Progress towards the United Nations 90‐90‐90 and 95‐95‐95 targets: the experience in British Columbia, Canada

Abstract Introduction Antiretroviral therapy (ART) scale‐up is central to the global strategy to control the HIV/AIDS pandemic. To accelerate efforts towards ending the AIDS epidemic, the Joint United Nations Programme on HIV/AIDS released the 90‐90‐90 and 95‐95‐95 targets, which have recently been approved by the United Nations (UN). This study characterizes the province of British Columbia (BC)'s progress towards achieving the UN targets, predicts a trajectory up to 2030 according to each of the individual steps (i.e. %Diagnosed, %On ART and %Virologically Suppressed), and identifies the population sub‐groups at higher risk of not achieving these targets. Methods The analyses were based on linked individual‐level datasets of people living with HIV (PLWH) in BC, aged ≥18 months, from 2000 to 2013. Using past trends in HIV prevalence and of each individual UN targets, we forecasted these outcomes until 2030 via generalized additive models. We ran a second set of analyses to assess the associations between individual demographic and behavioural factors and each of the individual steps of the UN targets. Lastly, we performed sensitivity analyses to account for uncertainty associated with prevalence estimates and suppression definitions. Results Among the estimated 10666 PLWH in BC in 2013, 82% were diagnosed, 76% of those diagnosed were on ART and 83% of those on ART were virologically suppressed. We identified that females, PLWH aged <30 years and those with unknown risk or who self‐identify as having a history of injection drug use were the population subgroups that experienced the most challenge in engaging on ART and achieving viral suppression. Our model projections suggest that BC will achieve 90%‐91%‐90% and 97%‐99%‐97% by 2020 and 2030 respectively. Conclusions As we approach 2020, BC is rapidly moving towards achieving the UN targets. However, region‐specific challenges persist. Identification of remaining regional challenges will be essential to achieving the proposed UN targets and therefore fulfilling the promise to end AIDS as a pandemic by 2030.


| INTRODUCTION
Expanded access to antiretroviral therapy (ART) has a significant role to play in the global strategy to control the HIV/ AIDS pandemic and alleviate its impact on people living with HIV (PLWH) [1]. There is ample scientific evidence suggesting that early ART initiation not only reduces morbidity and mortality, but also reduces incidence rates of HIV, resulting in the concept of "Treatment as Prevention (TasP)" [2]. Definitive evidence of the effect of ART on HIV transmission via heterosexual contact was provided by the HTPN 052 randomized control trial [3]. This finding was verified among men who have sex with men in the PARTNER study [4], and further supported by a number of observational and modelling population-based studies [5][6][7].
In order to achieve maximum benefits from the expansion of TasP, it is crucial that PLWH are diagnosed early in the course of their HIV infection, linked and retained in HIV care, offered treatment at diagnosis, and highly adherent to attain and sustain full virologic suppression [8]. As a result, monitoring engagement in the "HIV cascade of care" is now widely recognized as a critical element of HIV surveillance, aimed at ensuring PLWH can fully benefit from existing management strategies, including immediate access to ART to prevent disease progression to AIDS and premature death, and to minimize the risk of ongoing HIV transmission.
Controlling the global HIV pandemic will require renewed political will [2]. At the 2014 International AIDS Conference in Melbourne, the Joint United Nations Programme on HIV/ AIDS proposed the 90-90-90 and 95-95-95 targets to accelerate efforts towards ending the AIDS epidemic by 2030, which have recently been approved by the United Nations (UN) [9,10]. The 90-90-90 and 95-95-95 targets proposes that by 2020 and 2030, respectively, at least 90% and 95% of all PLWH should be diagnosed, at least 90% and 95% of those diagnosed should be on ART and at least 90% and 95% of those on ART should be virologically suppressed, translating into a final target of at least 73% and 86% of all PLWH achieving virologic suppression. UN Modelling has suggested that the full implementation of the 90-90-90 target could lead to an approximate 90% decrease in AIDS-related deaths and an approximate 90% decrease in new HIV infections by 2020, compared with the status quo using 2010 as the baseline [9].
In 2003, the province of British Columbia (BC), Canada, initiated and progressively scaled up efforts to expand HIV testing and publicly funded ART access, as part of a province-wide initiative to decrease HIV-related morbidity and mortality as well as HIV transmission [11,12]. This study characterizes BC's progress towards achieving the 90-90-90 and 95-95-95 targets, predicts a trajectory up to 2030 according to each of the individual steps (i.e. %Diagnosed, %On ART and %Suppressed), and identifies the population sub-groups at higher risk of not achieving these targets.

| Data sources
Eligible study participants were recruited from the BC Seek and Treat for Optimal Prevention of HIV/AIDS populationbased cohort (STOP HIV/AIDS), which contains longitudinal individual-level data on all PLWH ever diagnosed in the province. This population-based cohort is derived from various linkages among provincial databases (see Data S1) [11, [13][14][15][16][17][18]. The BC Centre for Excellence in HIV/AIDS (BC-CfE) is responsible for distributing ART, at no cost and without copayments, to all PLWH in the province, and its database captures HIV plasma viral load (pVL) results for all its participants. Previous studies have described these databases and the respective linkages [11,19]. Annual HIV prevalence estimates for BC from 2000 to 2013 were derived from the Public Health Agency of Canada (PHAC) available estimates of the percent of PLWH who were unaware of their infection, and the number of individuals diagnosed based on the STOP HIV/ AIDS cohort [20][21][22]. For more detail, please see Figure S1.

| Outcomes and statistical analysis
The inclusion criteria consisted of all diagnosed PLWH enrolled in the STOP HIV/AIDS cohort, aged ≥18 months old, in each calendar year from 2000 to 2013 in BC. The definition for each of the steps of the 90-90-90 or 95-95-95 targets, for each calendar year, is as follows: • %Diagnosed (among all PLWH in BC), the first instance of one of the following: a confirmed HIV-positive test, a detectable pVL >50 copies/ml, an HIV-related physician visit or hospitalization, a reported AIDS-defining illness or the dispensation of antiretroviral treatment during the calendar year. This definition has been previously validated [11]. • %On ART (among PLWH who have been diagnosed): receiving antiretroviral treatment at least three months apart within the calendar year [11]. • %Suppressed (among PLWH on ART): In our setting, multiple pVL tests are typically available per patient within a calendar year (median number of pVLs per year per patient was 4 (25th to 75th percentile three to five or six depending on the year)). As such, we used a conservative definition that utilizes all available pVLs from each patient, i.e. those who had an undetectable pVL test (<200 copies/ml) at all times tested in a given calendar year.
We modelled the non-linear trend in HIV prevalence and each of the individual steps of the UN targets via generalized additive models [23][24][25]. When modelling the HIV prevalence, we assumed a negative binomial distribution and the log link function, and for the proportions for each of the individual targets, we assumed a beta distribution and the cloglog link function (chosen after examining the data and goodness of fit statistics). In brief, we used a cubic regression spline to smooth the trend in these outcomes in which calendar year was the explanatory variable being smoothed. Additionally, different models were fitted based on the number of knots used to smooth the trend in all outcomes when the estimation was via restricted maximum likelihood. We also used a built-in smoothing parameter estimation method "GCV.Cp" to estimate the number of knots in each of the models. To assess the goodness of fit of each model, R outputs the adjusted R 2 , the percentage of the deviance explained, a test to check the appropriateness of the number of knots in the model, and the sum of the deviance residuals. Subsequently, we forecasted the evolution of the HIV prevalence and each target from 2014 to 2030. In these analyses, we used the statistical software R© version 3.2.2 mgcv library (functions gam, summary, gam.check, residuals.gam and forecast). All p-values were twosided and the significance level was set at 5%.
We ran a second set of analyses to assess the associations between individual demographic and behavioural factors and each of the individual UN targets. The binary outcome was classified as yes (i.e. successful in achieving a particular target) vs. no (i.e. failing to achieve a particular target). The possible explanatory variables, measured at diagnosis, were calendar year (continuous), age (<18, 18 to 29, 30 to 39, 40 to 49 and ≥50 years), gender (male/female) and HIV risk group (gay, bisexual and other men who have sex with men (MSM); history of injection drug use (IDU); MSM/IDU; other; unknown). Multivariable explanatory logistic regression models were built to assess which of these factors were associated with each of the individual steps of the UN targets. A modified backward stepwise technique developed and validated by our group, based on the Akaike Information Criterion and Type III p-values, was used in the selection of explanatory variables [26]. Categorical variables were compared using the Cochran-Mantel-Haenszel test, proportions were compared using a two proportion z-score test, and trends in counts were tested using a Poisson regression [27,28]. These analyses were performed using R© 3.2.2 MASS library and SAS, version 9.4 (SAS, Cary North CA, USA).

| Sensitivity analyses
There are several potential sources of uncertainty associated with HIV prevalence estimates. Thus, we performed a sensitivity analysis considering different scenarios of prevalence estimates. Since the percentages of the "on ART" and "virologically suppressed" steps of the UN targets remained unchanged irrespective of the prevalence estimates, we calculated the percentage of diagnosed PLWH for three scenarios of prevalence estimates utilizing the original PHAC estimates, and by decreasing these prevalence estimates by 2% and 5%. The second sensitivity analysis focused on the definition of viral suppression. In this case, we ran these analyses utilizing PHAC's definition, which is less conservative, and defines the %Suppressed (among PLWH on ART) as patients whose latest pVL test was undetectable (<200 copies/ml) in the calendar year.

| Ethical considerations
The University of BC Ethics Review Committee at the St. Paul Hospital site provided ethics approval for this study (H08-02095; H05-50123).

| Characteristics of PLWH ever diagnosed with HIV
Since 2000, there was a total of 12976 PLWH ever diagnosed with HIV in the province. The majority of these individuals were male (80%), aged 30 to 49 years (64%) and self-identified either as a MSM (29%) or as an IDU (27%, Table 1). Table 2 present demographic and behavioural factors associated with each of the individual steps of the UN targets, excluding diagnosed, given that all STOP HIV/AIDS participants have been diagnosed in compliance with the cohort entry criteria. We observed that, in comparison to females, males were 43% more likely to be on ART and 26% more likely to achieve virologic suppression. In terms of age, in comparison to PLWH aged 30 to 39 years, individuals <18 and 18 to 29 years were less likely to be on ART and to achieve virologic suppression, whereas those ≥40 years performed better in both outcomes. Regarding the HIV risk group, in comparison to MSM, all exposure groups were less likely to be on ART (IDUs 49%, other risk 21% and unknown risk 82%). For virologic suppression, all risk groups were less likely to succeed than MSM (IDUs 57% and remaining groups ranged from 5% to 57%). Individuals diagnosed more recently were more likely to be on ART and to achieve virologic suppression as compared to the previous year (odds ratio 1.15 and 1.16 respectively). Over time, we observed that the gap in the %On ART and %Suppressed by gender, age group and HIV risk group has decreased (see trend in the coefficient of variation in Figure 1). In Figure 1, we identified that females, PLWH aged <30 years and those with unknown risk or who self-identify as IDU were the population subgroups that experienced the most challenge in engaging on ART. These same sub-groups, except for those with unknown risk, had the most difficulty in achieving viral suppression.

| Sensitivity analyses
No scenario explored in our prevalence estimate sensitivity analysis met the 90% and 95% targets for the proportion diagnosed by 2020 and 2030 respectively ( Table 3). The proportion diagnosed varied between 82% to 86% and 90% to 93% respectively. It should be noted that these forecasted scenarios with higher prevalence estimates demonstrate that when the undiagnosed proportion of the prevalence increases relative to that of the original prevalence estimates, incident HIV infections associated with undiagnosed transmission will cause prevalent HIV cases to rise steadily from 2014 to 2020, and from 2020 to 2030.
When focusing on the operational definition of virologic suppression, we estimated the %Suppressed as the last pVL being <200 copies/ml (i.e. PHAC's definition). In this case, the data showed that we have already reached the 90% target in 2008 and the 95% target will be reached in 2015 for %Suppressed among those on ART. For the ultimate target of 73% and 86% for %Suppressed among all PLWH, we estimated that they will be reached in 2018 and 2022 respectively.  [29][30][31]. However, in both cases, the overall evidence is conflicting [32][33][34]. The gender differences found in different studies may be largely related to the heterogeneity in demographic factors (e.g. geography, income, ethnicity), psychosocial (e.g. substance use, sexual behaviour) and treatment-related factors (e.g. treatment adherence) pertaining to this population. The observed association between history of IDU and lower ART uptake and virologic suppression is also long-standing [35,36]. Individuals with a history of IDU may find themselves in situations where they struggle to optimally manage their health, which can be further complicated by active addictions and socio-structural, individual-level and provider-based barriers to care [35,37]. Thus, to optimize our ability to meet the UN targets, it is critical that further efforts be deployed towards optimizing engagement of specific subpopulations, as identified in our study.
At the end of 2016, the PHAC released the progress towards the UN 90-90-90 target for Canada for the year 2014 [38]. It estimated that in 2014, 80% of PLWH were diagnosed, 76% were on ART (among diagnosed PLWH) and 89% were suppressed (among PLWH on ART) in Canada. These compare unfavourably against BC if we use the same definition for virologic suppression, as we estimated that in 2014, in BC, 83% were diagnosed, 78% on ART and 94% were suppressed. BC's current and projected success is better interpreted with a brief acknowledgement of important contextual information. BC is the only jurisdiction in Canada that provides universally covered, fully subsidized ART as well as laboratory and medical monitoring services. At the core of BC's approach to the expansion of TasP were five guiding principles which included: combating stigma and discrimination, reach and engagement, community participation, aboriginal involvement and simplified consent for testing and engagement into care Our sensitivity analysis examining scenarios bring forward the challenge regarding the long-term feasibility and sustainability of the HIV care and treatment in BC in the context of increasing prevalence. It should be noted that in BC, we have shown that the expansion of ART is highly cost-effective even in a scenario of increased prevalence, largely due to substantial decreases in morbidity and mortality [43]. However, an increase in prevalent cases will translate into an increase in resource utilization in the province. With the ongoing commitment of the Provincial Government and the rapidly available generic formulations of the most utilized regimen combinations, we are confident that PLWH in BC will continue to receive appropriate care.
In light of the global consensus on the UN targets [10], the findings of this study should help inform the continued scaleup of TasP strategies in other contexts [44][45][46][47][48][49][50]. In the United States, 86% of PLWH were diagnosed, but only 42% of them were on ART, and 82% of those who were on ART were virologically suppressed in 2011 [44.] The majority (70%) of the countries in Europe that reported data on the percentage of diagnosed PLWH on ART were under 70%. Conversely, nearly two-thirds (65%) of the countries that reported data for the percentage of diagnosed PLWH were between 70% to 89% [45]. Only Sweden reported having met the UN 90-90-90 target [49]. The need for expanded testing is particularly evident in Sub-Saharan Africa, as well as much of Asia and the Pacific region. In sub-Saharan Africa, only 45% of PLWH were diagnosed, whereas 86% of the diagnosed were on ART, and 76% of those on ART were virologically suppressed in 2013 [9]. When comparing the performance between settings, we should note that we cannot establish with absolute certainty that the differences observed reflect a difference in testing and treatment coverage, quality of and access to care, regimen adherence and treatment options as opposed to a variation in the definitions or data sources and quality used to obtain estimates. However, individual, community and systemic barriers continue to negatively affect engagement of PLWH along each stage of the cascade. For example research has identified financial constraints as a significant barrier to ART adherence in resource-rich settings such as in the United States, where co-payments are higher [52]. Stigma also remains a crucial barrier, both perceived and actual stigma have been linked to reduced levels of HIV testing, retention in care and ART adherence in a variety of settings [53,54]. In resource-limited settings, capacity for viral load testing persists as a barrier with regards to monitoring progress. Many countries are not in a position to provide all PLWH on ART with viral load testing. Nonetheless, to meet and appropriately monitor the UN targets, all PLWH who initiate therapy will require access to viral load testing [9].
There are some limitations in our study. First, the actual prevalence of HIV cases from 2000 to 2013 is not definitively known and was thus derived from PHAC's modelling estimates which also provided an independent source for the number of PLWH who were unaware of their infection [20]. However, our sensitivity analysis examined different prevalence scenarios to account for the uncertainty associated with our prevalence estimates. Regarding demographic variables, we were unable to account for the role of ethnicity in achieving each step of the targets due to minimal data availability. Furthermore, while the definition for %Diagnosed used in this paper has been validated empirically [11], our definition differs from the one ascribed by the BC Centre for Disease Control in order to comprehensively capture all HIV cases in the province. Note that HIV cases identified through BC Centre for Disease Control accounts for 91% of the diagnosed individuals included in this study. Lastly, as in any model projection, caution is warranted regarding our projected numbers, especially for 2030, since they were based on historical trends. As such, important changes in policies, social and economic factors may influence these projections.

| CONCLUSIONS
In summary, BC (Canada) is in a position to reach the UN targets by 2020 and 2030. However, there are potential challenges in achieving these targets which should be further addressed in order to further reduce HIV morbidity, mortality and HIV transmission, thereby creating favourable conditions to meet the goal of "ending the AIDS pandemic by 2030. " Thus, despite of these challenges, our results provide strong evidence that integrated comprehensive free programs that facilitate testing and the delivery of treatment and care to PLWH can be effective in controlling, and eventually ending the AIDS pandemic.

C O M P E T I N G I N T E R E S T S
We have the following conflicts of interest: Montaner has received limited unrestricted funding, paid to his institution, from Abbvie, Bristol-Myers Squibb, Gilead Sciences, Janssen, Merck, and ViiV Healthcare. The remaining authors do not have conflicts to declare.

A U T H O R S ' C O N T R I B U T I O N S
Lima and Montaner involved in study concept and design. Lima, Sereda and Montaner involved in acquisition of data. Lima and St-Jean performed drafting of the manuscript. Lima, St-Jean, Shoveller, Nosyk, Hogg, Montaner, Rozada, Sereda and Barrios contributed to critical revision of the manuscript for important intellectual content and for final approval. Lima also performed analysis and interpretation of data, statistical analysis, obtained funding and administrative, technical or material support of the manuscript. Montaner involved in study supervision .

A C K N O W L E D G E M E N T S
We thank the patients enrolled in the BC Centre for Excellence in HIV/AIDS and the physicians, nurses, social workers and volunteers who support them.

F U N D I N G
Montaner is supported with grants paid to his institution by the British Columbia Ministry of Health and by the US National Institutes of Health (R01DA036307). Lima is funded by two grants from the Canadian Institutes of Health Research (MOP-125948 and PJT-148595), by a Scholar Award from the Michael Smith Foundation for Health Research and a New Investigator award from the Canadian Institutes of Health Research.

D I S C L A I M E R
The sponsors had no role in the design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all data in the study and had final responsibility to submit for publication.

SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article: Data S1. Description of the databases in the STOP HIV/AIDS cohort.  Figure S1. Methodology for adjusted prevalence estimates.