Introduction

Despite remarkable improvements in HIV prevention and treatment in the last decade, Indigenous peoples in Canada experience a disproportionate burden of HIV/AIDS. In the province of British Columbia, Indigenous peoples comprise approximately 5% of the population, yet accounted for 12.6% of new HIV infections in 2011/2012 [1]. Compared to the non-Indigenous population of Canada, Indigenous peoples experience a 3.6-fold increase in the burden of HIV incidence (9.2 per 100,000 population vs. 29.9 per 100,000 respectively) [2]. As such, evidence-based interventions are urgently required to reduce the burden of HIV/AIDS-associated morbidity, mortality and transmission among Indigenous peoples in Canada.

The benefits of potent antiretroviral therapy (ART) in reducing morbidity and mortality among HIV-positive individuals and curbing onward viral transmission are increasingly well understood [3]. Unfortunately, not all individuals living with HIV/AIDS have benefited equally from these life-saving treatments, including Indigenous peoples and people who use illicit drugs (PWUD) [4,5,6]. In response, various innovative and culturally-tailored programs have been developed in an effort to increase access and adherence to ART among Indigenous individuals living with HIV disease.

Directly-observed therapy (DOT) programmes have been successfully adapted from their original objective of treating tuberculosis to promote adherence to ART among vulnerable populations living with HIV/AIDS across various settings [7,8,9]. In 1999, the maximally-assisted therapy (MAT) programme was developed as a form of DOT that included additional intensive wrap-around services serving Vancouver’s downtown eastside; a post-industrial neighbourhood with a large open drug market and high levels of illicit drug use, poverty and HIV infection [10]. Specifically, MAT-based programmes go beyond DOT to address multiple barriers to care, including, housing instability, substance use and mental health issues faced by HIV-positive individuals. In Vancouver, the programme is located at a community health clinic in the Downtown Eastside, funded by the provincial government and clients are referred by health care providers whom are typically treatment naïve or disengaged from care [10]. For Indigenous clients, the Vancouver Native Health Society refers individuals in need of HIV treatment and support to the MAT programme, while providing continual holistic cultural support [11]. The MAT programme employs a multidisciplinary team of social workers, nurses, pharmacists, and clinicians and unique to this type of program, access to on-site pharmacists. All of the client’s prescription drugs are transferred to the clinic once they are enrolled in the programme, and the pharmacist provides advice on drug–drug interactions, optimal adherence and ensures continuity of access to ART if the client is incarcerated, hospitalized or enters addiction treatment [10]. Additional services include: the provision of complex HIV primary care, meals, arrangement of specialist appointments, referrals to access legal, social, mental health, and addictions services, and, when patients miss appointments, outreach workers will meet them in the community or at their home to deliver medication [12]. While MAT-based programmes have been previously evaluated in relation to adherence among the homeless [12], and injection drug users [13], the impact of MAT among Indigenous individuals living with HIV has yet to be evaluated.

Marginal structural models are an innovative statistical approach used for causal inference in epidemiology. They are particularly useful when a randomized controlled trial may not be ethically justified or when the design of a controlled trial may not be representative of the real-world application of the intervention [14, 15]. In the present study, we used marginal structural models to estimate the effect of being engaged in a MAT-based programme on the likelihood of achieving optimal adherence to ART among a cohort of HIV-positive Indigenous individuals who use illicit drugs in Vancouver, Canada.

Methods

Data for this analysis were collected from the AIDS Care Cohort to evaluate Exposure to Survival Services (ACCESS), an open and ongoing community-recruited prospective cohort study of HIV-positive illicit drug users in Vancouver, Canada [16]. Beginning in 2005, participants were recruited through self-referral and extensive street outreach from Vancouver’s downtown eastside. Individuals were eligible to participate in ACCESS if they were age 18 or older, HIV-seropositive, had used illicit drugs other than, or in addition to, cannabis in the month prior to enrolment, and provided written informed consent.

At baseline and semiannually, participants complete an interviewer-administered questionnaire soliciting information on demographics, drug use patterns, and related risk factors and exposures. At each visit, participants also complete an examination by a study nurse and provide blood samples for serologic analyses. At recruitment, participants provide their personal health number, a unique and persistent identifier issued for billing and tracking purposes to all residents of British Columbia by the government-run universal and no-cost medical system. Information gathered at each interview is augmented by comprehensive information on HIV care and treatment outcomes from the local centralized HIV/AIDS registry, which has been described previously [17]. Specifically, through a confidential linkage, a complete clinical profile of all CD4 T cell counts, HIV-1 RNA viral load (VL) observations and dispensation of specific antiretroviral agents for each participant are obtained. At each study visit, participants are compensated $30. The Providence Health Care/University of British Columbia Research Ethics Board has approved this study.

The study period for this analysis was from December 2005 to November 2016. Consistent with the approach used previously [18], we defined participants as Indigenous if they self-reported Indigenous ancestry at the baseline interview, defined in this context as First Nations, Aboriginal, Inuit, or Métis. We also included people who self-reported being a member of any Indigenous group or nation (e.g., Haida, Iroquois, etc.). Among these individuals, we included all 180 day observation periods which included ≥ 1 day of ART dispensation as indicated through the confidential linkage to the ART dispensary.

The primary outcome of interest was adherence to ART in the previous 180 days. We measured adherence in each 6 month period as the number of days for which ART had been dispensed over the number of days since that individual had first been dispensed ART, capped at 180 days. This quotient was dichotomized into optimal adherence vs. sub-optimal adherence (i.e., ≥ 95 vs. < 95%). We have used this previously validated measure of adherence based on prescription refill compliance extensively [17, 19, 20], and have shown it is strongly and positively associated with viral response and survival [17, 21, 22].

The primary explanatory variable of interest was engagement in a maximally-assisted therapy-based programme (MAT), defined as 180 day observation periods in which dispensary records indicated that ≥ 1 day of ART had been dispensed at a MAT-based programme. We also considered a range of confounding variables that could be associated with either engagement in MAT or adherence to ART, including: age at baseline (per year older); gender (non-male vs. male); residing in the Downtown eastside (yes vs. no); unstable housing, defined as living in a single-room occupancy hotel, homeless shelter, hostel, treatment or recovery house or having no fixed address (yes vs. no); living in a stable relationship, defined as being legally married, common law, or having a regular partner (yes vs. no); level of educational attainment (≥ high school diploma vs. < high school diploma); being incarcerated in the last 6 months, defined as spending at least one night in jail, prison or penitentiary (yes vs. no); injection drug use in the last 6 months (yes vs. no); non-injection drug use in the last 6 months (yes vs. no); employment in the last 6 months, defined as having at least one source of income from a regular or temporary employment or being self-employed (distinguished from non-legal forms of income generation by separate response options) (yes vs. no); enrollment in any type of addiction treatment in the last 6 months (yes vs. no); and CD4+ cell count (per 100 cells/mL). Except for housing status, which referred to current status, all other variables referred to behaviours or exposures in the last 6 months prior to the study interview. For CD4 cell count, we used the median of all observations in the previous 180 days conducted either through the study or by the participant’s physician or, if none, the most recent observation.

As a first step, we compared baseline characteristics of participants who had been exposed to a MAT-based programme versus those who had not, using Chi square test for binary measures and Wilcoxon’s rank sum test for continuous measures. Next, to determine the longitudinal relationship between engagement in MAT and optimal adherence, we used generalized linear mixed-effects models (GLMM), which account for within-individual similarities across repeated measures over the study period. First, bivariable GLMM analyses were performed to calculate the unadjusted odds ratios and 95% confidence intervals of the associations between the explanatory variables of interest and optimal adherence. We then fit a multivariable GLMM to measure the effect of MAT on optimal adherence, controlling for a range of potential confounders, which were fixed to the model. Variables that had a p value less than 0.05 in bivariable GLMM were included in the multivariable GLMM.

Next, to account for the nonrandomized nature of observational data, we employed a marginal structural modeling procedure. In brief, marginal structural models measure the causal effect of a time-dependent exposure in the presence of time-dependent covariates using observational data [14, 23, 24]. In the current study, we sought to estimate the difference in mean adherence that would have been observed if the entire cohort had received the intervention versus if no one had received the intervention. This model assumes that there are no unmeasured confounders, or that, among individuals who are identical with respect to all measured covariates, the observed adherence of individuals who received the intervention is representative of the counterfactual adherence—this is known as the randomization assumption [14].

Behavioural and drug use variables, based on activities in the last 6 months, were treated as time-updated covariates. Further, variables that had missing values were imputed by carrying the most recent observation forward. To ensure that potential confounders occurred before exposure to a MAT-based programme, time-lagged confounder measurements were used. Specifically, inverse probability-of-treatment weights (IPTW) calculated using pooled logistic regression was applied for the marginal structural model. IPTW can be used to adjust for measured confounding and selection bias in our sample by reweighting the dataset. Other models are susceptible to produce biased causal effect estimates when the existing time-dependent confounders are themselves affected by previous treatment [25]. Using IPTW, each individual was assigned a weight inversely proportional to the individual’s probability of receiving the treatment (engagement in MAT or not) that they in fact received [15]. By weighting the dataset, we can estimate the difference of 95% ART adherence that would have been observed between those who engaged in MAT versus those who did not, if MAT had been assigned randomly. This stabilized weighted procedure has been previously described by Hernán and colleagues [23]. Finally, a weighted repeated measures model analysis using generalized estimating equations was employed to estimate the effect of exposure to MAT. All statistical analyses were performed using the SAS software version 9.4 (SAS, Cary, NC). All p values are two sided with a significance level of 0.05.

Results

In total, 354 HIV-positive Indigenous participants who had ≥ 1 day of ART dispensation prior to the end of the study period were included in these analyses. Among these participants, 311 (87.9%) identified as First Nations, 41 (11.6%) as Métis and two (< 1%) as Inuit. Further, 179 (50.6%) participants in our sample identified as male and the median age was 42 (interquartile range [IQR] 35–47). These 354 participants contributed to 3319 observations over the study period, equal to 1819 person-years of observation. The median number of study visits was 9 (IQR 4–14), and the median follow-up time per participant was 63.4 months (IQR 29.1–90.5), with a loss to follow-up of 12.2% over the study period as previously defined [26]. Of note, this rate does not include the 83 (23.4%) participants that died over the study period. Individuals in our sample frequently reported several known barriers to optimal treatment outcomes. For example, 247 (69.8%) were unstably housed or homeless, 256 (72.3%) reported injection drug use, and 40 (11.3%) had been recently incarcerated at their baseline study visit. Figure 1 depicts self-reported drug use patterns among our sample over the study period.

Fig. 1
figure 1

Drug use patterns among Indigenous people living with HIV in Vancouver, Canada from Dec 2005–Nov 2016 (n = 354)

Full size image

At baseline, 24 (68.6%) of participants in MAT achieved ≥ 95% adherence compared to 160 (50.2%) not in MAT (p = 0.038). Among the 55 participants that were engaged in the MAT programme (n = 35 at baseline and 20 additional participants over follow-up), based on a participant’s total number of study visits, 40.0% were engaged in MAT ≥ 80% of the time; 10.9% were engaged in MAT 79–60% of the time; 18.2% were engaged 59–40% of the time; 18.2 ×% were engaged 40–20% of the time; and 12.7% were engaged in MAT < 20% of the time over the study period. See Fig. 2 for the proportion of the sample that was engaged in MAT 100% over the study period. Baseline characteristics of the study sample, stratified by exposure to a MAT-based programme in the 180 day period prior to the first study interview are presented in Table 1. The results of the bivariable and multivariable GLMM are shown in Table 2. In multivariable analysis, engagement in MAT was found to be significantly and positively associated with increased odds of achieving optimal adherence to ART, after controlling for factors including age, any illicit drug use (i.e., injection or non-injection, other than cannabis), enrollment in addiction treatment, and CD4+ cell count (Adjusted odds ratio [AOR]  = 4.92; 95% confidence interval [CI] 3.18–7.62). After employing a marginal structural model with the same covariates as above and adjusting for the stabilized weights in generalized estimating equations (Table 2), engagement in MAT was found to be strongly associated with the likelihood of achieving optimal adherence to ART (AOR = 5.76; 95% CI 3.34–9.96).

Fig. 2
figure 2

Percentage of sample engaged in MAT 100% of the time from Dec 2005–Nov 2016

Full size image
Table 1 Baseline sociodemographic characteristics and substance use behaviours stratified by exposure to a MAT programme among Indigenous individuals with ≥ 1 day of ART dispensation in the previous 180 days (n = 354)
Full size table
Table 2 Bivariable and multivariable results of the GLMM and marginal structural modeling for factors associated with ≥ 95% adherence to ART among Indigenous individuals with ≥ 1 day of ART dispensation in the previous 180 days (n = 354)
Full size table

Discussion

In this study, we observed a strong, independent and longitudinal link between engagement in a MAT-based programme and the achievement of optimal adherence among Indigenous people living with HIV/AIDS. Our findings stand in strong contrast to many earlier studies conducted across Canada, in which treatment and care systems were unable to deliver comparable HIV/AIDS treatment outcomes between Indigenous and non-Indigenous people living with HIV/AIDS [5, 6, 27]. Previous research has demonstrated that Indigenous populations do not, or are unable to, access HIV testing, care or be retained in treatment with the same frequency as non-Indigenous people [27, 28]; while multiple qualitative studies have linked the prevalence of institutional and provider distrust in HIV testing and care with the long history of discriminatory policies and practices of the Canadian government [29, 30]. Indeed, scholars have associated the generally poor health outcomes among Indigenous peoples to a health care system founded on systemic racism and colonialism [31]. As such, studies of HIV treatment outcomes among Indigenous peoples, including our own, should be interpreted within the context of the social and structural determinants of health including continued social, economic, political and cultural marginalization [32,33,34,35,36].

Previous studies have demonstrated that directly observed therapy-based programmes can improve HIV treatment adherence and virologic outcomes among HIV-positive people who use drugs, primarily through interventions nested within methadone programmes for opioid-dependent individuals [9, 37]. The current results add to this evidence by describing superior adherence levels among Indigenous individuals exposed to a maximally-assisted therapy-based intervention. When the MAT programme was established in 1999, Vancouver was experiencing an escalating rate of HIV infection among many populations, including Indigenous residents. Earlier studies conducted in our setting prior to the beginning of efforts to scale-up access and adherence to antiretroviral therapy in 2004 [26], found that Indigenous peoples were less likely to initiate ART [38], and more likely to die from AIDS without having ever received ART [39]. However, as Milloy et al. demonstrated in a recent study conducted in Vancouver from 2005 to 2014, no statistically significant differences were found between Indigenous and non-Indigenous people living with HIV on several key treatment outcomes; including, access to ART, viral load, and achieving optimal adherence to ART [18]. These findings, in conjunction with ours, suggest that MAT-based programmes were part of a comprehensive policy response to improve treatment outcomes among Indigenous individuals experiencing multiple barriers to care. As rates of new HIV infections on numerous reserves in Saskatchewan have reached levels equivalent to areas with generalized epidemics of HIV/AIDS [40], our findings suggest that the scale-up of MAT-based programmes may prove to be an important factor in responding to HIV among Indigenous peoples in Canada.

Our study has several limitations. As in all observational studies, our sample was not selected at random and our findings may not be representative of other vulnerable HIV-infected populations. We also recognize that many of our measures rely on self-report and are vulnerable to socially desirable reporting, although we note that in the current study both the outcome and primary explanatory variables of interest were gathered from administrative sources of data. Finally, there is increasing awareness of the importance of Indigenous-led systems to identify Indigenous individuals, especially in government and research data, to aid in decolonization, self-government and reconciliation efforts. Our analysis relied on self-reported status, which might under- or overstate important aspects of indigeneity in this setting. In future, all researchers should seek to use a standard approach to identifying Indigenous persons, such as the aboriginal administrative data standard [41].

In conclusion, this study demonstrated that engagement in the MAT programme significantly increased the odds in achieving optimal adherence for HIV treatment among Indigenous individuals living with HIV who use illicit drugs. Given the public health concerns regarding HIV transmission and the repeated calls for improved treatment options for Indigenous peoples living with HIV/AIDS, our study findings indicate that MAT-based programmes may be an effective part of the HIV care cascade and that efforts to scale-up similar programs in other settings are warranted.