Characteristics of the HIV cascade of care and unsuppressed viral load among gay, bisexual and other men who have sex with men living with HIV across Canada’s three largest cities

Abstract Introduction Treatment as prevention strategies have been variously applied across provinces in Canada. We estimated HIV care cascade indicators and correlates of unsuppressed viral load (VL) among gay, bisexual and other men who have sex with men (GBM) recruited in Vancouver, Toronto and Montreal. Methods Sexually active GBM, aged ≥16 years, were recruited through respondent‐driven sampling (RDS) from February 2017 to August 2019. Participants completed a Computer‐Assisted Self‐Interview and tests for HIV and other sexually transmitted infections (STIs). We conducted bivariate analyses comparing RDS‐adjusted proportions across cities. We used multivariable logistic regression to examine factors associated with having a measured VL ≥ 200 copies/mL with data pooled from all three cities. Results We recruited 1179 participants in Montreal, 517 in Toronto and 753 in Vancouver. The RDS‐adjusted HIV prevalence was 14.2% (95% CI 11.1 to 17.2) in Montreal, 22.1% (95% CI 12.4 to 31.8) in Toronto and 20.4% (95% CI 14.5 to 26.3) in Vancouver (p < 0.001). Of participants with confirmed HIV infection, 3.3% were previously undiagnosed in Montreal, 3.2% undiagnosed in Toronto and 0.2% in Vancouver (p = 0.154). In Montreal, 87.6% of GBM living with HIV were receiving antiretroviral therapy (ART) and 10.6% had an unsuppressed VL; in Toronto, 82.6% were receiving ART and 4.0% were unsuppressed; in Vancouver, 88.5% were receiving ART and 2.6 % were unsuppressed (p < 0.001 and 0.009 respectively). Multivariable modelling demonstrated that participants in Vancouver (adjusted odds ratio [AOR]=0.23; 95% CI 0.06 to 0.82), but not Toronto (AOR = 0.27; 95% CI 0.07 to 1.03), had lower odds of unsuppressed VL, compared to Montreal, as did older participants (AOR 0.93 per year; 95% CI 0.89 to 0.97), those at high‐risk for hazardous drinking (AOR = 0.19; 95% CI 0.05 to 0.70), those with a primary care provider (AOR = 0.11; 95% CI 0.02 to 0.57), and those ever diagnosed with other STIs (AOR = 0.12; 95% CI 0.04 to 0.32). Conclusions GBM living in Montreal, Toronto and Vancouver are highly engaged in HIV testing and treatment and all three cities have largely achieved the 90‐90‐90 targets for GBM. Nevertheless, we identified disparities which can be used to identify GBM who may require additional interventions, in particular younger men and those who are without a regular primary care provider.

federal government in Canada, policies and programmes to address HIV and other STI prevention, testing, care and treatment vary across provinces.
In 2014, UNAIDS formally proposed targets for HIV-related health services, known as 90-90-90, whereby countries should aspire to have 90% of people living with HIV (PLWH) aware of their diagnosis, 90% of these individuals should be receiving ART and 90% of these should have achieved virological suppression [11]. The government of Canada formally adopted these targets in 2014 and estimated then that approximately 80% of PLWH had been diagnosed, of whom 76% were receiving ART and 89% of these had achieved virological suppression [12]. Sustained virological suppression is the ultimate goal of the care cascade in that it maximizes the benefits of ART for PLWH [13] and eliminates the risk of onward HIV transmission [5]. Previous research has identified sociodemographic factors such as age, ethnicity and income [14,15], as being associated with virological suppression, as well as measures of mental health and substance use [15][16][17].
We designed a study to compare the GBM populations in Montreal, Toronto and Vancouver, by recruiting representative community-based samples. We estimated HIV care cascade indicators across all three cities and examined factors associated with having an unsuppressed viral load (VL). Our primary interest was to examine if there were differences in HIVrelated outcomes for GBM populations in these cities and whether access to specific health services were associated with achieving virological suppression.

| METHODS
Participants were recruited using respondent-driven sampling (RDS) in Montreal, Toronto and Vancouver from 1 February 2017 to 31 August 2019. RDS is a formalized chain recruitment method [18] by which participants are recruited through their own social networks. To be eligible GBM needed to identify as a man (cis-or transgender), be aged ≥16 years old, report having sex with another man in the previous six months and be able to complete a survey in either English or French (if a participant from Montreal) and agree to study procedures, including biological sampling. Furthermore, participants were required to either have been invited into the study as a "seed" participant or be in receipt of a valid invitation (see below) obtained from another participant whom they knew personally. The initial targeted sample size was 720 participants per site, in order to provide comparable data for the Vancouver site with a previous RDS-conducted study [16]. However, the Montreal site received additional funding from the Quebec government to increase its sample size to 1200 participants. All study sites began enrolment by purposely recruiting 30 seeds to initiate recruitment and study protocols allowed for the recruitment of additional seeds to maintain steady recruitment. Seeds were initially recruited through members of our Community Engagement Committees. As well. advertisements on mobile phone applications and websites used by GBM were also used to recruit additional seed participants in Toronto and Vancouver. Of the 30 initial seeds selected in each city, at least 10 were men living with HIV, 10 self-identified as ethnic minorities, two self-identified as transgender and two as bisexual. We also tried to recruit at least two seeds who were below the age of 18 years All seeds and subsequent recruits were provided with six invitations to offer to other GBM whom they knew personally. Invitations were either provided as a printed card or an electronic version which could be sent by text or email. Invitations had a unique identifier number to ensure that they could only be redeemed by one individual. Participants were provided with a $50 CAD honorarium for study participation and $15 CAD for each individual they recruited to participate in the study. The study samples in each city were monitored for equilibrium for key parameters including age, gender-identity, sexual orientation, ethnicity, being born in Canada, income group and HIV serostatus.
Potential participants were screened for eligibility with a short questionnaire at each of the study offices (one in each city), and, if eligible, underwent informed consent procedures. Participants then completed a Computer-Assisted Self-Interview (CASI) and met with a study nurse to undergo a rapid HIV test and provided venous blood samples for serological tests for HIV, hepatitis C virus (HCV) and syphilis. For participants known to be living with HIV, we offered them the option of either confirming their diagnosis with a point-of-care or HIV laboratory test, or by requesting confirmation from their primary care physician. For these individuals or those who were diagnosed with HIV at the visit, the nurse also drew blood for CD4 cell counts and HIV RNA VL. The CASI collected information regarding sociodemographics, health status, use of health services and individual characteristics, including reported substance-use. Mental health symptoms were measured by the Hospital Anxiety and Depression Scale (HADS) [19], using both the anxiety (a = 0.84) and depression (a = 0.76) sub-scales. Alcohol use was measured by the Alcohol Use Disorders Identification Test-Concise (AUDIT-C) scale [20]. Study protocols were approved by institutional review boards of Research Institute of the McGill University Health Centre, Ryerson University, The University of Toronto, St. Michael's Hospital, the University of British Columbia and the University of Victoria.
We conducted bivariate analyses comparing RDS-adjusted proportions across cities using RDS-II weights [21], using Pearson's chi-squared test for categorical variables and Type III Wald chi-squared test (from a weighted logistic regression) for continuous variables. RDS-II weights, which are inversely proportional to the size of participants' social network, were based on the question "How many men who have sex with men aged 16 years or older, including trans men, do you know who live or work in the [Metro Vancouver/Greater Toronto/ Metro Montreal] area?" An upper limit of network-size was fixed at 150 based on empirical research by Dunbar et al [22].
Among individuals self-reporting as HIV negative or unknown, we compared reports of their most recent HIV test and the number of tests in the previous year. For confirmed PLWH, we compared the proportions previously undiagnosed, the proportions currently receiving HIV treatment and the proportions with a plasma VL measured <200 copies/mL [23] across the three cities. We also compared sociodemographic, health service access, substance-use behaviour and mental health symptoms among PLWH across cities. We then conducted a pooled analysis of all PLWH across all three cities using multivariable logistic regression to examine factors associated with having an unsuppressed VL (≥200 copies/mL) [23].
Variables were weighted using the RDS II-derived weights from each city and city was used as a potential explanatory variable in the model. Variables of interest in the univariable models with a value of p < 0.2 were included for consideration in the multivariable model. For several variables, including age, AUDIT-C scores and HADS scores, we considered both continuous and categorical variables for inclusion in the final model. Only observations with complete data were included in the multivariable model. There is no consensus as to whether RDS weights should be used when conducting regression analyses in RDS studies [24]. However, an analysis conducted by members of our study team found that weighted logistic regression methods consistently outperform unweighted methods in terms of bias and precision when explanatory variables are correlated with the network size reported by participants [25]. As such, we have chosen to apply RDS weights to our regression analysis. The final model was selected based on potential candidate variables from the literature and using a backward selection technique, whereby the least significant (i.e., highest Type III p-value) variable was dropped until the final model reached the optimal (minimum) Akaike Information Criterion. All analyses were performed using SASâ Version 9.4 (SAS Corporation, Cary, NC, USA).

| RESULTS
We recruited a total of 1179 participants in Montreal, 517 in Toronto and 753 in Vancouver, from 27, 96 and 117 seed participants respectively. A total of 6822 invitations were issued to participants in Montreal, 3078 in Toronto and 4424 in Vancouver. The mean, median and ranges of recruitment waves were 6.67, 6 and 0 to 17 in Montreal, 2.67, 2 and 0 to 12 in Toronto and 2.67, 2 and 0 to 10 in Vancouver. Further details on RDS-specific recruitment outcomes are published elsewhere [26]. The study samples reached equilibrium for all of the key variables for which we monitored. The RDS-adjusted HIV prevalence based on serological testing or documentation was 14.2% (95% CI 11.1 to 17.2) in Montreal; 22.1% (95% CI 12.4 to 31.8) in Toronto, and 20.4% (95% CI 14.5 to 26.3) in Vancouver (p < 0.001) ( Table 1). Of participants who were found to be HIV negative at enrolment, 70.4% (95% CI 65.0 to 75.7) in Montreal reported having tested for HIV in the previous year, 67.5% (95% CI 57.5 to 77.6) in Toronto and 69.4% (95% CI 61.9 to 76.9) in Vancouver (p = 0.010). Among participants with confirmed HIV infection, 3.3% were previously undiagnosed in Montreal, with 3.2% undiagnosed in Toronto and 0.2% undiagnosed in Vancouver (p = 0.154). In Montreal, 87.6% of PLWH were receiving ART and 10.6% had an unsuppressed VL; in Toronto, 82.6% were receiving ART and 4.0% were unsuppressed; in Vancouver, 88.5% were receiving ART and 2.6% were unsuppressed (p < 0.001 and = 0.009, respectively). Notably, nine GBM living with HIV in Montreal, three in Toronto and six in Vancouver were found to have measured VLs < 200 copies/ mL despite reporting that they were not receiving ART.
Among PLWH we found differences in the age distribution across the three cities (Table 2), with Montreal having the highest RDS-adjusted proportion of participants ≥45 years (73.5%; 95% CI 64.0 to 83.0) and the lowest proportion aged <30 (9.2%; 95% CI 1.9 to 16.5) (p < 0.001). Montreal had the highest proportion of participants who self-identified as Canadian (69.7%; 95% CI 59.2 to 80.2; p < 0.001) and whom were born in Canada (77.6%; 95% CI 67.8 to 87.5; p = 0.003) with Vancouver having the lowest proportion reporting Canadian ethnicity (41.8%; 95% CI 26.3 to 57.4; p < 0.001) and Toronto having the lowest proportion born in Canada (62.1%; 95% CI 36.2 to 88.0; p = 0.003). In terms of health services, the proportions of PLWH who reported having access to a family doctor, ranged from 90.3% in Montreal to 99.4% in Toronto (p < 0.001). The median number of times participants had been tested for an STI in the past 2 years was highest in Toronto (5; Interquartile range [IQR] 3 to 8) and lowest in Vancouver (median of 2; IQR 1 to 5, p = 0.022), but differences in the proportion of participants who had ever been diagnosed with an STI (other than HIV) were not statistically significant (range of 80.3% to 88.4% across sites; p = 0.224).
In terms of substance use, Montreal had the highest proportion of participants classified at high risk for heavy drinking or abuse/dependence (40.5%; 95% CI 29.2 to 51.7) (p = 0.032), whereas Vancouver had the highest proportion of participants reporting methamphetamine use in the past six months (36.9%; 95% CI 21.5 to 52.2) (p < 0.001). Vancouver also had the highest proportion of reported opioid use in the previous six months, compared with 2.3% (95% CI 0.0 to 4.7) in Toronto and 10.3% in Montreal (95% CI 3.8 to 16.8) Among all 421 PLWH in the three cities with valid VL results, 27 (6.4% non-RDS adjusted) had measured VLs > 200 copies/mL. Our multivariable model (Table 3), based on 398 observations with complete data found that older age was associated with lower odds of unsuppressed VL (adjusted odds ratio [AOR] 0.93 per year; 95% CI 0.89 to 0.97), as was residing in Vancouver (AOR = 0.23; 95% CI 0.06 to 0.82) relative to Montreal. Toronto also had a lower odds compared to Montreal, but this was not statistically significant (AOR = 0.27; 95% CI 0.07 to 1.03). Having a primary care provider (AOR = 0.11; 95% CI 0.02 to 0.57), and ever being diagnosed with another STI (AOR = 0.12; 95% CI 0.04 to 0.32) also had lower odds of an unsuppressed VL. Participants with high-risk scores AUDIT-C scores, also had lower odds of having an unsuppressed VL (AOR = 0.19; 95% CI 0.05 to 0.70). No measures of mental health symptoms or diagnoses were retained in the final multivariable model.

| DISCUSSION
Our RDS-weighted analysis has demonstrated that GBM in Montreal, Toronto and Vancouver are highly engaged in HIV testing and treatment. While we did find some statistically significant differences across the three cities, the magnitude of these differences was quite small. We found that PLWH in Vancouver had lower odds of having an unsuppressed VL, compared with Montreal, but even in Montreal, only 10% of PLWH (whether diagnosed or not) had VL measurements ≥200 copies/mL. This suggests that policies and services to engage GBM at risk for acquiring HIV and those living with HIV have been largely successful across all three cities. In terms of the UNAIDS targets for this population, 97% to 99.8% of GBM living with HIV in our study were diagnosed, 82% to 89% of those were receiving treatment, and of those receiving ART, 94% to 99% had a suppressed VL. While the ART uptake values were all below 90%, we suspect that this may be due to underreporting of treatment status, as 18 participants were found to have a suppressed VL despite reporting not being on treatment.
Between 67.5 and 70.4% of HIV-negative or unknown serostatus participants reported having been tested for HIV in the past year, and only 7% to 12% of participants reported never being tested. Previous research has shown that, in Vancouver, less frequent HIV testing among GBM is associated with less HIV risk behaviour [27]. As such, previously undiagnosed HIV infections were very rare in the current study, ranging 0.2% of PLWH in Vancouver to 3% in Montreal and Toronto. These proportions of undiagnosed infections are far lower than the national estimate of undiagnosed HIV infections (13%) released by the Public Health Agency of Canada (PHAC) in 2018 [28]. However, these PHAC estimates do not disaggregate by HIV exposure category and no primary data collection which could be used to measure the undiagnosed fraction in GBM has been published in Canada in more than 10 years [29]. While it is reasonable to assume that GBM living in these metropolitan areas are perhaps more engaged in HIV and STI care than other GBM in these provinces, it is also important to note that HIV is highly concentrated in the urban cores of these three cities in each province, with over 70% of HIV diagnoses found in these metropolitan areas being among GBM [30][31][32]. Notably PHAC estimates for the proportion of all PLWH receiving ART (85%) and those with virological suppression (94%) [26] are similar to what we found in our study, for GBM living with HIV.
When examining the pooled analysis of GBM living with HIV with unsuppressed VL, we found that, similar to other North American studies, younger GBM living with HIV were less likely to have a suppressed VL [14,15]. Of note, unlike recent studies from the United States [15,17], but similar to another study from Canada [14], we did not find that ethnicity was    significantly associated with lack of viral suppression in this analysis. However, we had limited statistical power to examine associations with any single minority group, and our heterogeneous comparison of European/Canadian to all other-identifying GBM may mask differential experiences given the diversity of minority groups that were represented in our sample. With respect to health services, having a primary care provider and ever having been diagnosed with an STI were both associated with reduced odds of unsuppressed VL. In all three cities, much HIV care is provided by family physicians, who also manage other acute and chronic conditions. Even in circumstances where specialists may be providing HIV-specific care, having a primary care provider likely facilitates a continuity of healthcare that benefits of PLWH. Similarly, having been previously diagnosed with an STI likely reflects access and motivation to undertake STI testing, where questions regarding HIV treatment and adherence may also be addressed at the same time. As well, men who have been previously diagnosed with an STI may be more motivated to adhere to ART recognizing that they may be at greater risk of onward transmission if not adherent to their ART.
Other studies have found that syndemic factors related to mental health and substance use are associated with a lack of HIV virological suppression among GBM [15,17]. In our study, we did not find independent associations between measures of mental health symptoms or diagnosed disorders and lack of VL suppression, although some measures were associated with this outcome in univariable analyses. This suggests that other factors in our model, possibly having access to a family doctor or older age may explain the lack of these associations in our final model.
We also did not find associations with recent use of methamphetamine, cocaine, ecstasy, opioids or any injection drugs and our outcome. We did, however, find an association with high-risk AUDIT-C scores and a reduced odds of unsuppressed VL, which was unexpected. In a systematic review of the effects of alcohol-use disorder (AUD) on HIV treatment adherence among PLWH in general (not only GBM), five of seven prospective cohort studies found an association between AUD and reduced adherence to treatment and five other studies have found associations between AUD and increased VL [33]. It is worth noting that not all studies which have examined these outcomes have found associations between AUD and HIV-related outcomes [34] and there are wide variations as to how alcohol use and treatment adherence are measured, but we are unaware of other studies showing that high-risk consumption may be protective. GBM who had scores for high-risk drinking on the full AUDIT scale in Vancouver have been shown to have higher levels of social support and are more likely to read gay newspapers or community magazines [30]. This may lead such men who are living with HIV to be more aware and more motivated to seek HIV treatment and adhere to their prescribed regimens.
This study has a number of strengths, as well as several limitations. First, we used RDS as our recruitment strategy to overcome of some of the limitations of generating representative research samples of GBM in Canadian cities, which have typically used clinic-based samples, convenience samples, or on-line surveys or time-location sampling from venues or events [29]. Furthermore, the study teams in each city used harmonized study protocols with the same general recruitment strategies, study procedures and data collection tools. Nevertheless, study implementation varied by city, with Montreal able to recruit a larger sample, with fewer seed participants and more waves of recruitment than Toronto and Vancouver. As such, some of the differences we have observed between the city samples may be due to differences in the characteristics arising from differences in implementation. We have conducted analyses to identify differences in sociodemographic characteristics or motivations for study participation among participants in the study across the three cities [35], but have been unable to identify reasons as to why the RDS process was more successful in Montreal than the other two cities As well, conducting pooled regression analyses using RDS-recruited samples, violates one of the key assumptions of RDS [21], in that the participants are drawn from a distinct networked population. However, our pooled analysis did not seek to ascertain pooled prevalence estimates; instead, we sought to examine correlates of unsuppressed viral load.

A C K N O W L E D G E M E N T S
We thank the Engage study participants, office staff and community engagement committee members, as well as our community partner agencies.