Article

A peer-education intervention to reduce injection risk behaviors for HIV and hepatitis C virus infection in young injection drug users

George Washington University, Washington, Washington, D.C., United States
AIDS (Impact Factor: 5.55). 10/2007; 21(14):1923-32. DOI: 10.1097/QAD.0b013e32823f9066
Source: PubMed

ABSTRACT

To evaluate whether a behavioral intervention, which taught peer education skills, could reduce injection and sexual risk behaviors associated with primary HIV and hepatitis C virus infection (HCV) among young injection drug users (IDU).
We conducted a randomized controlled trial involving HIV and HCV antibody-negative IDU, aged 15-30 years, recruited in five United States cities. A six-session, small-group, cognitive behavioral, skills-building intervention in which participants were taught peer education skills (n = 431) was compared with a time-equivalent attention control (n = 423). Baseline visits included interviews for sociodemographic, psychosocial, and behavioral factors during the previous 3 months; HIV and HCV antibody testing; and pre/posttest counselling. Procedures were repeated 3 and 6 months postintervention.
The intervention produced a 29% greater decline in overall injection risk 6 months postintervention relative to the control [proportional odds ratio 0.71; 95% confidence limit (CL) 0.52, 0.97], and a 76% decrease compared with baseline. Decreases were also observed for sexual risk behaviors, but they did not differ by trial arm. Overall HCV infection incidence (18.4/100 person-years) did not differ significantly across trial arms (relative risk 1.15; 95% CL 0.72, 1.82). No HIV seroconversions were observed.
Interventions providing information, enhancing risk-reduction skills, and motivating behavior change through peer education training can reduce injection risk behaviors, although risk elimination might be necessary to prevent HCV transmission.

Full-text

Available from: Hanne Thiede
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
A peer-education intervention to reduce injection risk
behaviors for HIV and hepatitis C virus infection in
young injection drug users
Richard S. Garfein
a,g
, Elizabeth T. Golub
b
, Alan E. Greenberg
a,i
,
Holly Hagan
c
, Debra L. Hanson
a
, Sharon M. Hudson
d
,
Farzana Kapadia
e
, Mary H. Latka
e
, Lawrence J. Ouellet
f
,
David W. Purcell
a
, Steffanie A. Strathdee
g
and
Hanne Thiede
h
for the DUIT Study Team
Objectives: To evaluate whether a behavioral intervention, which taught peer edu-
cation skills, could reduce injection and sexual risk behaviors associated with primary
HIV and hepatitis C virus infection (HCV) among young injection drug users (IDU).
Design: We conducted a randomized controlled trial involving HIV and HCV antibody-
negative IDU, aged 1530 years, recruited in five United States cities. A six-session,
small-group, cognitive behavioral, skills-building intervention in which participants were
taught peer education skills (n ¼ 431) was compared with a time-equivalent attention
control (n ¼ 423). Baseline visits included interviews for sociodemographic, psychoso-
cial, and behavioral factors during the previous 3 months; HIV and HCV antibody testing;
and pre/posttest counselling. Procedures were repeated 3 and 6 months postintervention.
Results: The intervention produced a 29% greater decline in overall injection risk
6 months postintervention relative to the control [proportional odds ratio 0.71; 95%
confidence limit (CL) 0.52, 0.97], and a 76% decrease compared with baseline.
Decreases were also observed for sexual risk behaviors, but they did not differ by trial
arm. Overall HCV infection incidence (18.4/100 person-years) did not differ significantly
across trial arms (relative risk 1.15; 95% CL 0.72, 1.82). No HIV seroconversions were
observed.
Conclusion: Interventions providing information, enhancing risk-reduction skills, and
motivating behavior change through peer education training can reduce injection risk
behaviors, although risk elimination might be necessary to prevent HCV transmission.
ß 2007 Lippincott Williams & Wilkins
AIDS 2007, 21:19231932
Keywords: Hepatitis C virus, HIV, injection drug use, peer education,
randomized controlled trial, young adult
From the
a
Division of HIV/AIDS Prevention, National Center for HIV, STD, and TB Prevention, Centers for Disease Control and
Prevention, Atlanta, Georgia, USA, the
b
Johns Hopkins University, Bloomberg School of Public Health, Baltimore, Maryland,
USA, the
c
National Research and Development Institutes, New York, New York, USA, the
d
Health Research Association, Los
Angeles, California, USA, the
e
New York Academy of Medicine, Center for Urban Epidemiologic Studies, New York, New York,
USA, the
f
University of Illinois at Chicago, School of Public Health, Chicago, Illinois, USA, the
g
University of California San Diego,
School of Medicine, San Diego, California, USA, the
h
Public Health Seattle and King County, Seattle, Washington, USA, and the
i
George Washington University, School of Public Health and Health Services, Washington, DC, USA.
Correspondence to Richard S. Garfein, PhD, MPH, Division of International Health and Cross Cultural Medicine, Department of
Family and Preventive Medicine, University of California San Diego, School of Medicine, 9500 Gilman Drive, Mail Code 0622,
San Diego, CA 92093-0622, USA.
Tel: +1 858 822 3018; fax: +1 858 534 4642; e-mail: rgarfein@ucsd.edu
Received: 12 October 2006; revised: 30 March 2007; accepted: 11 April 2007.
ISSN 0269-9370 Q 2007 Lippincott Williams & Wilkins
1923
Page 1
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Introduction
Injection drug use remains the leading risk for HIV
infection in many countries worldwide [1]. The prevalence
of hepatitis C virus (HCV) infection among injection drug
users (IDU) ranges from 30% to more than 90% with an
incidence between 10 and 75/100 person-years [28].
These infections remain an enormous public health
challenge as each involves a chronic carrier state facilitating
ongoing transmission and costly medical consequences,
including AIDS, cirrhosis, and hepatocellular carcinoma.
Lacking vaccines to prevent infection, primary prevention
relies on reducing the risk of exposure. Beyond syringe
sharing, interventions are needed to prevent the sharing of
other drug paraphernalia [9,10], drug preparations [1113]
and risky sexual behaviors among IDU [1417].
Recently initiated IDU have higher HIV and HCV
seroincidence than IDU with longer duration of use
[1821]. A study of 1830-year-old IDU in six United
States cities showed that sexual [22] and injection [10,23]
risk behaviors were highly prevalent among young IDU,
underscoring the need to structure and target interventions
for recently initiated, typically younger IDU. Although
some interventions not specifically focusing on younger
IDU produced small reductions in risky injection behaviors
[2427], they were less effective at reducing sexual
risk behaviors, particularly among steady sex partners
[24,28,29], which may be more difficult to change
[3032].
Few primary HIV prevention interventions for IDU have
been evaluated through randomized controlled trials [33],
and none have directly addressed HCV infection. Although
behavior change interventions based on cognitive
behavioralskills-building[34]and peer-based interventions
[35] have been evaluated, those studies included non-IDU
or IDU who were not recent initiates. Whereas modest
effects were observed in cognitivebehavioral interven-
tions among younger individuals and IDU [3639], the
extent to which an intervention combining cognitive
behavioral theory and peer education can more effectively
reduce risk among young IDU has not been explored. We
evaluated the efficacy of an intervention incorporating
cognitivebehavioral skills-building into a programme
designed to teach IDU how to provide peer education
about sexual and injection risk reduction to decrease their
own HIV and HCV infection risk.
Methods
Participants
Between May 2002 and January 2004, IDU were recruited
through street outreach, advertising, and coupon-based
participant referrals in Baltimore, Chicago, Los Angeles,
New York, and Seattle. Eligible participants had injected
illicit drugs in the past 6 months, resided in the recruitment
city with no plans to move within 12 months, spoke
English, were aged 1530 years, were willing to undergo
HIV and HCV antibody testing, and provided written
informed consent. Trial-eligible participants had to test
HIV and HCV antibody-negative at baseline. Further
details are provided elsewhere [40].
Study design
The study, known to participants as ‘DUIT’, was a
randomized controlled trial (Fig. 1). Pretest counselling
was provided at baseline. Test results and face-to-face
posttest counselling were given approximately 2 weeks
later, before intervention trial enrollment. Counselling was
based on Centers for Disease Control and Prevention
(CDC) guidelines [41] and included comprehensive HIV
and viral hepatitis risk-reduction information. When
indicated, participants were referred for medical, drug
treatment, and social services. Participants testing antibody
negative for HIVand HCVat baseline were enrolled in the
trial upon return for a subsequent visit during which a
cohort of approximately 1030 eligible participants was
randomly assigned to trial arms. Sites varied by the number
of cohorts (range 1223). Study outcomes were assessed at
3 and 6-month postintervention follow-up visits. Partici-
pants were remunerated for time and travel after each visit
per local norms. Institutional review boards at CDC and all
collaborating institutions approved the study protocol.
Trial arm assignment was conducted by one staff member at
each site, who was not an intervention facilitator, using a
computer program provided by the central data manager,
which individually randomly assigned participants
immediately before the intervention began. Randomiz-
ation stratified participants by sex and age (< 18 and 18
years) with a block size of six to achieve similar participant
ratios across arms. Mid-enrollment, group assignment was
amended to stop the loss of participants who became
frustrated by rescheduling random assignments as a result of
low turnout. Thereafter, when the number of participants
arriving for random selection was between five and nine
the whole cohort was randomlyassigned to one trial arm by
drawing group assignments from sealed envelopes.
Intervention methods
Theories guiding the DUIT peer education intervention
(PEI) included social learning theory [42] and the
information, motivation, and behavioral skills model
[43]. As a framework for education and skills-building
activities, the intervention centered on teaching partici-
pants how to educate peers about HIV and HCV risk
reduction. By encouraging participants to adopt this new
prosocial role, we intended to motivate them to change
their own behaviors. The PEI consisted of six 2-h sessions
over a 3-week period. Sessions were conducted by two
facilitators, at least one of whom was female. Session 1
described HIV and HCV transmission through sex and
injection drug use, informed participants about disease
1924 AIDS 2007, Vol 21 No 14
Page 2
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
prevalence in their communities, and described the vital
role peer educators play in preventing further disease
spread. Sessions 2 and 3 provided peer education about safer
injection and sexual practices, respectively, with activities
designed to increase negotiation skills with sex and
injection partners. Session 4 added skills-building activities
and prepared participants to demonstrate peer education in
settings observable to intervention facilitators (e.g. an
information table on the street or near a syringe exchange
programme). During session 5, small teams of up to ve
participants conducted 90-min peer education sessions.
Participants debriefed afterwards with the facilitator, who
reinforced positive experiences and minimized potentially
negative reactions. Session 6 consisted of a large group
debrieng, goal-setting to encourage continued risk
reduction, and a graduation ceremony. Intervention details
and development methods are described elsewhere [44].
Participants randomly assigned to the attentioncontrol
arm received a video discussion intervention (VDI)
comprising equivalent hours and sessions as the PEI.
VDI participants watched hour-long lms addressing social
(e.g. gun violence, gangs, prejudice) and health (e.g.
cardiopulmonary resuscitation training, alcoholism, injury
prevention) issues followed by facilitated discussion using
scripted questions. Risk-reduction topics were diverted by
offering the same education pamphlets given to PEI
participants.
Data collection
At baseline and 3 and 6-month follow-up visits,
participants completed behavioral risk interviews using
audio computer-assisted self-interviewing (ACASI). Inter-
views were administered before pretest counselling and
venipuncture to minimize reporting bias.
Outcome measures
Multiple outcome measures were assessed to determine
whether the intervention had an effect on decreasing
several sexual and injection risk behaviors. Methods
employed to enhance the validity of participants self-
reported behaviors included: (i) using ACASI to minimize
interviewer bias and socially desirable responding [4547];
(ii) excluding intervention facilitators from follow-up
assessment activities to avoid a booster effect; (iii) using
study ID numbers instead of names during data collection;
and (iv) using calendars and a brief recall period (3 months)
to maximize recall. HCVantibody seroconversion during
follow-up provided a biological outcome measure.
Self-reported injection behaviors
Key injection risk indicators included the proportion of
all injections during the previous 3 months that involved:
(i) injecting with a syringe used previously by another
IDU; (ii) using a new sterile syringe to divide drugs with
another IDU when drugs were split; (iii) sharing cookers;
(iv) sharing cotton lters; and (v) sharing rinse water.
Reducing injection risk behaviors in young injection drug users Garfein et al. 1925
Fig. 1. Flow diagram of the 3rd Collaborative Drug Users Study/Drug Users Intervention Trial study visits and participation
rates, 20022004. HCV, Hepatitis C virus.
Included in analysis.
Page 3
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Questions regarding these risk indicators included seven-
item response categories (1, never;2,rarely;3,less
than half the time;4,about half the time;5,more than
half the time;6,almost always;7,always). The drug-
splitting variable had reverse-coded response categories
making higher values equate to greater risk. We also
assessed the number of injection partners in the past
3 months, and the proportion with whom participants
shared injection equipment with on a seven-item scale
(1, none;2,almost none;3,less than half ;4,about
half ;5,more than half ;6,almost all;7,all). A
composite injection risk variable was created by summing
the six proportion variables and dividing by six to produce
a single outcome measure with values ranging from one
to seven (Cronbachs alpha 0.83).
Self-reported sexual behaviors
Participants were asked for the number of vaginal and anal
sex acts, with and without condoms, stratied by partner
type (main steady, other steady, and casual or sex trade
partners). By denition, participants could have only one
main steady sex partner, but multiple other steady sex
partners. Six sexual risk outcomes were computed
enumerating unprotected vaginal or anal intercourse acts
with a main partner, other steady partners, or casual/sex
trade partners. Summing these variables provided the total
number of unprotected sex acts with all partners during
the past 3 months.
Serological testing
Incident HCV infection was dened as a positivelaboratory
test result for HCVantibodies (anti-HCV) on both enzyme
immunoassay (ORTHO HCV version 3.0 enzyme-linked
immunosorbent assay; Ortho-Clinical Diagnostics,
Raritan, New Jersey, USA) and recombinant immunoblot
assays (recombinant immunoblot assay; Chiron Corpor-
ation, Emeryville, California, USA) from blood samples
collected at 3 or 6-month follow-up assessments. HIV
antibody testing employed standard enzyme-linked immu-
nosorbent assay and Western blot procedures.
Statistical analyses
We conservatively calculated sample size by basing it on the
behaviors we expected to be least prevalent at baseline from
previous studies [22,48]. Estimating 20% baseline preva-
lence, 20% attrition over the 6-month follow-up period
and setting the type I error rate at 0.05 for a two-tailed test
with 80% power, we needed 725 participants per arm to
detect at least a 30% decrease in each risk behavior among
PEI participants compared with VDI participants.
All analyses were performed in a blinded fashion. Between-
group differences at baseline were determined using chi-
square tests for nominal variables or Wilcoxon rank sum
tests for count variables. A priori hypotheses guided
intention-to-treat analyses [49]. To account for variation
between participants in time to follow-up, 6-month
intervention effects were assessed testing whether differ-
ential rates of linear change occurred over the entire study
period using a group-by-time interaction.
Injection outcomes were modelled by using cumulative
logit models for repeated ordinal data with the inclusion of
a random intercept to account for baseline participant
differences [50,51]. Multivariate models were tadjusting
for city, race, sex, cohort size, age, and men who have sex
with-men (MSM) or women who have sex with women
sex when these variables were associated with the outcome
measures. Proportional odds ratios and 95% condence
limits (CL) were estimated using SAS Proc NLMIXED
[52] to determine intervention effects on 6-month changes
in risk behavior. A score chi-square test was used to
evaluate the proportional odds assumption that the
intervention effect was constant for all seven response
categories in each outcome measure.
The composite injection variable was created posthoc from
the six injection variables because the interventions effect
was in the same direction across variables. Given that
missing data on individual injection variables could have
affected the composite variable, we modelled each
individual injection variable separately and estimated the
overall injection risk as the weighted average of the odds
ratios from separate models for each injection outcome.
Weights were the inverse variances for the intervention
effect from each model. To computevariances and 95% CL,
correlations between these non-independent measures for
each behavior were calculated from participants average
monthly behavior change score for each outcome.
Unprotected sexual behavior outcomes were modelled
using a negative binomial t of repeated count data that
included a random intercept to account for within-subject
correlation [51,52]. Zero-inated negative binomial
models (STATA software, version 8; STATA Corp.,
College Station, Texas, USA) with robust variance
estimates and two components, a logit component of
the zero probability and a negative binomial component of
the frequency count, were implemented to capture
excessive zero responses and extreme values [53]. Wald
and likelihood-ratio tests for evaluating tandtheVuong
test for assessment of zero ination were computed [54]. A
summary measure for the six sexual outcome models was
not applicable because heterogeneous associations with the
outcomes by arm were observed. All model results are
reported in terms of 6-month change.
Results
A total of 3285 IDU completed the baseline assessment.
Of 2062 participants who tested HIV and HCV antibody
negative and were eligible for the trial, 854 were enrolled
(Fig. 1). Enrolled participants were slightly older than
non-enrolled eligible participants (Table 1). In addition,
1926 AIDS 2007, Vol 21 No 14
Page 4
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
enrolled participants were less likely to have been
homeless during the 6 months before baseline (43.2%
versus 50.8%; P < 0.001), but were similar on all other
sociodemographic and outcome variables (data not
shown). Of the participants enrolled, 431 were randomly
assigned to the PEI and 423 to the VDI arm. The only
signicant baseline difference between these arms was in
the percentage of MSM (Table 2). No adverse events
occurred in either arm.
Overall, 712 (83%) participants returned for at least one
follow-up assessment (Fig. 1) and were included in the
analysis; 485 (57%) returned for both. Slightly more PEI
than VDI participants returned for the 3-month
assessment (68 versus 61%; P ¼ 0.04), but return rates
were similar at 6 months (77 versus 74%; P ¼ 0.30).
Effect of the peer education intervention
All six injection outcome variables and the composite
index measure decreased signicantly at follow-up
compared with baseline among PEI participants, as did
all but one measure in the VDI arm (Table 3). Declines in
the PEI arm compared with the VDI arm ranged from 26
to 39% across measures, although none reached statistical
signicance individually. The intervention effect was,
however, statistically signicant for the composite
measure [unweighted average of participants responses
to the six outcome measures; proportional odds ratio
(POR) 0.64; 95% CL 0.44, 0.94]. Furthermore, a
weighted average from models of the six individual
outcome measures demonstrated a 29% greater decline in
overall risk among PEI compared with VDI participants
(POR 0.71; 95% CL 0.52, 0.97).
Participants within each trial arm reported fewer
unprotected sex acts in all categories at follow-up com-
pared with baseline, although not all outcomes were
statistically signicant (Table 4). A statistically signicant
difference between trial arms was observed for only one
sexual outcome measure. After adjusting for potential
confounders, unprotected anal sex with casual/sex trade
partnersdecreased more among VDI participants com-
pared with PEI participants. Given a high proportion of
zero responses, particularly for the frequency of unpro-
tected anal sex outcomes, we reanalysed the data using
zero-inated negative binomial mixture models. We
observed a signicantly greater reduction in the frequency
component of anal sex with a main partneramong PEI
versus VDI participants (risk ratio 0.54; 95% CL 0.33,
0.87). In addition, there was no longer a signicant
difference in anal sex with casual/sex trade partners by
trial arm (data not shown).
The overall incidence of HCV infection was 18.1/
100 person-years (95% CL 14.4, 23.0). Using Poisson
regression to control for site, race, sex, age, and cohort size,
we found no difference in HCV incidence rates between
PEI and VDI participants (relative risk 1.15; 95% CL 0.72,
1.82). No participants seroconverted to HIV positive in
either trial arm during the 427 person-years of follow-up.
Discussion
Our intervention, which provided information and skills
for reducing sexual and injection risk associated with HIV
and HCV infection and encouraged young adult IDU to
adopt prosocial roles as peer educators, produced a
29% greater reduction across six injection risk behaviors
compared with the attentioncontrol participants. Previ-
ous intervention trials among HIV-positive IDU [55] have
demonstrated decreases in unprotected sex and sexually
transmitted diseases, but not needle sharing as shown
here.
This is the rst intervention tested in a randomized
controlled trial specically focused on young HIV and
Reducing injection risk behaviors in young injection drug users Garfein et al. 1927
Table 1. Characteristics of 1530-year-old injection drug users, 3rd Collaborative Injection Drug Users Study/Drug Users Intervention Trial,
20022004.
Variable
Eligible for trial (HIV and HCV antibody negative)
Completed baseline
(n ¼ 3285)
Not enrolled
(n ¼ 1172)
Enrolled
(n ¼ 854)
Total
(n ¼ 2026)
Age in years Mean (IQR) 23.8 (2127) 22.9 (2025) 23.8 (2127)
M
23.8 (2127)
Minors 1517 years 2.2% 1.8% 2.7% 3.0%
Sex Male 68.9% 67.9% 66.5% 67.3%
Female 30.4% 31.4% 32.8% 32.0%
Transgender 0.8% 0.7% 0.7% 0.7%
Race/ethnicity NH black 7.7% 9.1% 8.4% 8.8%
NH white 64.0% 65.3% 63.3% 64.5%
Hispanic 16.8% 13.4% 17.1% 15.0%
Other/mixed 11.5% 12.2% 11.2% 11.8%
Homeless in the past 6 months 49.6% 50.8% 43.2%
M
47.6%
HIV antibody positive 2.9% –––
HCV antibody positive 34.4% –––
HCV, Hepatitis C virus; NH, non-Hispanic; IQR, interquartile range.
M
P < 0.001 comparing enrolled with non-enrolled participants.
Page 5
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
HCV-uninfected IDU that stressed risk reduction for
both infections. As HCV is primarily spread through
parenteral routes among IDU, interventions that decrease
unhygienic injection practices could reduce its spread;
and given that HCV-associated liver disease is often silent
for decades, interventions that merely delay infection
among young IDU could signicantly impact public
health. Similarly, parenteral HIV transmission would also
be decreased. Despite self-reported injection risk
reduction in the PEI group, HCV infection incidence
did not differ across trial arms, suggesting that exposure to
infected blood must be virtually eliminated, rather than
simply reduced, to prevent HCV infection. In studies of
needlestick injuries sustained by healthcare workers, the
proportion of infections after percutaneous injury was
10-fold higher for HCV than HIV [56]. This could also
explain why HCV but not HIV transmission readily
occurs among IDU who share injection paraphernalia
other than syringes.
The effect of the PEI on condom use was less clear. As with
injection risk behaviors, the frequency of all sexual risk
behaviors declined in both trial arms over time. The only
signicant difference across arms was a greater decrease in
unprotected anal sex with casual/sex trade partners among
VDI versus PEI participants. Of note is the fact that this was
the least commonly reported sexual risk behavior and
less than 8% of men reported being MSM, explaining some
of the instability in this estimate. Upon reanalysis using a
mixture model, the difference in anal sex with casual/sex
trade partners became non-signicant. Instead, results from
the frequency component indicated that unprotected anal
1928 AIDS 2007, Vol 21 No 14
Table 2. Results of random assignment to trial arm among 1530-year-old injection drug users, 3rd Collaborative Injection Drug Users Study/
Drug Users Intervention Trial, 20022004.
Variable
a
Peer education intervention
(n ¼ 431)
Video discussion intervention
(n ¼ 423)
Participant characteristics
Years of age at baseline visit 23.8 (3.5) 23.8 (3.7)
Male sex, N (%)
b
285 (66.1) 282 (66.7)
Did not complete 12th grade, N (%) 147 (34.2) 147 (34.8)
Most income from full or part-time job in past 6 months, N (%) 188 (43.8) 182 (43.3)
Mostly lived away from parents in past 6 months, N (%) 269 (62.7) 260 (61.8)
Considered self homeless in past 6 months, N (%) 187 (43.4) 180 (42.8)
Ever traded sex for money or drugs, N (%) 109 (25.5) 115 (27.6)
Ever incarcerated, N (%) 298 (69.1) 288 (68.1)
In drug treatment now (excl. support groups), N (%) 41 (11.1) 50 (13.4)
Age of rst injection in years 19.5 (3.8) 19.3 (4.0)
Duration of injection drug use in years 4.2 (3.4) 4.4 (3.5)
Injection behaviors in past 3 months
Mostly injected heroin not mixed with other drugs, N (%) 323 (77.6) 316 (77.5)
Total no. of injections 202 (186) 202 (231)
Receptively shared syringes
c
2.1 (1.5) 2.1 (1.6)
Cleaned syringe with bleach before receptively shared syringes
c
3.8 (2.4) 3.5 (2.4)
Used new syringes to divide drugs
c
2.8 (1.6) 2.9 (1.6)
Split drugs while in solution (e.g. backload)
c
2.1 (1.8) 2.1 (1.7)
Split drugs with a syringe used by another IDU
c
3.4 (2.0) 3.2 (2.1)
Shared cooker
c
3.1 (2.2) 3.1 (2.1)
Shared cotton
c
2.7 (2.1) 2.6 (2.0)
Shared rinse water
c
2.8 (2.1) 2.7 (2.1)
Shared any injection equipment with other IDU
d
2.9 (2.2) 2.8 (2.1)
Sexual behaviors in past 3 months
Total no. of sex partners 6.5 (42.2) 4.3 (10.8)
Man who had male sex partners, N (%) 15 (5.5) 30 (10.8)
e
Woman who had female sex partners, N (%) 26 (18.3) 26 (18.8)
No. of unprotected vaginal sex acts
With main steady partner 29.0 (62.1) 29.7 (69.9)
With non-main steady partner 1.6 (6.7) 1.3 (7.2)
With casual/sex trade partner 5.1 (22.9) 6.7 (53.0)
No. of unprotected anal sex acts
With main steady partner 2.7 (11.5) 2.4 (12.8)
With non-main steady partner 0.4 (3.1) 0.6 (7.0)
With casual/sex trade partner 0.8 (5.1) 3.3 (50.3)
IDU, Injection drug user.
a
Data are expressed as mean (SD) unless otherwise noted.
b
Male sex includes three participants in each arm who reported being transgendered male to female.
c
Response categories: 1, never;2,rarely;3,less than half the time;4,about half the time;5,more than half the time;6,almost always;
7, always; Cleaned syringe with bleach and Used new syringes to divide drugs were reverse-coded to make higher values represent greater risk.
d
Response categories for partners shared injection equipment with: 1, none;2,almost none;3,less than half;4,about half;5,more than half;
6, almost all;7,all.
e
Chi-square P ¼ 0.02 for the difference between trial arms.
Page 6
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
sex with main partners decreased 46% more among PEI
than VDI participants. The intervention design assumed
that behavior change with steady partners would translate
into behavior change among non-steady partnerships;
therefore, PEI messages focused mostly on steady partner-
ships. In retrospect, this assumption may be inaccurate, but
could explain why the PEI produced greater, although
non-signicant, decreases in unprotected sex only with
steady partners.
All injection and several sexual risk behaviors decreased
signicantly at follow-up compared with baseline among
PEI participants; however, decreases also occurred among
VDI participants, emphasizing the importance of measur-
ing efcacy relative to a concurrent control group.
Considering the possibility that participants in both
arms learned through successive assessments that under-
reporting risk behaviors shortened the interview, we
compared responses on repeated baseline assessments
from participants who had to be reassessed when
3 months had lapsed before they were enrolled. No
consistent downward trends in self-reported risk behaviors
were found.
Failure to detect signicantly greater risk reductions
among PEI versus the VDI participants may be caused by
features inherent in the trials design. For example, by
giving all participants identical pre and posttest counsel-
ling before enrollment, and making condoms, bleach kits,
HIVand HCV-related educational pamphlets and medical
Reducing injection risk behaviors in young injection drug users Garfein et al. 1929
Table 3. Comparison of 6-month changes in injection risk behaviors among 1530-year-old injection drug users participating in a peer
education intervention or video discussion intervention, 3rd Collaborative Injection Drug Users Study/Drug Users Intervention Trial,
20022004.
PEI arm 6-month change
POR (95% CL)
VDI arm 6-month change
POR (95% CL)
Intervention effect
a
(change in
PEI/VDI) POR (95% CL) P
Composite index of all six items
b
0.18 (0.14, 0.25) 0.29 (0.22, 0.39) 0.64 (0.44, 0.94) 0.024
Summary measure of all six models
b
0.24 (0.19, 0.31) 0.37 (0.29, 0.46) 0.71 (0.52, 0.97) 0.029
Individual items
Proportion of injections in which
c
Injected with used syringe 0.29 (0.22, 0.40) 0.42 (0.31, 0.57) 0.71 (0.47, 1.07) 0.105
Used new syringe to divide drugs
d
0.49 (0.29, 0.83) 0.88 (0.54, 1.42) 0.61 (0.31, 1.19) 0.145
Shared cooker 0.22 (0.16, 0.29) 0.35 (0.27, 0.46) 0.73 (0.50, 1.06) 0.099
Shared cotton 0.24 (0.17, 0.33) 0.38 (0.28, 0.50) 0.69 (0.46, 1.04) 0.074
Shared rinse water 0.27 (0.20, 0.36) 0.35 (0.27, 0.47) 0.74 (0.51, 1.09) 0.133
Proportion of partners with whom you
shared injection paraphernalia
e
0.13 (0.09, 0.18) 0.22 (0.16, 0.31) 0.69 (0.44, 1.10) 0.117
CL, Condence limit; PEI, peer education intervention; POR, proportional odds ratio; VDI, video discussion intervention.
a
Estimates result from random intercept proportional odds models (cumulative logit). The intervention effect is estimated with a time trend by study
arm, two-way interaction term in the model. Signicant covariates potentially included in models were city, race, sex, cohort size, and age.
b
The composite index was calculated by adding the response values from six outcome measures and dividing by six; the summary measure is the
weighted average of the odds ratios from separate models for each individual injection outcome.
c
Ordinal measures: 1, never;2,rarely;3,less than half the time;4,about half the time;5,more than half the time;6,almost always;
7, always.
d
Response categories were reverse-coded so that higher values represent greater risk.
e
Response categories for partners shared injection equipment with: 1, none;2,almost none;3,less than half;4,about half;5,more than half;
6, almost all;7,all.
Table 4. Comparison of 6-month changes in sexual risk behaviors among 1530-year-old injection drug users participating in a peer education
intervention or video discussion intervention, 3rd Collaborative Injection Drug Users Study/Drug Users Intervention Trial, 20022004.
No. of acts done without condom
PEI arm 6-month change
RR (95% CL)
VDI arm 6-month change
RR (95% CL)
Intervention effect
a
(change in
PEI/VDI) RR (95% CL) P value
Total acts with all partners 0.67 (0.55, 0.83) 0.77 (0.62, 0.95) 0.90 (0.67, 1.21) 0.486
Vaginal sex with main partner 0.65 (0.52, 0.81) 0.72 (0.58, 0.90) 0.90 (0.65, 1.23) 0.489
Vaginal sex with non-main other steady 0.72 (0.36, 1.45) 0.56 (0.29, 1.10) 1.10 (0.41, 2.97) 0.849
Vaginal sex with casual/sex trade partner(s) 0.85 (0.52, 1.38) 0.69 (0.43, 1.10) 1.17 (0.60, 2.29) 0.648
Anal sex with main partner 0.52 (0.31, 0.85) 0.84 (0.56, 1.25) 0.66 (0.35, 1.23) 0.192
b
Anal sex with non-main other steady 0.13 (0.01, 1.13) 0.82 (0.25, 2.67) 0.22 (0.02, 2.05) 0.185
Anal sex with casual/sex trade partner(s) 0.88 (0.43, 1.83) 0.29 (0.14, 0.59) 3.15 (1.13, 8.79) 0.028
b
CL, Condence limit; PEI, peer education intervention; RR, risk ratio; VDI, video discussion intervention.
a
Estimates result from negative binomial regression models with inclusion of a random intercept to account for within-subject correlation in which
city, race, sex, cohort size, age, and men who have sex with men or women who have sex with women status were considered as potential
covariates; women who have sex with women and cohort size were not signicant in any model.
b
To account for excessive zero responses and extreme values in the data, we reanalysed the data using zero-inated negative binomial models
(STATA software, version 8) with robust variance estimates and two components, a logit component of the zero probability and a negative binomial
component of the frequency count. We observed a signicantly greater reduction in the frequency component of unprotected anal sex with a main
partner among PEI participants (RR 0.54; 95% CL 0.33, 0.87), and there was no longer a signicant difference in unprotected anal sex with casual/
sex trade partner(s) by trial arm using this mixture model.
Page 7
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
and drug treatment referrals available in both interven-
tions, all participants received similar information. An
overly powerful control condition has the effect of
minimizing the relative efcacy of the intervention,
which has been observed in previous behavioral
intervention trials [28]. Although this trial was not
designed to measure community-level effects of the PEI,
such an effect was desired. If randomization placed
network members into opposing trial arms, PEI
participants may have practised what they learned with
their VDI arm peers. Consequent reductions in risk
behavior among VDI participants would result in
underestimating the effect of the PEI on behavior change
compared with the control arm.
Our study was limited by the fact that participants had to
return for test results to learn of their eligibility, and then
had to return for a third visit, sometimes weeks later, for
random selection. Consequently, many eligible partici-
pants were lost in the process. We thus lost statistical power
to detect differences in less prevalent behaviors, such as
unprotected sex. Second, intensive risk-reduction training
exclusive to the PEI may have caused PEI participants to
feel greater pressure than VDI participants to underreport
risk behaviors at follow-up in an effort to please their
trainers, thus biasing results in favor of the PEI. The use of
ACASI to minimize socially desirable responding, and
seeing an intervention effect for injection but not sexual
behaviors, makes this bias appear unlikely. Third, attrition
could have biased results, but 83% returned for at least one
follow-up visit and we observed no signicant differences
between returners and non-returners. Fourth, it is
unknown whether our sample is representative of all
young adult IDU; however, by employing multiple
recruitment strategies in ve cities, a broad cross-section
of IDU was included. Fifth, intention-to-treat analyses
are subject to bias if intervention attendance is low or
uneven across arms. Although only 56% of participants
overall attended all six intervention sessions, all participants
attended at least the rst session and attendance at each of
the remaining sessions was reasonably high; on average 77%
(range 68100%) of PEI participants and 78% (range 71
100%) of VDI participants attended each session.
Furthermore, attendance was similar across trial arms.
Finally, sustainability of the intervention effect may not
have been measurable in a 6-month follow-up period.
The strengths of this study include the use of a randomized
controlled trial design with a dose-equivalent control
condition. This design minimized the effect of merely
paying attention to participants assigned to the intervention
group (Hawthorne effect), which could be signicant in
marginalized populations such as young IDU. We
examined multiple injection and sexual behavioral out-
comes, plus a biological outcome, seeking consistency
across outcomes to weight empirical evidence of an
intervention effect. As behavioral intervention efcacy
often degrades over time, we averaged efcacy over a
6-month period to be conservative. Finally, our multisite
design provides better generalizability than a single-site
trial.
This peer education intervention offers a means for
substantially reducing injection risk behaviors among
IDU, particularly those who have been injecting for a
short time and are at high risk of blood-borne infections.
The potential community-wide effects of training peer
educators could further impact the spread of HIV and
HCVamong young IDU. Intensifying the intervention to
eliminate, rather than reduce, injection risk may be
required to decrease HCV incidence among IDU
signicantly.
Acknowledgements
The authors would like to thank Drs Ann OLeary, Carl
Latkin, and Janet Moore for their expert consultation on
both the intervention and efcacy trial designs; Drs
Kathleen Sikkema and Susan Tross for reviewing and
providing invaluable feedback on the pilot intervention;
Brigette Finkelstein-Ulin and Linda Moyer (CDC,
Division of Viral Hepatitis) for their contribution to
the development of the hepatitis educational and
counselling materials; and members of Community/Peer
Advisory Boards and HIV Program Review Panels at
each site for providing constructive feedback on the
intervention and trial designs.
The authors would like to recognize the following
individuals who made important contributions to the
development and implementation of the study: Yvette
Bowser, Peter ODriscoll, Janet Reeves, Marcella Sapun
(Baltimore); Angus Atkins-Trimnell, Mary Bonilla, David
Cosey, Jaime Delgado, Julio Garcia, Michelle Giles, Erin
Kubalanza, Michael Phillips, Edward Snulligan (Chicago);
Marrisa Axelrod, Elizabeth Faber, Lawrence Fernandez Jr,
Christian Geannette, Roberto Rojas (Los Angeles); Ebele
Benjamin, Sebastian Bonner, Micaela Coady, Joanna Cruz,
Sandra DelVecchio, Dirk Jackson, Gregory Malave, Joan
Monserrate, Danielle Ompad, Clarisse Miller OShea,
Yingfeng Wu, Manny Yonko (New York); Stanley Brown,
Rong Lee, Susan Nelson, Jef St De Lore, Carrie Shriver,
Jeanette Frazier, Jean Pass, Paul Swenson (Seattle); Vincent
Raimondi, Scott Santibanez, Roberto Valverde (CDC);
Wendi Kuhnert, Himal Dhotre, Leigh Farrington, (CDC
Division of Viral Hepatitis); Suzette Bartley, Dollene
Hemmerlein (CDC Serum Bank Branch).
Other members of the 3rd Collaborative Injection Drug
Users Study/Drug Users Intervention Trial (CIDUS III/
DUIT) study group include (listed alphabetically): Susan
Bailey, Marie Bailey-Kloch, Jennifer V. Campbell, Joyce
Fitzgerald, Paige Ingram, Peter Kerndt, Yuko Mizuno,
1930 AIDS 2007, Vol 21 No 14
Page 8
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Nadine Snyder, Andrea Swartzendruber, Karla Wagner,
David Vlahov, Ian Williams, and Karen Yen-Hobelman.
Sponsorship: This study was funded in its entirety by a
cooperative agreement from the Centers for Disease
Control and Prevention U64/CCU317662, U64/
CCU517656, U64/CCU917655, U64 CCU217659,
U64/CCU017615; Institutional Review Board no.
CDC-NCHSTP-2934.
Disclaimer: The findings and conclusions in this report
are those of the authors and do not necessarily
represent the views of the Centers for Disease Control
and Prevention. All authors made significant contribu-
tions to the study conceptualization, design and
implementation, as well as to conceiving and writing
of this manuscript.
Trial registration: Collaborative Injection Drug Users
Study III/Drug Users Intervention Trial (CIDUS III/
DUIT), NCT00164372, http://clinicaltrials.gov/ct/
show/NCT00164372?order=1.
References
1. Sullivan LE, Metzger DS, Fudala PJ, Fiellin DA. Decreasing
international HIV transmission: the role of expanding access
to opioid agonist therapies for injection drug users. Addiction
2005; 100:150158.
2. Chamot E, De Saussure P, Hirschel B, Deglo JJ, Perrin LH.
Incidence of hepatitis C, hepatitis B and HIV infections among
drug users in a methadone-maintenance programme. AIDS
1992; 6:430431.
3. Crofts N, Hopper JL, Bowden DS, Breschkin AM, Milner R,
Locarnini S. Hepatitis C virus infection among a cohort of
Victorian injecting drug users. Med J Aust 1993; 159:237
241.
4. Thomas DL, Vlahov D, Solomon L, Cohn S, Taylor E, Garfein R,
Nelson KE. Correlates of hepatitis C virus infection among
injection drug users. Medicine 1995; 74:212220.
5. Garfein RS, Vlahov D, Galai N, Doherty MC, Nelson KE. Viral
infections in short-term injection drug users: the prevalence of
the hepatitis C, hepatitis B, human immunodeficiency, and
human T-lymphotropic viruses. Am J Public Health 1996;
86:655661.
6. Smyth BP, Barry J, Keenan E. Retrospective cohort study
examining incidence of HIV and hepatitis C infection among
injecting drug users in Dublin. J Epidemiol Commun Health
2003; 57:310311.
7. Judd A, Hickman M, Jones S, McDonald T, Parry JV, Stimson
GV, Hall AJ. Incidence of hepatitis C virus and HIV among new
injecting drug users in London: prospective cohort study. BMJ
2005; 330:2425.
8. Van Beek I, Dwyer R, Dore GJ, Luo K, Kaldor JM. Infection with
HIV and hepatitis C virus among injecting drug users in a
prevention setting: retrospective cohort study. BMJ 1998;
317:433437.
9. Hagan H, Thiede H, Weiss NS, Hopkins SG, Duchin JS,
Alexander ER. Sharing of drug preparation equipment as a risk
factor for hepatitis C. Am J Public Health 2001; 91:4246.
10. Thorpe LE, Ouellet LJ, Hershow R, Bailey SL, Williams IT,
Williamson J, et al. Risk of hepatitis C virus infection among
young adult injection drug users who share injection equip-
ment. Am J Epidemiol 2002; 155:645653.
11. Hahn JA, Page-Shafer K, Lum PJ, Bourgois P, Stein E, Evans JL,
et al. Hepatitis C virus seroconversion among young injection
drug users: relationship and risk. J Infect Dis 2002; 186:1558
1564.
12. Grund JP, Kaplan CD, Adriaans NF, Blanken P, Huisman J. The
limitations of the concept of needle sharing: the practice of
‘frontloading’. AIDS 1990; 4:819821.
13. Koester S, Glanz J, Baron A. Drug sharing among heroin net-
works: implications for HIV and hepatitis B and C prevention.
AIDS Behav 2005; 9:2739.
14. Schoenbaum EE, Hartel D, Selwyn PA, Klein RS, Davenny K,
Rogers H, et al. Risk factors for human immunodeficiency virus
infection in intravenous drug users. N Engl J Med 1989;
321:874879.
15. Strathdee SA, Galai N, Safaiean M, Celentano DD, Vlavov D,
Johnson L, Nelson KE. Sex differences in risk factors for HIV
seroconversion among injection drug users: a 10-year perspec-
tive. Arch Intern Med 2001; 161:12811288.
16. Strathdee SA, Sherman SG. The role of sexual transmission of
HIV infection among injection and noninjection drug users.
J Urban Health 2003; 80:iii7iii14.
17. Kral AH, Bluthenthal RN, Lorvick J, Gee L, Bacchetti P, Edlin BR.
Sexual transmission of HIV-1 among injection drug users in San
Francisco, USA: risk-factor analysis. Lancet 2001; 357:1397
1401.
18. Garfein RS, Doherty MC, Monterroso ER, Thomas DL, Nelson
KE, Vlahov D. Prevalence and incidence of hepatitis C virus
infection among young adult injection drug users. J Acquir
Immune Defic Syndr 1998; 18:S11S19.
19. Nicolosi A, Leite ML, Musicco M, Molinari S, Lazzarin A.
Parenteral and sexual transmission of human immunodefi-
ciency virus in intravenous drug users: a study of seroconver-
sion. The North Italian Seronegative Drug Addicts (NISDA)
Study. Am J Epidemiol 1992; 135:225233.
20. Van Ameijden EJ, Van den Hoek JA, Van Haastrecht HJ, Coutinho
RA. The harm reduction approach and risk factors for human
immunodeficiency virus (HIV) seroconversion in injection drug
users, Amsterdam. Am J Epidemiol 1992; 136:236243.
21. Nelson KE, Vlahov D, Solomon L, Cohn S, Mun
˜
oz A. Temporal
trends of incident HIV infection in a cohort of injection drug
users in Baltimore, Maryland. Arch Intern Med 1995; 155:
13051311.
22. Latka M, Ahern J, Garfein RS, Ouellet L, Kerndt P, Morse P, et al.
Prevalence, incidence, and correlates of chlamydia and
gonorrhea among young adult injection drug users. J Subst
Abuse 2001; 13:7388.
23. Des Jarlais DC, Diaz T, Perlis T, Vlavov D, Maslow C, Latka M,
et al. Variability in the incidence of human immunodeficiency
virus, hepatitis B virus, and hepatitis C virus infection among
young injecting drug users in New York City. Am J Epidemiol
2003; 157:467471.
24. Neaigus A, Suan M, Friedman SR, Goldsmith DS, Stepherson B,
Mota P, et al. Effects of outreach intervention on risk reduction
among intravenous drug users. AIDS Educ Prev 1990; 2:253
271.
25. Stephens RC, Feucht TE, Roman SW. Effects of an intervention
program on AIDS-related drug and needle behavior among
intravenous drug users. Am J Public Health 1991; 81:568571.
26. McClusker J, Stoddard AM, Zapka JG, Morrison CS, Zorn N,
Lewis BF. AIDS education for drug users: evaluation of short-
term effectiveness. Am J Public Health 1992; 82:533540.
27. Needle RH, Coyle SL. Community-based outreach risk reduc-
tion strategy to prevent HIV risk behaviors in out-of-treatment
injection drug users (IDUs). In: National Institutes of Health
Consensus Development Conference on Interventions to
Prevent HIV Risk Behaviors. Rockville, MD, 1123 February
1997. National Institute on Drug Abuse (US). 36.
28. Van den Hoek A, Van Haastrecht HJA, Coutinho RA. Little
change in sexual behavior in drug users in Amsterdam. In: Sixth
International Conference on AIDS. San Francisco, CA 2025
June 1990. Abstract 122.
29. Watkins KE, Metzger D, Woody G, McLellan AT. Determinants
of condom use among intravenous drug users. AIDS 1993;
7:719723.
30. Barnard MA. Needle sharing in context: patterns of sharing