods, and definitions of non‐medical use limit clear comparison, as does the common grouping of gabapentin with pregabalin under the umbrella of gabapentinoid misuse. Evidence suggests that non‐medical use of gabapentin is more common in the United States, with pregabalin dominating the non‐medical use of gabapentinoids in Europe [21]. These differences are likely due to regional variations in prescribing practices and pharmaceutical coverage, influencing the subsequent non‐medical availability of both medications in their respective continents [3, 21]. With non‐medical gabapentin use only reported by 3% of the cohort participants over the entire study period, the present proportion aligns more closely with population‐level estimates of non‐medical use from the United States and Europe of 0.4%–4.4% [51, 52, 53, 54].

Finally, efforts to characterise gabapentin use within the context of alcohol use treatment revealed low levels of self‐reported prescriptions, with only 2.6% of the participants ever reporting such treatment across the study period. In Canada, gabapentin is recommended off‐label for the management of low‐risk alcohol withdrawal or as a second‐line option in the pharmacotherapy for alcohol use disorder [55]. While present, the low occurrence of self‐reported gabapentin use for alcohol treatment in this sample is not surprising given the relatively low levels of heavy alcohol use compared to other substances [56]. Even among the general population, pharmacotherapy with approved alcohol use treatment (e.g., naltrexone, acamprosate) is underutilised, with less than 3% of potential recipients estimated to receive it [57, 58, 59].

It is important to note that our estimate of non‐medical use may be underestimating the proportion of people who use their prescribed gabapentin in ways or for purposes that are not aligned with the prescription, as this type of use was not explicitly probed. Further, while recall bias may be leading to an underestimation of the prevalence of non‐medical gabapentin use in the present analysis, this limitation is likely also present in the comparable self‐reported observational studies showing higher rates of non‐medical use elsewhere. Additionally, as the present sample was community‐recruited and exists apart from participants' clinical care, self‐reported data around unregulated use may be more reliable than the studies cited above from clinical settings. Social desirability bias is less likely to be as prominent as participants do not have to fear the consequences of such disclosure on their care or the continuation of their prescription medications. Together, these points suggest that the substantially lower rates of non‐medical gabapentin use among people who use opioids in Vancouver are unlikely to be because of bias in the data reporting. Though beyond the scope of the present analysis due to limitations of the cohort instrument (non‐medical pregabalin use was not prompted), future research in Canadian settings may also seek to explore the potential non‐medical use of pregabalin, as it is commonly reported internationally [21].

This analysis has a number of limitations to consider. As an observational study, we emphasise that the statistical associations observed are not evidence of causation. Further, as a non‐random sample of people who use opioids, cohort study findings have limited generalizability to other settings and populations of people who use opioids. Third, we cannot exclude the possibility of unmeasured confounding in our multivariable model, such as having a regular healthcare provider or an unassessed indication for gabapentin. Fourth, our primary outcome of interest was restricted to assessing prescriptions for pain that overlapped with the study visit. Meaning gabapentin prescriptions for indications other than pain and prescriptions received over the six‐month follow‐up window but discontinued before the study visit, were not captured. As such, prevalence rates described above are likely an underestimation of total gabapentin prescriptions among this cohort. Finally, aside from participants' HIV serostatus, all data used are self‐reported, introducing the potential for recall and social desirability bias as discussed above. While previous research has demonstrated self‐reported data from people who use unregulated drugs to be reasonably reliable and valid [60, 61, 62], the outcome variable's self‐reported nature likely led to the underestimation of gabapentin's prevalence for pain. However, overall, this study's use of self‐reported data served as a strength, allowing us to explore associations that are not currently possible in medical records, pharmaceutical dispensation datasets, or other forms of administrative data. These included the ability to specify pain as the indication for the gabapentin prescription and to explore potential associations with various chronic conditions and substance use patterns.

5. Conclusions

In conclusion, consistent with general prescribing trends, we observed an increase in gabapentin prescriptions for pain among a longitudinal, community‐recruited sample of people who use opioids between 2014 and 2020. Further, we observed that the non‐medical use of gabapentin among people who use opioids in Vancouver was relatively uncommon. Given growing prescribing rates, more research is needed to inform clinical guidelines regarding the balance between the possible short‐ and long‐term benefits of gabapentin and its potential risks among people who use opioids.

Author Contributions

Evelyne Marie Piret: conceptualisation, methodology, visualisation, writing – original draft. Heather Palis: supervision, writing – review and editing. Alexis Crabtree: writing – review and editing. JinCheol Choi: formal analysis, writing – review and editing. Kora DeBeck: funding acquisition, project administration, resources, writing – review and editing. Kanna Hayashi: funding acquisition, project administration, resources, writing – review and editing. Michael R. Law: supervision, writing – review and editing. M.‐J. Milloy: supervision, funding acquisition, project administration, resources, writing – review and editing. Each author certifies that their contribution to this work meets the standards of the International Committee of Medical Journal Editors.

Funding

The study was supported by the US National Institutes of Health (NIH) (U01DA038886, U01DA021525). This research was undertaken, in part, thanks to funding from the Canadian Institutes of Health Research (CIHR) Canadian Research Initiative on Substance Misuse (SMN‐139148). The ARYS cohort received support from CIHR (MOP‐286532; SKF‐149507; and PJT‐175162).

Conflicts of Interest

E.M.P. received funding from UBC's Four Year Fellowship and the Cordula and Gunter Paetzold Fellowship. M.‐J.M. reports financial support from the US NIH (U01‐DA0251525). He is also the Canopy Growth professor of cannabis science at the University of British Columbia, a position established through arms‐length gifts to the university from Canopy Growth, a licensed producer of cannabis and the Government of British Columbia's Ministry of Mental Health and Addictions. K.H. holds the St. Paul's Hospital Chair in Substance Use Research and is supported in part by the NIH (U01DA038886), a Michael Smith Foundation for Health Research Scholar Award and the St. Paul's Foundation. M.R.L. received salary support from a Canada Research Chair in Access to Medicines. He has also consulted for Health Canada and Canada's Drug Agency and acted as an expert witness for several labour unions. K.D. is supported by a Dorothy Killam Fellowship from the National Killam Program, an Applied Public Health Chair from the Canadian Institutes of Health Research and the Public Health Agency of Canada (PP7 192591) and in part by NIH (U01DA038886). Funding sources were not involved in the study design, analysis, interpretation of the data, writing or decision to submit the manuscript for publication. The authors declare no conflicts of interest.

Supporting information

Figure S1: Count and proportion of participant observations reporting non‐medical gabapentin use in the last 6 months per interview period, December 2016 to May 2024.

Figure S2: Count and proportion of participant observations reporting gabapentin for the treatment of alcohol use in the last 6 months per interview period, December 2016 to May 2024.

Table S1: Prevalence of self‐reported prescription gabapentin for pain among people who use opioids, June 2014 to March 2020.

Data Availability Statement

Assurances of strict confidentiality given to participants during the consenting process preclude public sharing of datasets.