Needle syringe programmes and opioid substitution therapy for prevention HCV transmission among people who inject drugs: Cochrane Systematic Review Lucy Platt, Silvia Minozzi, Peter Vickerman, Clare French, Jennifer Reed, Vivian Hope, Louisa Degendhardt, Lisa Maher, Julie Bruneau, Holly Hagan, Ashly Jordan, Matthew Hickman Many thanks for opportunity to talk about our systematic review – acknowledge my co-authors – especially Lucy

Acknowledgements & Disclosure/CoI NIHR (HS&DR) (12/3070/13) - Assessing the impact and cost-effectiveness of NSP NIHR Health Protection Research Unit in Evaluation of Interventions European Commission Drug Prevention and Information Programme (DIPP) “Treatment as Prevention in Europe: Model Projections [JUST/2013/DPIP/AG/4812] MH received honoraria from Abbvie, MSD, Janssen, Gilead. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. Acknowledge also funders NIHR and EU – and disclosures

NSP review level evidence:- Background NSP review level evidence:- sufficient that reduce injecting related risk behaviour insufficient that prevent HCV transmission; tentative that prevent HIV transmission; OST associated with reduce risk of HIV transmission Evidence to support effectiveness of NSP to reduce HIV transmission (0.66 95%CI 0.4-1.01) Background well known. Until very recently review level evidence suggested that good evidence that OST and NSP reducing reported sharing equipment but insufficient that reduced transmission – and tentative to good evidence that OST and NSP reduce HIV transmission.

Methods Search Inclusion Risk of bias: www.riskofbias.info/ Analyses Medline, PsycInfo, Embase, Cochrane and Web of Science up to July 2014 ; Reference lists; Grey literature Inclusion observational and intervention studies measuring exposure to NSP and/or OST (compared to no intervention) among PWID and HCV incidence Risk of bias: www.riskofbias.info/ Studies classified as Low, Moderate, Serious, Critical risk according to 7 domains Critical confounders: duration of injection or age; injecting frequency Analyses Meta-analysis using random effects models & heterogeneity explored Separate analyses for adjusted and unadjusted, exclusion of studies classified as ‘critical’ and subgroup analyses, geographical region, recruitment site, type of drug injected, frequency of injecting So we conducted a global systematic review for Cochrane to quantify evidence using random effect meta-analyses of the impact of OST and NSP on HCV incidence among PWID We searched multiple databases – no language restrictions. We did two separate systematic search strategies – one for studies that directly measured OST or NSP intervention effect- the other for cohort studies that reported HCV incidence among PWID to identify whether they reported effect of OST or NSP on HCV and if not to contact the researchers for unpublished estimates if they exist. We classified risk of bias using novel tool developed by Cochrane – which compares each study against what a perfect study would be like – all fell short

Flow Chart – Included studies 6814 records identified through database searching with IMPACT search (MEDLINE: 932; PsycINFO: 451; Global Health: 575; EMBASE: 2298; CINAHL: 123; WOS: 1679; CENTRAL: 630; DARE: 13; NHSEED: 20; HTA: 3; CLIB: 21; CDAG register:69) 6720 records after duplicated removed Records screened (n =6720) Records excluded (n =6576) Full-text articles assessed for eligibility (n =144) Full-text articles excluded, with reasons (n = 103 referring to 101 studies) Studies awaiting classification (n=10) Studies included in qualitative synthesis (n = 28, (31 articles)) Studies included in quantitative synthesis (meta analysis) (n= 28) 6996 records identified through database searching with LONGITUDINAL search (MEDLINE: 1926; PsycINFO: 284; Global Health: 248; EMBASE: 2276; CINAHL: 74; WOS: 1808) Flow Chart – Included studies This is our flow diagram we screened over 6700 records Looked at 144 full papers and identified 28 studies

Description of studies - I 21 published, 7 unpublished 13 North America, 5 UK, 4 other Europe, 5 Australia and 1 China 1817 HCV infections 8806.95py follow-up. HCV incidence - 0.09 to 42 cases per 100 py Risk of Bias Only 2 at moderate overall risk of bias, 17 at serious risk, 7 at critical risk; 2 unpublished studies no RoB No RCT intervention effects – so typically evidence quality categorised as low. The 28 studies included 21 published and 7 unpublished estimates …. We didn’t identify any RCT – so this means quality of evidence tended to be categorized as low… and most of the observational studies were at serious or critical risk of bias – because of problems with adjustment of confounders, study design, measurement of exposure, and selection of participants…

Description of studies - II Impact of OST (17), NSP (15), combination NSP/MMT (4) Interventions current OST (within last 3-6 months) High NSP coverage (regular NSP attendance/ all injections covered by a new needle/syringe) Variation across studies: Sample size (range 46-2788) Method of recruitment: street outreach; RDS; snowball sampling; service users Study design: Prospective cohort (20); Cross-sectional (3); case-control studies (2); 2 retrospective cohort, 1 serial cross-sectional Published 1995-2014 Follow-up time 1 and 22 years So in terms of intervention we measured current/recent OST And a variety of measured of high frequency/coverage of NSP – ideally in terms of whether picked up sufficient number of sterile equipment for injecting frequency – but if not in terms of how regular attended NSP. There was variation in the studies in terms of sample size, follow-up, recruitment and study design. Most were cohort studies but we did also have some cross-sectional studies that identify recent sero-conversions as the outcome.

Impact of current OST exposure (adjusted estimates) 12 studies: 6361 participants 1030 HCV cases 50% reduction in risk of HCV Little heterogeneity GRADE: Low Evidence. Next few slides are our results. Here you see forest plot for opioid substitution treatment – and we show studies with adjusted estimates only in this slide. This includes twelve observational studies – including over 6300 participants and 1000 HCV incident events. Consistent evidence across geographical areas (Australia, North America, and Europe) – with little heterogeneity that OST reduces HCV transmission by 50%.   Because no trials (and there wont be any as would be unethical now) we have to grade the evidence as low quality – but strong good low quality evidence…

Impact of current OST Effect maintained if exclude unpublished estimates, studies at critical or no information on risk of bias, cross-sectional studies Effect strengthened if exclude unpublished estimates and weakened slightly if pool unadjusted studies (10,647 participants, RR =0.57, I2=32.4) No evidence that effectiveness varied by geographic region or site of recruitment. Intervention effect reduced with 10% increase in sample of PWID who were women Now we did do some sensitivity analyses – and found that the intervention effect was maintained if exclude unpublished studies or others. And stronger if look at published studies only, and slightly weaker if pool across all unadjusted studies – but still strong effect. Importantly no evidence of differential effect by region or site of recruitment or type of study – but a slight diminution of effect with every 10% increase in proportion of women in the sample.

No evidence of publication bias We did not detect any publication bias – any over-representation of small studies with large effects Funnel plot – OST meta-analysis - Egger’s bias coefficient (−0.87 p=0.106)

Impact of high NSP by region (unadjusted analyses) 7 studies High heterogeneity (I2=79%) Weak evidence overall – RR 0.77 In Europe NSP associated with 66% reduction in HCV Grade: very low evidence Now to Needle and syringe programmes – evidence is more mixed. Here we show Forest plot for all studies – using unadjusted estimates because didn’t want to drop too many studies. Overall we find weak evidence – across seven studies – that high coverage NSP High coverage NSP defined either as regular attendance or people receiving sufficient syringes for their injecting frequency reduces HCV transmission by 23%. There is a great deal of heterogeneity. For studies in Europe which had a more consistent measure of NSP exposure there is evidence that can reduce HCV transmission by 50% or more.

Impact of NSP and OST High NSP with OST 4 studies 3356 participants 518 HCV cases Reduced HCV by 71% moderate heterogeneity Low NSP with OST 3 studies 3071 participants 449 HCV cases, Reduced HCV by 24% GRADE: low evidence Four studies reported combined exposure to both NSPs and OST – the combination of high coverage NSP and OST reduced HCV by 71% - shown in top forest plot 3 studies also could assess combination of low NSP and OST – and suggested that impact of OST on HCV transmission was reduced in absence of high coverage NSP

Impact of current NSP Meta-regression No evidence differential impact by region N. America vs Europe (Ratio of rate ratios= 3.73, p=0.06). No evidence publication bias differential impact by proportion of female participants in the sample, homelessness or experience of prison. Unsurprisingly we found strong evidence in meta-regression that there was differential impact of NSP by region. But we also found no evidence of publication bias – admittedly across small number of studies – and no differential impact with other factors – such as % of women in the sample, or homelessness or prison exposure

Impact of current NSP: Key limitations & potential explanations heterogeneity in measurement of NSP exposure In North America – studies had varied definitions of NSP use relating to frequency of attendance at NSPs. In Europe – studies tend to measure 100% NSP coverage “person reports that they receive sufficient or a greater number of sterile syringes per injecting frequency” confounding/ selection bias lower overall coverage in US may mask intervention effect higher proportion of stimulant injectors in US studies that without OST may contribute to lower impact regular NSP attenders at greater risk of HCV acquisition US studies conducted when incidence was higher So key limitations and potential explanations for why we have no evidence from N America are related The meta-regression analysis also showed evidence of differential impact by region comparing North America with Europe (Ratio of rate ratios= 3.73, p=0.06) (Table 3). Univariable meta regression analysis also suggested some association between high coverage of NSP and study design (Ratio of rate ratios=3.5, p value=0.087 comparing cross-sectional with longitudinal study design), this was reduced when adjusted for geographical region (Ratio of rate ratios=1.7, p value=0.577). We found no evidence of differential impact by proportion of female participants in the sample, homelessness or experience of prison. The two key limitations of the review are heterogeneity in measurement of exposure to NSPs and confounding. Consistent measures of NSP exposure through coverage of injections by clean needles/syringes were used across the European studies12, 29, 32, 61, whereas the North American studies drew on varied definitions of NSP use that focussed on frequency of attendance at NSPs.27, 33, 55 The measure of 100% NSP coverage corresponds to the situation where a person reports that they receive sufficient or a greater number of sterile syringes per reported injecting frequency and is found in sites which allow PWID to collect a large number of syringes or attend very regularly rather than in sites which adopt a more restricted form of exchange. It is possible also that the population exposed to 100% coverage are more compliant than the comparison group in terms of regular attendance and uptake of needles/syringes. However, we cannot assume that they necessarily use all needles/syringes obtained since HCV transmission still occurs in this population. Inconsistencies in NSP measurement contributed to heterogeneity observed among studies (I2=77%, p=0.002), while differences in study design, exposure measurement, and patterns of injecting may have contributed to the lack of effect of NSPs on HCV transmission observed in North America. The European definition of coverage may include needles/syringes obtained from pharmacies, secondary distribution via friends or via outreach whereas the North American definition is specific to fixed site NSP use. In reality PWID may obtain needles/syringes from multiple sources and, as a consequence, use NSPs less frequently for social support, HIV/HCV testing and counselling and other specialist advice. These additional services are essential to prevent the spread of blood-borne viruses and reduce inequalities in health over time63, but the immediate effect on HCV incidence may be difficult to assess where measurement of exposure doesn’t capture uptake of sterile needles/syringes from other sources. Measurement of NSP use that focuses on needles/syringes without taking into account acquisition of other drug preparation equipment from NSPs may also have contributed to the lack of an association in some settings, particularly in the US where nearly half of HCV seroconversions have been attributed to the shared use of cookers (spoons) and cottons.5 It is also hypothesised that less frequent use of NSPs and lack of federal funding for NSPs in the United States has resulted in lower coverage among PWID overall which may mask an intervention effect, although in cities where the studies took place coverage is not as low as for the US overall.64 65 The higher proportion of stimulant injectors in US studies also may contribute to lower impact. Meta regression analysis suggested no differential impact between intervention effect and study design for current use of OST, but did suggest that longitudinal studies of NSP found a lower effect than cross-sectional studies which were associated with European settings. The control of confounders was limited and inconsistent across the studies. However, synthesised effect estimates for OST and combined effect of OST and high coverage NSP were consistent across multiple studies and maintained between analyses that adjusted for confounders and those that did not, suggesting the variation did not affect the results. We cannot rule out the effect of residual confounding on NSP – especially for the lack of association between NSP use and HCV from studies in North America. For example, it has been shown that people who attend NSPs regularly in North America also report greater injecting risk behaviours and other social vulnerabilities (including sex work or homelessness) and that after adjustment for these factors any positive association between HCV or HIV transmission and NSP attendance is reduced.66 67

strong consistent evidence that OST reduces HCV transmission Implications strong consistent evidence that OST reduces HCV transmission weaker evidence for high coverage NSP more heterogeneity NSP highly cost effective/ cost saving See Zoe Ward talk corroborates importance of combining interventions (NSP and OST) Model evidence that OST/NSP enhance HCV TasP & minimize re-infection

Studies with high level of bias Implications Lack of published data from outside Europe, (Australia) and North America Studies with high level of bias No RCT possible Global coverage estimates confirm that NSP/OST scale-up needed in most countries Improve reporting of observational studies Better natural experiments Strengthen evidence base as improve coverage

Joint publication We had hoped studies would be available you today – but Cochrane computer said no – so will be released by Cochrane and Addiction on 11th Sept… hopefully Cochrane Database of Systematic Reviews DOI: 10.1002/14651858.CD012021