What works: A synthesis of research on HCV prevention for drug injectors Holly Hagan Don C. Des Jarlais Enrique R. Pouget Corina Lelutiu-Weinberger Center for Drug Use and HIV Research NDRI New York, NY
Background What works for HIV prevention does not necessarily control HCV Host, agent and environmental factors favor endemic HCV in IDUs Risk reduction vs. risk elimination Large reservoir of infectious individuals Injection as a complicated process in a chaotic setting
Outline of this talk Summarize Meta-analysis of association between HCV seroconversion and prevention Meta-regression of interval from onset of injection to HCV infection Discuss Implications for research and prevention
Methods - HCV Synthesis Project Design and scope Meta analysis and synthesis of research on HCV epidemiology and prevention in drug users throughout the world Data collection Published and unpublished data sought Encompassed studies published since the discovery of HCV Sample January 1989 to December 2006 2,375 reports identified and screened 628 eligible reports
Meta-analysis of association between HCV seroconversion and prevention Included all HCV incidence studies reporting on the association with prevention measures: Drug treatment Needle exchange Used adjusted OR/RR/HR when available All observational epidemiology studies Evaluated heterogeneity & calculated summary estimates of the effect ect on HCV seroconversion Comprehensive Meta-Analysis software Examined operational definition of prevention Dose or type?
Drug Treatment 1 st Author Study Dates Location N Definition of Treatment Brunton 1996 New Zealand (multi-site) 39 Hagan 1994-2001 Seattle 484 Lamothe 1992 Montreal 63 Maher 1999-2002 Sydney 368 Patrick 1996-1999 Vancouver 155 In treatment (not specified) at follow-up interview In treatment e t (not specified) ed) at baseline In treatment (not specified) at follow-up interview In treatment (not specified) at baseline Addiction therapy (not methadone) during follow-up Addiction treatment (not Smythe 1992-19991999 Dublin 98 specified) more than 3m vs. less during follow-up
Drug Treatment t Model First Author Statistics for each study Odds ratio and 95% CI Odds Lower Upper ratio limit limit Z-Value p-value Brunton 11.50 1.93 68.52 2.68 0.01 Hagan '04 090 0.90 064 0.64 127 1.27-0.60 060 055 0.55 Lamothe 0.69 0.25 1.88-0.74 0.46 Maher 0.84 0.48 1.47-0.61 0.54 Patrick 3.58 1.35 9.49 2.57 0.01 Smythe 0.69 0.42 1.15-1.41 0.16 Fixed 0.94 0.74 1.19-0.52 0.61 Random 1.15 0.69 1.92 0.53 0.59 0.1 0.2 0.5 1 2 5 10 Less Risk More Risk Significant Heterogeneity
Opioid Replacement Therapy 1 st Author Study Dates Location N Definition of Opioid Replacement Therapy Crofts 1991-1995 Victoria 73 Dolan 1997-2002 New South Wales prisons Continuous vs. none or interrupted MMT during follow-up o 222 Any vs. no MMT during follow-up Hall 1996-2003 Sydney y 54 Continuous vs. interrupted ORT during follow-up Lucidarme 1999-2001 France (multisite) 131 In substitutive treatment vs. not during 3m before baseline Patrick 1996-1999 Vancouver 155 Methadone therapy last 6m of follow-up Rezza 1991-1993 Naples 106 Any vs. no methadone treatment during last 6m of follow-up Thiede 1994-1998 Seattle 78 Continuous vs. interrupted MMT during follow-up van Beek 1992-19951995 Sydney 144 Ever vs. never receive MMT van den Berg 1985-2005 Amsterdam 903 Any methadone treatment during follow-up
Opioid id Replacement Therapy Model First Author Statistics for each study Odds ratio and 95% CI Odds Lower Upper ratio limit limit Z-Value p-value Crofts 1.05 0.35 3.17 0.09 0.93 Dl Dolan 098 0.98 052 0.52 185 1.85-0.06 006 095 0.95 Hall 0.12 0.01 1.17-1.82 0.07 Lucidarme 0.53 0.17 1.65-1.09 0.27 Patrick 0.47 0.14 1.52-1.27 0.20 Rezza 0.34 0.11 1.13-1.76 0.08 Thiede 0.40 0.04 3.50-0.83 0.41 van Beek 1.39 0.50 3.87 0.62 0.53 van den Berg 114 1.14 063 0.63 204 2.04 043 0.43 067 0.67 Fixed 0.84 0.61 1.15-1.08 0.28 Random 0.80 0.56 1.15-1.18 0.24 0.1 0.2 0.5 1 2 5 10 Less Risk More Risk Non-significant Heterogeneity
Opioid id Replacement Therapy (Continuous Treatment) t) Model First Author Statistics for each study Odds ratio and 95% CI Odds Lower Upper ratio limit limit Z-Value p-value Crofts 1.05 0.35 3.17 0.09 0.93 Hall 0.12 0.01 1.17-1.82 0.07 Thiede 0.40 0.04 3.50-0.83 0.41 Fixed 0.63 0.26 1.56-1.00 0.32 Random 051 0.51 015 0.15 177 1.77-1.06 106 029 0.29 0.1 0.2 0.5 1 2 5 10 Less Risk More Risk Non-significant Heterogeneity
NEP Participation 1 st Author Study Dates Location N Definition of NEP participation Hagan 95 1991-19931993 Tacoma 66 Ever vs. never Hagan 04 1994-2001 Seattle 484 Lamothe 1992 Montreal 63 Patrick 1996-1999 Vancouver 155 Roy 1997-2003 Montreal 359 Thorpe 1997-1999 1999 Chicago 353 Ever vs. never during follow- up; also frequency, recency Any vs. none last 6m followup Frequent attendance vs. less last 6m of follow-up Any vs. none last 6m followup Any vs. none last 6m follow- up
NEP Participation i Model First Author Statistics for each study Odds ratio and 95% CI Odds Lower Upper ratio limit limit Z-Value p-value Hagan '95 0.14 0.03 0.62-2.59 0.01 Hagan '04 140 1.40 096 0.96 203 2.03 177 1.77 008 0.08 Lamothe 2.79 1.14 6.84 2.24 0.03 Patrick 2.56 1.37 4.79 2.94 0.00 Roy 3.02 2.38 3.83 9.15 0.00 Thorpe 1.18 0.53 2.63 0.40 0.69 Fixed 2.27 1.89 2.71 8.90 0.00 Random 1.64 0.95 2.83 1.78 0.08 0.1 0.2 0.5 1 2 5 10 Less Risk More risk Sgnificant Heterogeneity
Control of Confounding? Note Rezza et al 1996: The association between HCV seroconversion and MMT during follow-up OR = 2.8 (1.1, 7.4) AOR* = 0.34 (0.10, 1.11) *Controlled for age, duration injecting, i daily injecting, injecting cocaine, sharing of drug preparation p and injection equipment, sexual partnership characteristics
Meta-regression of interval from onset of injection to HCV infection No standardized categories of time since onset of injection For each study, calculated the midpoint of time- categories Each category represented by a single value Potential ti effect modifiers Data collection 1985-95 vs. later 1995 chosen a priori to represent a division between an early period and the expansion of harm reduction programs Developing/transitional country Linear mixed effects meta-regression models of HCV rates for time at risk categories
Results: HCV prevalence in relation to time at risk 72 studies reporting prevalence in relation to time since onset of injection time at risk 293 categories of time at risk 31% conducted 1985-95 f d l l 19% from developing or transitional countries
Observed and fitted HCV prevalence in relation to time at risk
Mean fitted values: HCV prevalence two years after onset of injection Prevalence 95% CI Developing/transitional countries 57.9% 56.6 59.1% Non-developing/transitional countries 1985-95 52.8% 47.1 58.5% 5% Non-developing/transitional countries post-1995 40.6% 39.5 41.6%
Summary Confounding in individual-level observational studies of the association between HCV infection and harm reduction Volunteer bias in needle exchange Hagan et al, 2001 Frequent needle exchange use and higher risk Wood et al, 2007 Community-level trends and analysis of time to infection strongly suggest that harm reduction is working BUT need more knowledge regarding dose-response relationship contribution of individual components
What works: Encouraging findings are nice, but Knowledge is limited Exactly what aspects of prevention are working is unclear What elements should be increased? What is unnecessary? How much [more] prevention is needed? What level of endemic HCV will we have to live with? Is it conceivable that prevalence will decline over the next ten years?
Acknowledgements The HCV Synthesis Project was funded by the National Institute on Drug Abuse RO1 DA 18609