WHO/HIV/2012.25
Annex 4. GRADE evidence profiles
Question 1: Should rapid HBV vaccination regimen versus a standard HBV vaccination regimen be used among PWID?
Author(s): Elie Akl Date: 2012-‐05-‐8
Bibliography: Christensen 2004. Brisette 2002. Systematic review by Amato et al.
Quality assessment No of patients Effect
Quality Importance No of
studies Design Risk of
bias Inconsistency Indirectness Imprecision Other considerations
Rapid HBV vaccination
regimen
Standard HBV vaccination regimen
Relative
(95% CI) Absolute Completion of vaccination regimen (follow-‐up mean 6 months)
21,2 randomised trials
very serious3
no serious inconsistency4
serious5 no serious imprecision
none 297/405
(73.3%)
184/400 (46%)
RR 1.6 (1.42 to 1.81)
276 more per 1000 (from 193 more to 373
more)
ÅOOO VERY LOW
CRITICAL
Immune response 11 randomised
trials
serious3 serious6 serious5 serious7 none 148/365
(40.5%)
155/407 (38.1%)
RR 1.12 (0.84 to 1.49)8
46 more per 1000 (from 61 fewer to 187 more)
ÅOOO VERY LOW
CRITICAL
Satisfaction -‐ not measured
0 -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ IMPORTANT
1 Brissette 2002 conducted in Quebec, Canada among 908 cocaine/heroin users. It compared an accelerated schedule regimen (0, 2, 4 weeks) using a high dose vaccine (40 mcg) to a standard schedule regimen (0, 4, 24 weeks) using a standard dose vaccine (10 mcg). There was no difference in seroprotection between the two regimens.
2 Christensen 2004 was conducted in Copenhagen, Denmark among prisoners who were injection drug users. It compared an accelerated regimen (0-‐1-‐3 weeks) to a standard regimen (0-‐1-‐6 months).
3 Brissette 2002: sequence generation adequate (computer generated) and allocation concealment probably adequate (“opening an envelope”). However, no blinding of outcome adjudicator reported; no mention whether analysis was intention to treat; unclear rate of loss to follow-‐up. Christensen 2004: essentially not randomized (34 randomized while 566 not randomized); no blinding of outcome adjudicator reported; no mention whether analysis was intention to treat; significant rate of loss to follow-‐up
4 I2=39%
5 Rated down for indirectness of outcome (regimen completion and immune response being surrogates for HBV infection). Also both studies were conducted in high income countries and Christensen recruited only prisoners.
6 I2=81%
7 Relatively small number of participants and of events
8 Calculated by meta-‐analyzing (using fixed effect model) data from high dose and low dose groups
2
Question 2: Should incentives for HBV vaccination completion versus no incentives be used among PWID?
Author(s): Elie Akl Date: 2012-‐04-‐23
Bibliography: Seal 2003. Stitzer 2010. Systematic review by Amato et al.
Quality assessment No of patients Effect
Quality Importance No of
studies Design Risk of
bias Inconsistency Indirectness Imprecision Other considerations
Incentives for completion of HBV vaccination
No incentives
Relative
(95% CI) Absolute Vaccine completion (follow-‐up mean 6 months)
21,2 randomised trials
serious3 no serious inconsistency4
no serious indirectness5
Serious6 none 43/61
(70.5%)
17/61 (27.9%)
RR 2.53 (1.64 to 3.9)
426 more per 1000 (from 178 more to 808 more)
ÅÅOO LOW
CRITICAL
1 Seal 2003 was conducted among IDUs in San Francisco, USA. Eligible participants received their first dose of vaccine and were randomized to either monthly monetary incentives (monthly visit to neighbourhood field site once each month for 6 months to receive $20) or a weekly contact with an outreach worker (contacted participants weekly to provide safe injection information and reminder of upcoming vaccine appointments). Participants were also randomized to different schedules (rapid or accelerated) and 2 doses of vaccine (Recombivax 10 microg or 40 microg).
2 Stitzer 2010 was conducted in the USA among cocaine drug users. All participants received $20 for completing study intake procedures (screening, consent and initial injection) and $10 after each weekly visit to cover transportation costs. In the incentive participants, they were draws for attendance at each weekly visit with the chance to win prizes. They also received cash bonuses for attending injection visits. These started at $20 (intake visit) and increased by $5 each month, up to $50with total possible = $265. Overall incentive group earnings averaged $501 out of $751 possible. The control
participants received no incentive.
3 Seal 2003: unclear methods of sequence generation and allocation concealment; no blinding of outcome adjudicator reported; analysis was intention to treat; unclear rate of loss to follow-‐up. Stitzer 2010: unclear methods of sequence generation and allocation concealment; no blinding of outcome adjudicator reported; unclear whether analysis was intention to treat; unclear rate of loss to follow-‐up
4 I2=48%
5 Regimen completion is a surrogate for HBV infection and this indirectness of outcome was accounted for when rating down the quality of evidence to low.
6 Relatively small number of participants and small number of events
Question 3: Should low dead space-‐syringes versus high dead space syringes be provided to PWID?
Author(s): Elie Akl Date: 2012-‐04-‐24
Settings: needle syringe programs
Bibliography: Gyarmathy 2010. Zule 2009. Systematic review by Amato et al.
Quality assessment No of patients Effect
Quality Importance No of
studies Design Risk of
bias Inconsistency Indirectness Imprecision Other considerations
Low dead space syringes (LDSS)
High dead space syringes (HDSS)
Relative
(95% CI) Absolute HBV infection -‐ not reported
0 -‐ -‐ -‐ -‐ -‐ none -‐ -‐ -‐ -‐ CRITICAL
HCV infection (assessed with: serologic testing) 21,2 observational
studies
serious3 no serious inconsistency4
no serious indirectness
no serious imprecision
none5 264/692
(38.2%)
501/645 (77.7%)
RR 0.49 (0.44 to 0.55)
396 fewer per 1000 (from 350 fewer to 435
fewer)
ÅOOO VERY LOW
CRITICAL
HIV infection (assessed with: serologic testing) 21,2 observational
studies
serious3 no serious inconsistency4
no serious indirectness
no serious imprecision
none5 24/692
(3.5%)
70/645 (10.9%)
RR 0.29 (0.18 to 0.46)
77 fewer per 1000 (from 59 fewer to 89 fewer)
ÅOOO VERY LOW
1 Gyamarthy 2010 was conducted among IDUs in Hungary and Lithuania (mean age 28; 77% males).
2 Zule 2009 was conducted among IDUs in USA (mean age 40, 74% males).
3 Unclear how the SR calculated RR based on cross-‐sectional data
4 High I2 but individual effect estimates of study both suggest major reduction in outcome
5 No rating up for large effect size given the concern about the analytical method
4
Question 4: Should psychosocial interventions vs. no psychosocial interventions be used in people who inject drugs?
Author(s): Elie Akl Date: 2012-‐05-‐08
Bibliography: Abou Saleh 2008; Gilbert 2010; Stein 2009; Tucker 2004; Wu 2007; Zule 2009. Systematic review by Amato et al.
Quality assessment No of patients Effect
Quality Importance No of
studies Design Risk of bias Inconsistency Indirectness Imprecision Other considerations
Psychosocial interventions
No psychosocial interventions
Relative
(95% CI) Absolute HCV infection
21,2 randomised trials
serious3,4 no serious inconsistency
no serious indirectness5
serious6 none 9/93
(9.7%)
12/94 (12.8%)
RR 0.75 (0.33 to 1.71)
32 fewer per 1000 (from 86 fewer to 91
more)
ÅÅOO LOW
CRITICAL
Needle sharing behavior (assessed with: Self reported; surrogate for HIV infection) 27,8,9 randomised
trials
serious10,11 no serious inconsistency
serious12 serious13 none 41/395
(10.4%)
61/416 (14.7%)
RR 0.7 (0.49 to 1.02)14
44 fewer per 1000 (from 75 fewer to 3
more)
ÅOOO VERY LOW
IMPORTANT
High risk sexual behavior (assessed with: Self reported; surrogate for HIV infection) 18,9 randomised
trials
serious10 no serious inconsistency
serious15 serious none 104/265
(39.2%)
101/286 (35.3%)
RR 1.11 (0.89 to 1.38)16
39 more per 1000 (from 39 fewer to 134
more)
ÅOOO VERY LOW
IMPORTANT
Satisfaction (assessed with: 5 point Likert scale (satisfied or very satisfied)) 117 randomised
trials
no serious risk of bias18
no serious inconsistency
no serious indirectness
very serious6
none 56/57
(98.2%)
50/54 (92.6%)
RR 4.48 (0.48 to 41.42)
from 481 fewer to 1000 more)
ÅÅOO LOW
IMPORTANT Quality of life -‐ not measured
0 -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ IMPORTANT
1 Abou Saleh 2008 was conducted in the UK and included 95 IDUs (mean age 32 years, 70% male, mean duration of injecting drug use: 5.9 years). The enhanced prevention counseling (EPC) intervention consisted of 4 sessions of manual-‐guided intervention. The manual was based particularly on the Brief Intervention (Motivational Enhancement Therapy) and was concerned with the reduction of high-‐
risk sexual behavior and the introduction of safer sex. Duration: 8 weeks. Sessions were carried out by a drug clinic worker who was trained in delivering the intervention. All sessions last between 40-‐60 minutes. The control simple educational counseling (SEC) consisted of ten-‐minute session of information-‐giving intervention about the nature and the risk factors of HCV, with advice on prevention including the need to reduce sharing of injecting equipment and safer injecting practices. It was intended to be a non-‐interactive intervention.
2 Stein 2009 was conducted in the US and included 277 participants using heroin or cocaine (39% IDUs) at least three times weekly, HCV antibody negative, mean age 37.2 years, male: 65% (the SR considered only IDUs). The intervention consisted of 4 session motivational intervention and provision of a written informational handout about local treatment resources or to a control condition receiving. HCV risk behaviors specifically discussed during the session were initiating injection drug use (for non-‐injectors), continued injection drug use, sharing both injection (e.g., syringe, cooker) and non-‐injection (e.g. straws) drug equipment, sharing drug in certain ways (e.g., backloading), receiving tattoos, and having unprotected sex. The individual sessions each lasted 30-‐45 minutes and were conducted at baseline, 1, 3, and 6-‐months post-‐baseline. The control consisted of the provision of only the informational treatment resource list handout.
3 Abou Saleh 2008: allocation was concealed; no blinding reported; analysis was intention to treat; follow-‐up was 82% and 65% at 6 and 12 months.
4 Stein 2009: unclear whether randomization was concealed; no blinding reported; unclear whether analysis was intention to treat; follow-‐up rates were 80% at 6-‐months and 75% at the 24-‐month.
5 Some indirectness of population (Saleh 2008: PWID in the UK; Stein 2009: PWID in the US)
6 The CI did not rule out either a beneficial or a harmful effect on the outcome of interest
7 Gagnon 2010 was conducted in Canada and included 260 IUDs (mean age 34.9 years, 69% male). The intervention consisted of standard intervention plus 4 session of computed tailored messages on the website. The control consisted of standard intervention: needle exchanges, psychosocial support, social and health service referral. Participants to both groups could have discussion with their assigned community worker. Interaction were tailored to the need of the participant and could last from 10 to 30 minutes
8 Zule 2009 was conducted in the USA and included 851 IDUs (mean age 41 years, male: 63%). The intervention consisted of 6 sessions; contents were motivation to change, developing a plan for change, reviewing progress and reaffirming commitment to change, build self efficacy and self regulation skills. Participants were given 10 dollars coupon at the end of each session. The control consisted of an educational intervention; 6 sessions; contents were information of Hepatitis, sharing practices and addiction. Participants were given 10 dollars coupon at the end of session 3 through 6. All participants were offered the HCV, HIV and hepatitis test
9 We excluded Wu 2007 (initially included by the systematic review) because the authors state that the study design was modified to cross-‐sectional due to a very low follow-‐up rate of 12%.
10 Zule 2009: unclear whether randomization was concealed; no blinding reported; unclear whether analysis was intention to treat; follow-‐up rate was 73%.
11 Gagnon 2010: allocation was not concealed; study was not blinded; analysis was not intention to treat; 9.6% lost at short term follow-‐up; 33% lost at long term follow-‐up
12 Outcome is indirect in the 2 studies (reported behavior). Population was indirect in the 2 studies (Zule 2009: PWID in the USA; Gagnon 2009: IUD in Canada).
13 CI includes both values suggesting benefit and values suggesting no effect
14 Analysis using continuous data for this outcome found highly heterogeneous data and no statistically significant effect
15 Outcome is indirect (reported behavior). Population was indirect (PWID in the USA).
16 Analysis using continuous data for this outcome found no statistically significant effect
17 Tucker 2004 was conducted in Australia and included 145 IDUs (mean age 41 years, 60% male, had been injecting for 18 years on average). The intervention consisted of standard written educational materials regarding HCV plus a single-‐session, 30 minutes , individually tailored brief behavioural intervention. The control consisted of only the standard written educational materials
18 Tucker 2004: allocation was concealed, participants and data collectors were blinded, analysis was ITT, data for satisfaction was available for 76% of those randomized.
6
Question 5: Should peer education and mentoring vs. no peer education and mentoring be used in people who inject drugs?
Author(s): Elie Akl Date: 2012-‐04-‐27
Question: Should peer education and mentoring vs no peer education and mentoring be used in people who inject drugs?
Bibliography: Garfein 2007. Latka 2008. Systematic review by Amato et al.
Quality assessment No of patients Effect
Quality Importance No of
studies Design Risk of
bias Inconsistency Indirectness Imprecision Other considerations
Peer education and mentoring
No peer education and mentoring
Relative
(95% CI) Absolute Needle sharing behavior (assessed with: Self reported; surrogate for HCV and HIV infection)
21,2 randomised trials
serious3,4 no serious inconsistency
serious5,6 no serious imprecision
none -‐ -‐ OR 0.61 (0.44 to
0.85)
-‐7 ÅÅOO LOW
CRITICAL High risk sexual behavior (assessed with: Self reported (total acts with all partners); surrogate for HCV and HIV infection)
12 randomised trials
serious3 no serious inconsistency
serious6 serious8 none -‐ -‐ OR 0.90 (0.67 to
1.21)9
-‐7 ÅOOO VERY LOW
CRITICAL
Quality of life -‐ not measured
0 -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ -‐ IMPORTANT
1 Latka 2008 included 418 HCV-‐positive injection drug users, mean age 26.5 years, 77% male. The intervention consisted of peer mentoring. 6 sessions 2 hours in length, held twice weekly. The intervention delivered risk reduction information by training participants to mentor other injection drug users about reducing HCV transmission risks. The intervention was guided by social cognitive theory. The control consisted of video discussion. 6 sessions 2 hours in length, held twice weekly; participants watched a docudrama television series about injection drug users and then took part in a facilitated group discussion focusing on family, education, self-‐respect, relationships, violence, parenting, and employment; participants who sought information about risk reduction or health care were referred to a resource table
2 Garfein 2007 inlcuded 854 IDUs, HIV and HCV antibody negative, mean age 23.8 years, 66.5% male. The peer education intervention (PEI) consisted of peer education skills, intervention incorporating cognitive-‐behavioural skills-‐building; six 2-‐h sessions over a 3-‐week period. The intervention was centered on teaching participants how to educate peers about HIV and HCV risk reduction. Contents were HIV and HCV transmission through sex and injection drug use, peer education and skills-‐building activities about safer injection and sexual practices, with activities designed to increase negotiation skills with sex and injection partners. The control included arm received a video discussion intervention (VDI) comprising equivalent hours and sessions as the PEI. VDI participants watched hour-‐long films 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.
3 Garfein 2007: unclear whether randomization was concealed; only data analysts were clearly blinded; analysis was described as intention to treat; and lost to follow-‐up was at least 17%.
4 Latka 2008: unclear whether randomization was concealed; study was unblinded; analysis was intention to treat; retention rates were 66% and 80% at the 3-‐ and 6-‐month visits.
5 Latka 2008 provides indirect evidence: (1) population was HCV positive PWID in 5 US cities; (2) outcome was surrogate (participants’ self reported behaviors)
6 Garfein 2007 provides indirect evidence: (1) population was PWID in 5 US cities; (2) outcome was surrogate (participants’ self reported behaviors)
7 Absolute effects not calculable as rates and numbers of events not reported
8 CI includes both values suggesting harms and values suggesting benefits
9 Results consistent for all types of sexual acts (vaginal and anal) with different types of partners (main, non main other steady, sex trade partner) except for anal sex with casual/sex trade partner(s) (OR 3.15 (1.13 to 8.79))