Association between opioid agonist therapy use and HIV testing uptake among people who have recently injected drugs: a systematic review and meta‐analysis

Abstract Background and aim Globally, nearly one in five people who inject drugs (PWID) are living with HIV, and the rate of new HIV infections in PWID is increasing in some settings. Early diagnosis is crucial for effective HIV control. We reviewed the evidence on the association between opioid agonist therapy (OAT) and HIV testing uptake among PWID. Methods We conducted a systematic review searching MEDLINE, Scopus, Web of Science, Cochrane Central Register of Controlled Trials and PsycINFO for studies published from January 2000 to March 2019. Reference lists and conference proceedings were hand‐searched. Observational and intervention studies were eligible for inclusion. Risk of bias was assessed using the Risk of Bias in Non‐Randomised Studies of Interventions (ROBINS‐I) tool. Meta‐analyses were conducted using random‐effects models. Results Of 13 373 records identified, 11 studies from Australia, Europe, Malaysia and the United States were included. All studies had at least a serious risk of bias, largely due to confounding and selection bias, making it difficult to draw causal conclusions from the evidence. Ten studies provided data on the association between current OAT use and recent HIV testing. Six showed a positive association, while four provided little evidence of an association: pooled odds ratio (OR) = 1.71, 95% confidence interval (CI) = 1.28–2.27. Looking at having ever been on OAT and having ever been HIV tested, seven studies showed a positive association and three showed either weak or no evidence of an association: pooled OR = 3.82, 95% CI = 2.96–4.95. Conclusions Opioid agonist therapy may increase uptake of HIV testing among people who inject drugs, providing further evidence that opioid agonist therapy improves the HIV treatment care cascade.


INTRODUCTION
Globally, there are an estimated 15.6 million people aged 15-64 years who inject drugs (PWID) [1]. Blood-borne virus infections from injecting drug use are a major contributor to the global burden of disease [2,3]. There are an estimated 2.8 million PWID living with HIV [1], representing 18% of the global population of PWID and contributing an estimated 4% of disability-adjusted life years (DALYs) due to HIV [2].
In contrast to an overall 25% decline in global HIV incidence between 2010 and 2017, the incidence among PWID is increasing in some regions [4]. To reverse this trend, there is a need to improve engagement in every aspect of the HIV cascade of care [4]. Early diagnosis, timely linkage to HIV care and initiation of anti-retroviral therapy (ART) are important, both for the health of the individual and in reducing the risk of onward transmission of HIV [5][6][7][8].
HIV prevention among PWID requires individual, structural and combination approaches [9]. Harm reduction programmes are endorsed by the World Health Organization (WHO) as a key strategy in reducing morbidity and mortality in PWID [4]. Global coverage of harm reduction programmes for PWID improved between 2010 and 2017 [10,11]. However, in 2017 only approximately half of the 179 countries with evidence of injecting drug use had any form of harm reduction programme [10]. Furthermore, only 34 countries report having targeted programmes for HIV testing among PWID, and the number of PWID receiving HIV tests or accessing ART varies widely [10].
When integrated with HIV services, harm reduction in the form of opioid agonist therapy (OAT) is associated with improved initiation and adherence to ART and viral suppression in PWID living with HIV [12,13]. OAT is also associated with a reduction in risk of HIV [14] and hepatitis C virus (HCV) acquisition [15]. However, to date, evidence on the association between OAT use and HIV testing uptake has not been synthesized, nor has the strength of the evidence been critically assessed. We conduct the first systematic review and meta-analysis of the association between OAT use and HIV testing uptake among people who have recently injected drugs. The aims of our review are to assess: (1) the association between current or recent OAT use and HIV testing in the last year; (2) the association between ever having used OAT and ever having been HIV tested; and (3) the risk of bias of the primary studies included in the review.

METHODS
We registered our review protocol with the International Prospective Register of Systematic Reviews (PROSPERO; CRD42019131095). Our report of the review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [16].

Eligibility criteria
Our review included studies that met the following criteria: • Population: adults (> 15 years) with a current or recent history of injection drug use (within the last year). We excluded studies in prison populations, as OAT provision in this setting is likely to be significantly different from treatment delivered in other settings. • Intervention: current or recent (within the last year) OAT use was the primary intervention of interest. OAT use ever was the secondary intervention of interest. We included studies on OAT (both methadone and buprenorphine), irrespective of whether they were delivered in isolation or in conjunction with other harm reduction interventions. • Comparator: no current OAT use was the primary intervention of interest; no history of OAT use was the secondary intervention. • Outcome: recent (within the last year) HIV antibody testing was the primary outcome of interest. Ever HIV antibody testing was the secondary outcome. • Study design: randomized and non-randomized trials and observational studies with a sample size > 40.

Search strategy
We We hand-searched the reference lists of papers selected for inclusion in our review and relevant review papers. Forward citation searches were conducted using Web of Science, Google Scholar and Scopus. We also sought data from known repositories of data on PWID and contacted authors for supplementary data where required.

Study selection
After the removal of duplicates, titles and abstracts were screened by one reviewer, with 10% checked by a second reviewer. Full texts were retrieved for records identified as potentially eligible and were independently reviewed by two authors using pre-defined eligibility criteria. Discrepancies were resolved through discussion, with a third reviewer consulted as necessary.

Data extraction and synthesis
Data were independently extracted by two reviewers into a pre-defined and piloted Microsoft Excel database. Any discrepancies were identified and resolved. Data extracted included study design, study period, country, method of recruitment, inclusion criteria, sample size, age, gender, OAT type and HIV testing (either self-reported or documented in medical or other records). Unadjusted and adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were extracted or calculated for each study. Where data were insufficient or missing, we contacted the study authors. Where there was more than one publication from the same study, the one with the most comprehensive and relevant data to our primary outcome was selected as the main information source. Where a manuscript did not present the specific data that we required, but indicated that this may have been collected, this was requested from the study authors.
Statistical heterogeneity was assessed using the Cochran χ 2 test (Q-test), with the I 2 statistic used to assess the percentage of variability between studies due to heterogeneity rather than sampling error. Due to the presence of between-study variability, summary outcomes were calculated using random-effects meta-analysis. Prediction intervals were calculated to reflect heterogeneity by indicating the interval within which the true effects of 95% of any similar future studies would be expected to fall [17,18]. Logit-transformed outcome estimates were used in all meta-analyses and the estimates were back-transformed for reporting. Where available, ORs adjusted for all key confounders throughout the whole study data set were used in the meta-analyses, otherwise unadjusted estimates were used. A fixed continuity correction of 0.5 was applied where the outcome of a study was 0 or 100%. Meta-regression was planned to assess factors contributing to heterogeneity across studies. All analyses were conducted using the metan package in Stata version 15 [19].

Assessment of risk of bias
We assessed the risk of bias using the ROBINS-I tool ('Risk of Bias in Non-randomised Studiesof Interventions'). The tool assesses bias across seven domains: confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of the outcome and selection of the reported result. Assessments are made at the outcome rather than study level, so for studies that provided data on both the primary and secondary outcome a risk of bias assessment were conducted for each. Studies were judged as having 'low', 'moderate', 'serious' or 'critical' risk of bias for each domain, and the overall risk of bias was then derived. A study was judged as being at low risk of bias overall if it was at low risk of bias for all domains, at moderate risk of bias if it was at a low or moderate risk of bias for all domains, at serious risk of bias if it was at serious risk of bias for at least one domain but not at critical risk of bias in any domain and at critical risk of bias if it was judged to be at critical risk of bias in at least one domain [20].
Risk of bias assessments were undertaken by two reviewers independently with any discrepancies resolved through discussion, and the involvement of a third reviewer as necessary. A priori key confounders were age, gender, time since last injection drug use and markers of severity of addiction (e.g. homelessness, stimulant use in conjunction with injected opioids.)
All studies were cross-sectional, except for one randomized controlled trial. The number of participants in each study ranged from 80 to 19 481. A total of 53 012 individuals were included in our analysis, the majority (73%) of whom are male. The mean age of participants within studies ranged from 23.4 to 39.4 years ( Table 1).
Studies were based in nine countries: Australia, Croatia, Malaysia, Ukraine, the four countries in the United Kingdom and the United States. Individuals from the United Kingdom (52.7%, n = 27 925) and Ukraine (38.9%, n = 20 644) comprised the majority of the sample.

Risk of bias
All studies were assessed as being at serious risk of bias, primarily due to the risk of confounding and selection bias. In the absence of any strong evidence indicating which, if any, factors are likely to confound the association between OAT use and HIV testing we judged studies presenting unadjusted analyses only to be at serious rather than critical risk of bias due to confounding. Two studies provided adjusted estimates for both the primary and secondary analyses and these were assessed as being at moderate risk of bias due to confounding, although at serious risk of bias overall [26,28]. One study provided adjusted estimates for a proportion of their data [32]. As this study remained at a serious risk of bias and provided similar estimates to the unadjusted data (more information is presented in Appendix S2), we chose to use the larger, unadjusted data to increase the power in our analyses. A further study provided adjusted analyses for the secondary analysis but remained at serious risk of bias due to confounding, as only age and sex were adjusted for [21]. Consequently, it was not appropriate to conduct sensitivity analyses to assess how excluding studies at serious risk of bias influenced the findings. Where there was sufficient information to make an assessment, risk of bias for all other domains was low or moderate (Figs 2 and 3).

Impact of OAT on HIV testing uptake
Ten studies provided data for our primary analysis: the impact of current OAT use on HIV testing uptake in the previous year ( Table 2). Six showed a positive association, while four provided little evidence of an association; ORs ranged from 1.03 to 3.58 across studies. Two adjusted estimates were available for the complete study population and were included in the meta-analysis [26,28] (see Appendix S2 for comparisons of unadjusted and adjusted effect estimates). The pooled OR across all 10 studies was 1.71 (95% CI = 1.28-2.27, prediction interval = 0.66-4.43). Inconsistency of results across studies was high (I 2 = 93.9, P < 0.001). No meta-regression was undertaken due to the small number of studies and the lack of clear candidate factors for explaining heterogeneity (Fig. 4).
Ten studies provided data for our secondary analysis: the impact of ever using OAT on uptake of HIV testing ever ( Table 2). Seven studies showed a positive association and three showed either weak or no evidence of an association. Two adjusted estimates were available for inclusion in the meta-analysis (Appendix S2). The pooled OR among all studies was 3.82 (95% CI = 2.96-4.95, prediction interval = 1.69-8.65). Heterogeneity was again high (I 2 = 87.2%, P < 0.001) (Fig. 5).   Multi-site study defined as PWID recruited from more than one geographical location (e.g. city).

DISCUSSION
Our comprehensive review found evidence from published and unpublished observational studies and one randomized controlled trial that current OAT use is associated with an increased uptake of HIV testing in the last year among PWID. We also found evidence to suggest that people who had ever taken OAT were more likely to have ever been tested for HIV. This is the first systematic review, to our knowledge, of the impact of OAT on HIV testing uptake among PWID. Our results are consistent with findings of other studies demonstrating the benefits of OAT use in reducing harms across multiple health outcomes for people who are opioid-dependent [36]. OAT use has been shown to improve engagement in HIV treatment uptake, adherence and viral suppression, as well as engagement in hepatitis C testing and treatment [12,37,38]. OAT use is also associated with a reduced risk of HIVand HCV infections [14,15].
Reasons for the observed positive association between OAT use and HIV testing uptake are likely to be multifactorial. That some OAT programmes require entrants to be HIV tested may be one explanation [39]. The characteristics and preferences of those seeking OAT may also play a role; entry into an OAT programme is indicative of an individual's level of engagement with health-care providers with regard to drug use and associated health-care needs. Furthermore, people in OAT will probably also have increased opportunities to be offered HIV testing, although global data on HIV testing programmes for PWID are   scarce [10]. Finally, the multiple benefits of OAT use, such as improved functional outcomes [40] and reductions in injecting risk behaviour [39], criminal activity [41], overdose and all-cause mortality [42], may lead to better engagement in services and a greater readiness to test, and be treated, for HIV.
The major strength of our review is that we included several sources of unpublished data and completed secondary analyses of data that have not previously been applied to this research question. We recognize that publication bias may be an issue-although we screened more than 13 000 study reports for eligibility, there may be relevant data sets that were not identified through our searches because authors had not published data pertaining to OAT use and HIV testing. There is also a small risk that we missed some studies due to lack of complete duplicate screening at title and abstract screening. However, of the 10% of studies screened in duplicate on title and abstract there were no important discrepancies. All studies that reached the full text screening stage were screened in duplicate.
The findings presented in this review are limited by the relatively small number of studies eligible for inclusion. The majority of data in the review originated from either the United Kingdom or Ukraine, which may limit the generalizability of the findings. All studies were conducted in urban areas, and excluded prison populations where drug injecting and HIV testing practices may differ from other settings [43,44].
Furthermore, the serious risk of bias identified in all the studies included in this review means that findings should be interpreted with caution, and that estimates of association should not be interpreted as convincing evidence of causality. Lack of adjustment for confounding factors, including age, gender, time since last injection drug use, markers of severity of addiction and bias in the selection of participants, were the main reasons that studies were assessed as being at serious risk of bias. However, the three studies providing effect estimates for the primary analysis that were adjusted for all important confounders were similar to the unadjusted estimates, which may suggest that these factors do not strongly influence the association between OAT use and HIV testing.
The temporality of the association between exposure and outcome is also an issue, as the majority (n = 10) of studies included in our review were cross-sectional. We cannot, therefore, assume that OAT use commenced before, rather than after, HIV testing in those studies. To try to minimize this issue we chose current OAT use and HIV testing in the last year as our primary analysis. In future studies it would be interesting and informative to collect data on both the date that OAT started and the date of HIV testing to more clearly understand the temporality of the association.
It was not possible to explore reasons for heterogeneity across the studies in our analysis. It is likely that the characteristics of the study participants and the different settings in which studies were conducted can explain much of the heterogeneity observed. For example, there are variations in both HIV prevalence and OAT coverage by setting. Estimated HIV prevalence among PWID in regions where our included studies were conducted vary between 1.1% (95% CI = 0.8-1.4) in Australasia and 24.7% (95% CI = 15.6-33.9) in eastern Europe [1], while OAT coverage varies from one to two OAT clients per 100 PWID in eastern Europe to 46-95 in western Europe [10].
It was not possible to explore differences in the association between OAT use and HIV testing by gender, due to a lack of disaggregated data. Evidence from some countries suggests that HIV is more prevalent in women who inject drugs than in men [45], and females may face social, structural and psychological barriers to accessing harm reduction services which their male counterparts do not experience [46]. Further studies to explore and address gender differences in the access and outcomes for harm reduction services for PWID would be a welcome addition to the evidence base.
Finally, HIV testing uptake in people who inject stimulants rather than opioids, and consequentially for whom OAT is not indicated, may differ in HIV testing behaviour. This may be due to the different risk factors associated with stimulant, rather than heroin injection, including the sexual transmission of HIV within the men who have sex with men (MSM) community [47]. However, although some PWID in the studies included in our review may be injecting stimulants rather than opioids, a recent systematic review on OAT use and hepatitis C testing and treatment noted that the proportion of PWID who reported ever using opioids was 95-100%, indicating that this issue is unlikely to have substantially biased our findings [38].
For health-care policymakers and practitioners seeking to reduce the harms associated with injecting drug-use, the findings of this study provide further support for OAT as a critical component of harm reduction. Together with evidence on the benefits of OAT on the HIV treatment care cascade [12] our findings suggest that, in addition to the benefits of OAT across multiple health domains [36], OAT may also improve HIV testing and treatment among PWID. OAT programmes may also support work in settings where HIV testing programmes for PWID [10] are in development or do not comply with the recommendations made by the WHO [48]. However, even where OAT and HIV testing programmes are well established, there is more work to do to understand barriers and enablers to HIV testing uptake. Recent findings from the United Kingdom, for example, suggest that there are still many PWID who are accessing health and harm reduction services and yet report no history of HIV testing [49].

CONCLUSIONS
Our findings indicate that OAT may increase uptake of HIV testing among PWID, providing further evidence of the benefits of OAT use on the HIV treatment and care cascade.